conda-build - Man Page

For proper functioning, it is strongly recommended to install conda-build in the conda base environment. Not doing so may lead to problems.

With earlier conda / conda-build versions, it was possible to build packages in your own defined environment, e.g. my_build_env. This was partly driven by the general conda recommendation not to use the base env for normal work; see Conda Managing Environments for instance. However, conda-build is better viewed as part of the conda infrastructure, and not as a normal package. Hence, installing it in the base env makes more sense. More information: Must conda-build be installed in the base envt?

• An installation or update of conda-build (in fact, of any package) in the base environment needs to be run from an account with the proper permissions (i.e., the same permissions as were used to install conda and the base env in the first place via the Miniconda or Anaconda installers). For example, on Windows that might mean an account with administrator privileges.
• conda-verfiy is a useful package that can also be added to the base environment in order to remove some warnings generated when conda-build runs.
• For critical CI/CD projects, you might want to pin to an explicit (but recent) version of conda-build, i.e. only update to a newer version of conda-build and conda once they have been first verified "offline".

To install conda-build, in your terminal window or an Anaconda Prompt, run:

conda activate base
conda install conda-build

Keep your versions of conda and conda-build up to date to take advantage of bug fixes and new features.

To update conda and conda-build, in your terminal window or an Anaconda Prompt, run:

conda activate base
conda update conda
conda update conda-build

For release notes, see the conda-build GitHub page.

Conda-build supports standard conda channel behavior.

If the channel and package name are identical, it's possible to encounter a build problem if the short channel name is used.

Let's say your Anaconda.org username or an organization name is example. And suppose you created a package example, whose files' layout is similar to:

setup.py
conda/meta.yaml
example/

If your build depends on some other packages inside your channel, you will need to add -c example, however, the following code:

conda-build ./conda/ -c example

will fail with the following error message (the path will be different):

requests.exceptions.HTTPError: 404 Client Error: None for url:
file:///path/to/your/local/example/noarch/repodata.json
[...]
The remote server could not find the noarch directory for the requested channel with
url: file:///path/to/your/local/example/noarch/repodata.json
[...]
As of conda 4.3, a valid channel must contain a `noarch/repodata.json` and
associated `noarch/repodata.json.bz2` file, even if `noarch/repodata.json`
is empty. please request that the channel administrator create
`noarch/repodata.json` and associated `noarch/repodata.json.bz2` files.

This happens because conda-build will consider the directory ./example/ in your project as a channel. This is by design due to conda's CI servers, where the build path can be long, complicated, and not predictable prior to build.

There are several ways to resolve this issue.

1. Use the URL of the desired channel:

   conda-build ./conda/ -c https://conda.anaconda.org/example/

2. Run the build from inside the conda recipe directory:

   cd conda
   conda-build . -c example

3. Use the label specification workaround:

   conda-build ./conda/ -c example/label/main

   which technically is the same as -c example, since main is the default label, but now it won't mistakenly find a channel example/label/main on the local filesystem.

.
├── channeldata.json
├── linux-32
|   ├── repodata.json
│   └── package-0.0.0.tar.bz2
├── linux-64
|   ├── repodata.json
│   └── package-0.0.0.tar.bz2
├── win-64
|   ├── repodata.json
│   └── package-0.0.0.tar.bz2
├── win-32
|   ├── repodata.json
│   └── package-0.0.0.tar.bz2
├── osx-64
|   ├── repodata.json
│   └── package-0.0.0.tar.bz2
...

• channeldata.json contains metadata about the channel, including:

  • What subdirs the channel contains.
  • What packages exist in the channel and what subdirs they are in.
• Subdirs are associated with platforms. For example, the linux-64 subdir contains packages for linux-64 systems.
• repodata.json contains an index of the packages in a subdir. Each subdir will have its own repodata.
• Channels have packages as tarballs under corresponding subdirs.

{
  "channeldata_version": 1,
  "packages": {
    "super-fun-package": {
      "activate.d": false,
      "binary_prefix": false,
      "deactivate.d": false,
      "home": "https://github.com/Home/super-fun-package",
      "license": "BSD",
      "post_link": false,
      "pre_link": false,
      "pre_unlink": false,
      "reference_package": "win-64/super-fun-package-0.1.0-py310_0.tar.bz2",
      "run_exports": {},
      "subdirs": [
        "win-64"
      ],
      "summary": "A fun package! Open me up for rainbows",
      "text_prefix": false,
      "version": "0.1.0"
    },
    "subdirs": [
      "win-64",
      ...
    ]
}

{
  "packages": {
    "super-fun-package-0.1.0-py310_0.tar.bz2": {
      "build": "py37_0",
      "build_number": 0,
      "depends": [
        "some-depends"
      ],
      "license": "BSD",
      "md5": "a75683f8d9f5b58c19a8ec5d0b7f796e",
      "name": "super-fun-package",
      "sha256": "1fe3c3f4250e51886838e8e0287e39029d601b9f493ea05c37a2630a9fe5810f",
      "size": 3832,
      "subdir": "win-64",
      "timestamp": 1530731681870,
      "version": "0.1.0"
    },
    ...
  }

For each subdir:

• Look at all the packages that exist in the subdir.
• Generate a list of packages to add/update/remove.
• Remove all packages that need to be removed.

• For all packages that need to be added/updated:

  • Extract the package to access metadata, including full package name, file modification time (mtime), size, and index.json.
  • Aggregate package metadata to repodata collection.
• Apply repodata hotfixes (patches).
• Compute and save the reduced current_index.json index.

To build a local channel and put a package in it, follow the directions below:

1. Make the channel directory.

   $ mkdir local-channel
   $ cd local-channel

2. Now, download your favorite package.  We'll use SciPy in our example.  The next steps depend on your platform:

   1. Windows

      $ mkdir win-64
      $ curl -L https://anaconda.org/anaconda/scipy/1.9.1/download/win-64/scipy-1.9.1-py310h86744a3_0.tar.bz2 -o win-64\scipy-1.9.1-py310h86744a3_0.tar.bz2

   2. Linux

      1. Most Linux systems come with curl pre-installed.  Let's install it if you don't already have it.

         1. Check if you have curl:

            $ which curl

         2. If curl is not found, then install it:

            $ conda install curl

      2. Create a local copy of the package you want to include in your channel:

         $ mkdir linux-64
         $ curl -L https://anaconda.org/anaconda/scipy/1.9.1/download/linux-64/scipy-1.9.1-py310hd5efca6_0.tar.bz2 -o linux-64\scipy-1.9.1-py310hd5efca6_0.tar.bz2

   3. macOS, Intel chip

      $ mkdir osx-64
      $ curl -L https://anaconda.org/anaconda/scipy/1.9.1/download/osx-64/scipy-1.9.1-py310h09290a1_0.tar.bz2 -o osx-64/scipy-1.9.1-py310h09290a1_0.tar.bz2

   4. macOS, Apple chip

      $ mkdir osx-arm64
      $ curl -L https://anaconda.org/anaconda/scipy/1.9.1/download/osx-arm64/scipy-1.9.1-py310h20cbe94_0.tar.bz2 -o osx-arm64/scipy-1.9.1-py310h20cbe94_0.tar.bz2

   5. Other

      To find the latest SciPy on a platform not included in the list above, go to the Anaconda Packages file list for SciPy.

3. Run a conda index. This will generate both channeldata.json for the channel and repodata.json for the linux-64 and osx-64 subdirs, along with some other files:

   $ conda index .

4. Check your work by searching the channel:

   $ conda search -c file:/<path to>/local-channel scipy

   SciPy should be available in several channels, including local-channel.

Caching utilizes the existing repodata.json file if it exists. Indexing checks which files to update based on which files are new, removed, or changed since the last repodata.json was created. When a package is new or changed, its metadata is extracted and cached in the subdir to which the package belongs. The subfolder is the .cache folder. This folder has one file of interest: stat.json, which contains results from the stat command for each file. This is used for understanding when a file has changed and needs to be updated. In each of the other subfolders, the extracted metadata file for each package is saved as the original package name, plus a .json extension. Having these already extracted can save a lot of time in fully re-creating the index, should that be necessary.

An aside: one design goal of the .conda package format was to make indexing as fast as possible. To achieve this, the .conda format separates metadata from the actual package contents. Where the old .tar.bz2 container required extracting the entire package to obtain the metadata, the new package format allows extraction of metadata without touching the package contents. This allows indexing speed to be independent of the package size. Large .tar.bz2 packages can take a very long time to extract and index.

It is generally never necessary to manually alter the cache. To force an update/rescan of all cached packages, you can delete the .cache folder, or you can delete just the .cache/stat.json file. Ideally, you could remove only one package of interest from the cache, but that functionality does not currently exist.

Package repodata is bootstrapped from the index.json file within packages. Unfortunately, that metadata is not always correct. Sometimes a version bound needs to be added retroactively. The process of altering repodata from the values derived from package index.json files is called "hotfixing." Hotfixing is tricky, as it has the potential to break environments that have worked, but it is also sometimes necessary to fix environments that are known not to work.

On your own server, you're probably fine to run arbitrary python code that you have written to apply your patches. The advantage here is that the patches are generated on the fly every time the index is generated. That means that any new packages that have been added since the patch python file was last committed will be picked up and will have hotfixes applied to them where appropriate.

Anaconda applies hotfixes by providing a python file to conda index that has logic on how to alter metadata. Anaconda's repository of hotfixes is at https://github.com/AnacondaRecipes/repodata-hotfixes

Unfortunately, you can't always run your python code directly - other people who host your patches may not allow you to run code. What you can do instead is package the patches as .json files. These will clobber the entries in the repodata.json when they are applied.

This is the approach that conda-forge has to take, for example. Their patch creation code is here: https://github.com/conda-forge/conda-forge-repodata-patches-feedstock/tree/main/recipe

What that code does is to download the current repodata.json, then runs their python logic to generate the patch JSON file. Those patches are placed into a location where Anaconda's mirroring tools will find them and apply them to conda-forge's repodata.json at mirroring time.

The downside here is that this JSON file is only as new as the last time that the repodata-patches feedstock last generated a package. Any new packages that have been added to the index in the meantime will not have any hotfixes applied to them, because the hotfix JSON file does not know about those files.

The number of packages available is always growing. That means conda is always having to do more and more work. To slow down this growth, in conda 4.7, we added the ability to have alternate repodata.json files that may represent a subset of the normal repodata.json. One in particular is current_repodata.json, which represents:

1. the latest version of each package
2. any earlier versions of dependencies needed to make the latest versions satisfiable

current_repodata.json also keeps only one file type: .conda where it is available, and .tar.bz2 where only .tar.bz2 is available.

For Anaconda's defaults "main" channel, the current_repodata.json file is approximately 1/7 the size of repodata.json. This makes downloading the repodata faster, and it also makes loading the repodata into its python representation faster.

For those interested in how this is achieved, please refer to the code at https://github.com/conda/conda-build/blob/90a6de55d8b9e36fc4a8c471b566d356e07436c7/conda_build/index.py#L695-L737

To enable building conda packages, install and update conda and conda-build.

Building a conda package requires a recipe. A conda-build recipe is a flat directory that contains the following files:

• meta.yaml — A file that contains all the metadata in the recipe. Only package/name and package/version are required.
• build.sh — The script that installs the files for the package on macOS and Linux. It is executed using the bash command.
• bld.bat — The build script that installs the files for the package on Windows. It is executed using cmd.
• run_test.[py,pl,sh,bat] — An optional Python test file, a test script that runs automatically if it is part of the recipe.
• Optional patches that are applied to the source.
• Other resources that are not included in the source and cannot be generated by the build scripts. Examples are icon files, readme files and build notes.

Review Defining metadata (meta.yaml) to see a breakdown of the components of a recipe, including:

• Package name
• Package version
• Descriptive metadata
• Where to obtain source code
• How to test the package

TIP:

When you use the conda skeleton command, the first 3 files — meta.yaml, build.sh, and bld.bat — are automatically generated for you.

Conda-build performs the following steps:

1. Reads the metadata.
2. Downloads the source into a cache.
3. Extracts the source into the source directory.
4. Applies any patches.
5. Re-evaluates the metadata, if source is necessary to fill any metadata values.
6. Creates a build environment and then installs the build dependencies there.
7. Runs the build script. The current working directory is the source directory with environment variables set. The build script installs into the build environment.
8. Performs some necessary post-processing steps, such as adding a shebang and rpath.
9. Creates a conda package containing all the files in the build environment that are new from step 5, along with the necessary conda package metadata.

10. Tests the new conda package — if the recipe includes tests — by doing the following:

    • Deletes the build environment and source directory to ensure that the new conda package does not inadvertantly depend on artifacts not included in the package.
    • Creates a test environment with the package and its dependencies.
    • Runs the test scripts.

The archived conda-recipes repo, AnacondaRecipes aggregate repo, and conda-forge feedstocks repo contain example recipes for many conda packages.

CAUTION:

The conda skeleton command can help to make skeleton recipes for common repositories, such as PyPI.

Let's take a closer look at how conda-build uses a recipe to create a package.

When you build a conda package, conda-build renders the package by reading a template in the meta.yaml. See Templating with Jinja.

Templates are filled in using your conda build config, which shows the matrix of things to build against. The conda build config determines how many builds it has to do. For example, defining a conda_build_config.yaml of the form and filling it defines a matrix of 4 packages to build:

foo:
  - 1.0
  - 2.0
bar:
  - 1.2.0
  - 1.4.0

After this, conda-build determines what the outputs will be. For example, if your conda build config indicates that you want 2 different versions of Python, conda-build will show you the rendering for each Python version.

To build the package, conda-build will make an environment for you and install all of the build and run dependencies in that environment. Conda-build will indicate where you can successfully build the package. The prefix will take the form:

<file path to conda>/conda-bld/<package name and string>/h_env_placeholder…

Conda-build downloads your package source and then builds the conda package in the context of the build environment. For example, you may direct it to download from a Git repo or pull down a tarball from another source. See the Source section for more information.

What conda-build puts into a package depends on what you put into the build, host, or run sections. See the Requirements section for more information. Conda-build will use this information to identify dependencies to link to and identify the run requirements for the package. This allows conda-build to understand what is needed to install the package.

Once the content is downloaded, conda-build runs the build step. See the Build section for more information. The build step runs a script. It can be one that you provided. See the Script section for more information on this topic.

If you do not define the script section, then you can create a build.sh or a bld.bat file to be run.

The build environment is created in a placeholder prefix. When the package is bundled, the prefix is set to a "dummy" prefix. Once conda is ready to install the package, it rewrites the dummy prefix with the final one.

Once a package is built, conda-build has the ability to test it. To do this, it creates another environment and installs the conda package. The form of this prefix is:

<file path to conda>/conda-bld/<package name + string>/_test_env_placeholder…

At this point, conda-build has all of the information from meta.yaml about what its runtime dependencies are, so those dependencies are installed as well. This generates a test runner script with a reference to the testing meta.yaml that is created. See the Test section for more information. That file is run for testing.

After the package is built and tested, conda-build cleans up the environments created during prior steps and outputs the metadata. The recipe for the package is also added in the output metadata. The metadata directory is at the top level of the package contents in the info directory. The metadata contains information about the dependencies of the package and a list of where all of the files in the package go when it is installed. Conda reads that metadata when it needs to install.

Running conda install causes conda to:

1. Reach out to the repodata containing the dependencies for the package(s) you are installing.
2. Determine the correct dependencies.
3. Install a list of additional packages determined by those dependencies.
4. For each dependency package being installed: #. Unpack the tarball to look at the information contained within. #. Verify the file based on metadata in the package. #. Go through each file in the package and put it in the right location.

For additional information on conda install, please visit the conda documentation deep dive page on that topic.

To facilitate communication and documentation, conda observes the package naming conventions listed below:

Package name

The name of a package, without any reference to a particular version. Conda package names are normalized and they may contain only lowercase alpha characters, numeric digits, underscores, hyphens, or dots. In usage documentation, these are referred to by package_name.

Package version

A version number or string, often similar to X.Y or X.Y.Z, but it may take other forms as well.

Build string

An arbitrary string that identifies a particular build of a package for conda. It may contain suggestive mnemonics, but these are subject to change, and you should not rely on it or try to parse it for any specific information.

Canonical name

The package name, version, and build string joined together by hyphens: name-version-buildstring. In usage documentation, these are referred to by canonical_name.

Filename

Conda package filenames are canonical names, plus the suffix .tar.bz2 or .conda.

The following figure compares a canonical name to a filename:

Conda supports both .conda and .tar.bz2 package extensions. The .conda format is generally smaller and more efficient than .tar.bz2 packages. Read our blog post about it to learn more.

The build string is created as the package is built. Things that contribute to it are the variants specified either by the command line or the configuration from the conda_build_config.yaml, and the build number in the recipe. If there are no variants, then the build string is the build number that is specified in the recipe.

A package name together with a package version — which may be partial or absent — joined by an equal sign.

Examples:

• python=2.7.3
• python=2.7
• python

In usage documentation, these are referred to by package_spec.

• A package is anything you install using your package manager.

• A "conda package" is a compressed tarball that contains

  • the module to be installed
  • information on how to install the package
• You can use conda-build to build a conda package.

• Channels contain packages.
• They conform to a standard structure and contain an index of available packages.

• An index of the available packages can be generated by running:

  $ conda index <path to channel>

• conda is able to install from channels and uses the indexes in the channel to solve for requirements and dependencies.

Before starting the tutorials, consider reviewing how to install and update conda-build and conda-build concepts.

You may also find our resources collection helpful.

Before starting the tutorials, you need to install:

• Miniconda or Anaconda
• conda-build
• Git

The most straightforward way to do this is to install Miniconda or Anaconda, which contain conda, and then use conda to install conda-build and Git. Make sure you install these packages into your base environment.:

conda install -n base conda-build git

For more information on installing and updating conda-build, see Installing and updating conda-build.

Have an idea for a tutorial? You can submit your suggestions to Anaconda by emailing us at documentation@anaconda.com.

To create your own tutorials, follow the writing style guide and tutorial template.

Before starting the tutorials, review the Getting started guide.

This tutorial describes how to use conda-build to create conda packages on Windows, macOS, and Linux using the examples of SEP and GDAL. Additional Windows-specific instructions are provided in the Toolkit section.

The final built packages from this tutorial are available on anaconda.org (formerly known as Anaconda Cloud):

• SEP
• GDAL

This tutorial also describes writing recipes. You can see the final SEP recipe and the GDAL recipe on GitHub in the conda-build documentation repository.

This tutorial is for Windows, macOS, and Linux users who wish to build more complex conda packages. This tutorial will involve building scientific packages, which require compilers for several different Python versions.

Before you start, make sure you have installed:

• Conda
• Conda-build
• Any compilers you want

In the standard practices of the conda developers, conda packages for different versions of Python are each built with their own version of Visual Studio (VS):

• Python 2.7 packages with Visual Studio 2008
• Python 3.4 packages with VS 2010
• Python 3.5 packages with VS 2015, (default) 2017
• Python 3.6 packages with VS 2015, (default) 2017
• Python 3.7 packages with VS 2015, (default) 2017

Using these versions of VS to build packages for each of these versions of Python is also the practice used for the official python.org builds of Python. Currently VS 2008 and VS 2010 are available only through resellers, while VS 2015 and VS 2017 can be purchased online from Microsoft. Note there is also a community edition of VS 2015 and VS 2017 which may be used.

There are free alternatives available for each version of the VS compilers:

• Instead of VS 2008, it is often possible to substitute the free Microsoft Visual C++ Compiler for Python 2.7.
• Instead of VS 2010, it is often possible to substitute the free Microsoft Windows SDK for Windows 7 and .NET Framework 4.
• Make sure that you also install VS 2010 Service Pack 1 (SP1).
• Due to a bug in the VS 2010 SP1 installer, the compiler tools may be removed during installation of VS 2010 SP1. They can be restored as described in Microsoft Visual C++ 2010 Service Pack 1 Compiler Update for the Windows SDK 7.1.
• Visual Studio 2015 has a full-featured, free Community edition for academic research, open source projects, and certain other use cases.

The MS Visual C++ Compiler for Python 2.7 and the Microsoft Windows SDK for Windows 7 and .NET Framework 4 are both reasonably well tested. Conda-build is carefully tested to support these configurations, but there are known issues with the CMake build tool and these free VS 2008 and 2010 alternatives. In these cases, you should prefer the "NMake Makefile" generator, rather than a Visual Studio solution generator.

You can use any recent version of Windows. These examples were built on Windows 10.

Some environments initially lack tools such as patch or Git that may be needed for some build workflows.

On Windows, these can be installed with conda using the following command:

$ conda install git m2-patch

On macOS and Linux, replace m2-patch with patch.

Conda recipes are typically built with a trial-and-error method. Sometimes, the first attempt to build a package will fail with compiler or linker errors, often caused by missing dependencies. The person writing the recipe then examines these errors and modifies the recipe to include the missing dependencies, usually as part of the meta.yaml file. Then the recipe writer attempts the build again and, after a few of these cycles of trial and error, the package builds successfully.

Miniconda2 and Miniconda3 can each build packages for either Python 2 or Python 3 simply by specifying the version you want. Miniconda2 includes only Python 2 and Miniconda3 includes only Python 3.

Installing only one makes it easier to keep track of the builds, but it is possible to have both installed on the same system at the same time. If you have both installed, use the where command on Windows, or the which command on macOS or Linux, to see which version comes first on PATH, since this is the one you will be using:

$ where python

To build a package for a Python version other than the one in your Miniconda installation, use the --python option in the conda-build command. For example, to build a Python 3.5 package with Miniconda2:

$ conda-build recipeDirectory --python=3.5

NOTE:

Replace recipeDirectory with the name and path of your recipe directory.

After the build, if the recipe directory contains a test file. This test file is named run_test.bat on Windows, run_test.sh on macOS or Linux, or run_test.py on any platform. The file runs to test the package and any errors that surface are reported. After seeing "check the output," you can also test if this package was built by using the command:

$ conda build --test <path to package>.tar.bz2

NOTE:

Use the "test" section of the meta.yaml file to move data files from the recipe directory to the test directory when the test is run.

The SEP documentation states that SEP runs on Python 2 and 3, and it depends only on NumPy. Searching for SEP on PyPI shows that there is already a PyPI package for SEP.

Because a PyPI package for SEP already exists, the conda skeleton command can make a skeleton or outline of a conda recipe based on the PyPI package. The recipe outline can then be completed manually and conda can build a conda package from the completed recipe.

If you have not already done so, install the appropriate version of Visual Studio:

• For Python 3 — Visual Studio 2017:

  1. Choose the "Custom install" option.
  2. Under Programming Languages, choose to install Visual C++.

• For Python 2 — Visual Studio 2008:

  1. Choose the "Custom install" option.
  2. Choose to install X64 Compilers and Tools. Install Service Pack 1.

1. Run the skeleton command:

   $ conda skeleton pypi sep

   The skeleton command installs into a newly-created directory called sep.

2. Go to the sep directory to view the files:

   $ cd sep

   You will see that one skeleton file has been created: meta.yaml

For this package, bld.bat and build.sh need no changes. You need to edit the meta.yaml file to add the dependency on NumPy and add an optional test for the built package by importing it. For more information about what can be specified in meta.yaml, see Defining metadata (meta.yaml).

1. In the requirements section of the meta.yaml file, add a line that adds NumPy as a requirement to build the package.
2. Add a second line to list NumPy as a requirement to run the package.
3. Set the NumPy version to the letters x.x.
4. Make sure the new line is aligned with - python on the line above it, so as to ensure proper yaml format.

EXAMPLE:

requirements:
  host:
    - python
    - numpy     x.x

  run:
    - python
    - numpy     x.x

Notice that there are two types of requirements, host and run (build is another valid parameter, but is not shown in this example). host represents packages that need to be specific to the target platform when the target platform is not necessarily the same as the native build platform. run represents the dependencies that should be installed when the package is installed.

NOTE:

Adding this optional test will test the package at the end of the build by making sure that the Python statement import sep runs successfully:

1.

Add - sep, checking to be sure that the indentation is consistent with the rest of the file:

test:
  # Python imports
  imports:
    - sep

Build the package using the recipe you just created:

$ conda build sep

1. Check the output to make sure that the build completed successfully. The output contains the location of the final package file and a command to upload the package to Anaconda Cloud. The output will look something like:

   # Automatic uploading is disabled
   # If you want to upload package(s) to anaconda.org later, type:
   anaconda upload /Users/builder/miniconda3/conda-bld/osx-64/sep-1.0.3-np111py36_0.tar.bz2
   # To have conda build upload to anaconda.org automatically, use
   # $ conda config --set anaconda_upload yes
   anaconda_upload is not set.  Not uploading wheels: []
   ####################################################################################
   Resource usage summary:
   Total time: 0:00:56.4
   CPU usage: sys=0:00:00.7, user=0:00:07.0
   Maximum memory usage observed: 220.1M
   Total disk usage observed (not including envs): 3.9K
   ####################################################################################
   Source and build intermediates have been left in /Users/builder/miniconda3/conda-bld.
   There are currently 437 accumulated.
   To remove them, you can run the ```conda build purge``` command

2. If there are any linker or compiler errors, modify the recipe and build again.

To begin, install Anaconda or Miniconda and conda-build. If you are using a Windows machine, also use conda to install Git and m2-patch.

$ conda install git
$ conda install m2-patch

Because GDAL includes C and C++, building it on Windows requires Visual Studio. This procedure describes how to build a package with Python 2 or Python 3. Follow the instructions for the version with which you want to build.

To build a GDAL package:

1. Install Visual Studio:

   • For Python 3, install Visual Studio 2017:

     • Choose "Custom install".
     • Under "Programming Languages", select workloads that come from Visual Studio so that you can choose the Desktop Development with C++ and Universal Platform C.

   • For Python 2, install Visual Studio 2008:

     • Choose "Custom install".
     • Choose to install X64 Compilers and Tools.
     • Install Visual Studio 2008 Service Pack 1.

2. Install Git. Because the GDAL package sources are retrieved from GitHub for the build, you must install Git:

   $ conda install git m2-patch conda-build

3. Get gdal-feedstock. For the purpose of this tutorial, we will be using a recipe from Anaconda:

   $ git clone https://github.com/AnacondaRecipes/gdal-feedstock.git

4. Use conda-build to build the gdal-feedstock:

   $ conda build gdal-feedstock

5. Check the output to make sure that the build completed successfully. The output also contains the location of the final package file and a command to upload the package to Cloud. For this package in particular, there should be two packages outputted: libgdal and GDAL.
6. In case of any linker or compiler errors, modify the recipe and run it again.

Let's take a closer look at what's happening inside the gdal-feedstock, specifically in the meta.yaml file.

The first interesting bit happens under source in the patches section:

patches:
  # BUILT_AS_DYNAMIC_LIB.
  - 0001-windowshdf5.patch
  # Use multiple cores on Windows.
  - 0002-multiprocessor.patch
  # disable 12 bit jpeg on Windows as we aren't using internal jpeg
  - 0003-disable_jpeg12.patch

This section is basically saying "when this package is being built on a Windows platform, apply the following patch files". Notice that the patch files are in the patches directory of the recipe. These patches will only be applied to Windows since the # [win] selector is applied to each of the patch entries. For more about selectors, see Preprocessing selectors.

In the requirements section, notice how there are both a build and host set of requirements (run is another valid parameter, but is not shown in this example). For this recipe, all the compilers required to build the package are listed in the build requirements. Normally, this section will list out packages required to build the package. GDAL requires CMake on Windows, as well as C compilers. Notice that the C compilers are pulled into the recipe using the syntax {{ compiler('c') }}. Since conda-build 3, conda-build defines a jinja2 function compiler() to specify compiler packages dynamically. So, using the compiler('c') function in a conda recipe will pull in the correct compiler for any build platform. For more information about compilers with conda-build see compiler-tools.

Also note that the compilers used by conda-build can be specified using a conda_build_config.yaml. For more information about how to do that, see Using your customized compiler package with conda-build 3.

Notice that this package has an outputs section. This section is a list of packages to output as a result of building this package. In this case, the packages libgdal and GDAL will be built. Similar to a normal recipe, the outputs can have build scripts, tests scripts, and requirements specified. For more information on how outputs work, see the Outputs section.

Now, let's try to build GDAL against some build matrix. We will specify building against Python 3.7 and 3.5 using a conda-build config. Add the following to your conda_build_config.yaml:

python:
   - 3.7
   - 3.5

Now you can build GDAL using conda-build with the command:

$ conda build gdal-feedstock

Or explicitly set the location of the conda-build variant matrix:

$ conda build gdal-feedstock --variant-config-file conda_build_config.yaml

If you want to know more about build variants and conda_build_config.yaml, including how to specify a config file and what can go into it, take a look at Creating conda-build variant config files.

TIP:

We recommend Grayskull, a newer alternative to conda-skeleton, to generate conda recipes for PyPI packages. Please check out their project page on GitHub for more information.

This tutorial describes how to quickly build a conda package for a Python module that is already available on PyPI.

In the first procedure, building a simple package, you build a package that can be installed in any conda environment of the same Python version as your root environment. The remaining optional procedures describe how to build packages for other Python versions and other architectures, as well as how to upload packages to your Anaconda.org account.

NOTE:

This tutorial is for Windows, macOS, and Linux users who wish to build a conda package from a PyPI package. No prior knowledge of conda-build or conda recipes is required.

Before you start, check the Prerequisites.

The conda skeleton command picks up the PyPI package metadata and prepares the conda-build recipe. The final step is to build the package itself and install it into your conda environment.

It is easy to build a skeleton recipe for any Python package that is hosted on PyPI, the official third-party software repository for the Python programming language.

In this section you are going to use conda skeleton to generate a conda recipe, which informs conda-build about where the source files are located and how to build and install the package.

You'll be working with a package named Click that is hosted on PyPI. Click is a tool for exposing Python functions to create command line interfaces.

First, in your user home directory, run the conda skeleton command:

conda skeleton pypi click

The two arguments to conda skeleton are the hosting location, in this case pypi, and the name of the package.

This creates a directory named click and creates one skeleton file in that directory: meta.yaml. Many other files can be added there as necessary, such as build.sh and bld.bat, test scripts, or anything else you need to build your software. For simple, pure-Python recipes, these extra files are unnecessary and the build/script section in the meta.yaml is sufficient. Use the ls command on macOS or Linux or the dir command on Windows to verify that this file has been created. The meta.yaml file has been populated with information from the PyPI metadata and in many cases will not need to be edited.

Files in the folder with meta.yaml are collectively referred to as the "conda-build recipe":

• meta.yaml---Contains all the metadata in the recipe. Only the package name and package version sections are required---everything else is optional.
• bld.bat---Windows commands to build the package.
• build.sh---macOS and Linux commands to build the package.

Now that you have the conda-build recipe ready, you can use conda-build to create the package:

conda-build click

When conda-build is finished, it displays the exact path and filename of the conda package. See Troubleshooting a sample issue if the conda-build command fails.

Windows example file path:

C:\Users\jsmith\miniconda\conda-bld\win-64\click-7.0-py37_0.tar.bz2

macOS example file path:

/Users/jsmith/miniconda/conda-bld/osx-64/click-7.0-py37_0.tar.bz2

Linux example file path:

/home/jsmith/miniconda/conda-bld/linux-64/click-7.0-py37_0.tar.bz2

NOTE:

Your path and filename will vary depending on your installation and operating system. Save the path and filename information for the next step.

Now you can install your newly built package in your conda environment by using the use-local flag:

conda install --use-local click

Notice that Click is coming from the local conda-build channel.

(click) 0561:~ jsmith$ conda list
# packages in environment at /Users/Jsmith/miniconda/envs/click:
# Name                    Version                   Build  Channel
ca-certificates           2019.1.23                     0
certifi                   2019.3.9                 py37_0
click                     7.0                      py37_0    local

Now verify that Click installed successfully:

conda list

Scroll through the list until you find Click.

At this point you now have a conda package for Click that can be installed in any conda environment of the same Python version as your root environment. The remaining optional sections show you how to make packages for other Python versions and other architectures and how to upload them to your Anaconda.org account.

By default, conda-build creates packages for the version of Python installed in the root environment. To build packages for other versions of Python, you use the --python flag followed by a version. For example, to explicitly build a version of the Click package for Python 2.7, use:

conda-build --python 2.7 click

Notice that the file printed at the end of the conda-build output has changed to reflect the requested version of Python. conda install will look in the package directory for the file that matches your current Python version.

Windows example file path:

C:\Users\jsmith\Miniconda\conda-bld\win-64\click-7.0-py27_0.tar.bz2

macOS example file path:

/Users/jsmith/miniconda/conda-bld/osx-64/click-7.0-py27_0.tar.bz2

Linux example file path:

/home/jsmith/miniconda/conda-bld/linux-64/click-7.0-py27_0.tar.bz2

NOTE:

Your path and filename will vary depending on your installation and operating system. Save the path and filename information for the next task.

Now that you have built a package for your current platform with conda-build, you can convert it for use on other platforms with the conda convert command. This works only for pure Python packages where there is no compiled code. Conda convert does nothing to change compiled code, it only adapts file paths to take advantage of the fact that Python scripts are mostly platform independent. Conda convert accepts a platform specifier from this and a platform specifier from this list:

• osx-64.
• linux-32.
• linux-64.
• win-32.
• win-64.
• all.

In the output directory, 1 folder will be created for each of the 1 or more platforms you chose and each folder will contain a .tar.bz2 package file for that platform.

Windows:

conda convert -f --platform all C:\Users\jsmith\miniconda\conda-bld\win-64\click-7.0-py37_0.tar.bz2
-o outputdir\

macOS and Linux:

conda convert --platform all /home/jsmith/miniconda/conda-bld/linux-64/click-7.0-py37_0.tar.bz2
-o outputdir/

NOTE:

Change your path and filename to the exact path and filename you saved in Optional---Building for a different Python version.

To use these packages, you need to transfer them to other computers and place them in the correct conda-bld/$ARCH directory for the platform, where $ARCH can be osx-64, linux-32, linux-64, win-32, or win-64.

A simpler way is to upload all of the bz2 files to Anaconda.org as described in the next task.

If you find yourself needing to use conda convert, you might instead prefer to change your recipe to make your package a "noarch" package. Noarch packages run anywhere and do not require conda convert. Some of the ecosystem tools don't yet support noarch packages but, for the most part, noarch packages are a better way to go.

Anaconda.org is a repository for public or private packages. Uploading to Anaconda.org allows you to easily install your package in any environment with just the conda install command, rather than manually copying or moving the tarball file from one location to another. You can choose to make your files public or private. For more information about Anaconda.org, see the Anaconda.org documentation.

1. Create a free Anaconda.org account and record your new Anaconda.org username and password.
2. Run conda install anaconda-client and enter your Anaconda.org username and password.
3. Log into your Anaconda.org account from your terminal with the command anaconda login.

Now you can upload the new local packages to Anaconda.org.

Windows:

anaconda upload C:\Users\jsmith\miniconda\conda-bld\win-64\click-7.0-py37_0.tar.bz2

macOS and Linux:

anaconda upload /home/jsmith/miniconda/conda-bld/linux-64/click-7.0-py37_0.tar.bz2

NOTE:

Change your path and filename to the exact path and filename you saved in Optional---Building for a different Python version. Your path and filename will vary depending on your installation and operating system.

If you created packages for multiple versions of Python or used conda convert to make packages for each supported architecture, you must use the anaconda upload command to upload each one. It is considered best practice to create packages for Python versions 2.7, 3.4, and 3.5 along with all of the architectures.

TIP:

You can log out of your Anaconda.org account with the command:

anaconda logout

Conda-build may produce the error message "Build Package missing."

To explore this error:

1. Create a conda skeleton package for skyfield. The conda skeleton command is:

   conda skeleton pypi skyfield

   This command creates the skyfield conda-build recipe.

2. Run conda-build skyfield and observe that it fails with the following output:

   Removing old build environment
   Removing old work directory
   BUILD START: skyfield-0.8-py35_0
   Using Anaconda Cloud api site https://api.anaconda.org
   Fetching package metadata: ......
   Solving package specifications: .
   Error:  Package missing in current osx-64 channels:
     - sgp4 >=1.4

In this example, the conda recipe requires sgp4 for the skyfield package. The skyfield recipe was created by conda skeleton. This error means that conda could not find the sgp4 package and install it.

Since many PyPI packages depend on other PyPI packages to build or run, the solution is sometimes as simple as using conda skeleton to create a conda recipe for the missing package and then building it:

conda skeleton sgp4
conda build sgp4

You may also try using the --recursive flag with conda skeleton, but this makes conda recipes for all required packages, even those that are already available to conda install.

For more options, see the full conda skeleton command documentation.

This tutorial describes how to build a conda package for Click by writing the required files in the conda-build recipe.

This tutorial is for Windows, macOS, and Linux users who wish to generate a conda package by writing the necessary files. Prior knowledge of conda-build and conda recipes is helpful.

• Check the Prerequisites.
• You should have already completed Building conda packages with conda skeleton.

1. Make a new directory for this tutorial named click, and then change to the new directory:

   mkdir click
   cd click

2. To create a new meta.yaml file, open your favorite editor. Create a new text file and insert the information shown below. A blank sample meta.yaml follows the table to make it easier to match up the information.

   NOTE:

   To allow correct sorting and comparison, specify version as a string.

   name	click
   version	"7.0" (or latest from https://github.com/pallets/click/releases)
   git_rev	6.7 (or latest from https://github.com/pallets/click/releases)
   git_url	https://github.com/pallets/click.git
   imports	click
   home	https://github.com/pallets/click
   license	BSD

   package:
     name:
     version:
   
   source:
     git_rev:
     git_url:
   
   requirements:
     build:
       - python
       - setuptools
   
     run:
       - python
   
   test:
     imports:
       -
   
   about:
     home:

3. Save the file in the same click directory as meta.yaml. It should match this meta.yaml file.

Besides meta.yaml, 2 files are required for a build:

• build.sh---Shell script for macOS and Linux.
• bld.bat---Batch file for Windows.

These 2 build files contain all the variables, such as for 32-bit or 64-bit architecture---the ARCH variable---and the build environment prefix---PREFIX. The 2 files build.sh and bld.bat must be in the same directory as your meta.yaml file.

This tutorial describes how to make both build.sh and bld.bat so that other users can build the appropriate package for their architecture.

1. Open a text editor and create a new file named bld.bat. Type the text exactly as shown:

   "%PYTHON%" setup.py install
   if errorlevel 1 exit 1

   NOTE:

   In bld.bat, the best practice is to to add if errorlevel 1 exit 1 after every command so that if the command fails, the build fails.

2. Save this new file bld.bat to the same directory where you put your meta.yaml file.

3. Open a text editor and create a new file named build.sh. Enter the text exactly as shown:

   $PYTHON setup.py install     # Python command to install the script.

4. Save your new build.sh file to the same directory where you put the meta.yaml file.

You can run build.sh with bash -x -e. The -x makes it echo each command that is run, and the -e makes it exit whenever a command in the script returns nonzero exit status. If you need to revert this in the script, use the set command in build.sh.

Now that you have your 3 new build files ready, you are ready to create your new package with conda-build and install the package on your local computer.

1. Run conda-build:

   conda-build click

   If you are already in the click folder, you can type conda build ..

   When conda-build is finished, it displays the package filename and location. In this case the file is saved to:

   ~/anaconda/conda-bld/linux-64/click-7.0-py37_0.tar.bz2

   NOTE:

   Save this path and file information for the next task. The exact path and filename vary depending on your operating system and whether you are using Anaconda or Miniconda. The conda-build command tells you the exact path and filename.

2. Install your newly built program on your local computer by using the use-local flag:

   conda install --use-local click

   If there are no error messages, Click installed successfully.

   NOTE:

   Explicitly installing a local package bypasses the dependency resolver, as such the package's run dependencies will not be evaluated. See conda install --help or the install command reference page for more info.

Now that you have built a package for your current platform with conda-build, you can convert it for use on other platforms by using the 2 build files, build.sh and bld.bat.

Use the conda convert command with a platform specifier from the list:

• osx-64.
• linux-32.
• linux-64.
• win-32.
• win-64.
• all.

EXAMPLE: Using the platform specifier all:

conda convert --platform all ~/anaconda/conda-bld/linux-64/click-7.0-py37_0.tar.bz2 -o outputdir/

NOTE:

Change your path and filename to the path and filename you saved in Building and installing.

You can use PyPI or another repository instead of GitHub. There is little difference to conda-build between building from Git versus building from a tarball on a repository like PyPI. Because the same source is hosted on PyPI and GitHub, you can easily find a script on PyPI instead of GitHub.

Replace this source section:

git_rev: v0.6.7
git_url: https://github.com/pallets/click.git

With the following:

url: https://files.pythonhosted.org/packages/f8/5c/f60e9d8a1e77005f664b76ff8aeaee5bc05d0a91798afd7f53fc998dbc47/Click-7.0.tar.gz
sha256: 5b94b49521f6456670fdb30cd82a4eca9412788a93fa6dd6df72c94d5a8ff2d7

NOTE:

The url and sha256 are found on the PyPI Click page.

After converting your files for use on other platforms, you may choose to upload your files to Anaconda.org, formerly known as binstar.org. It only takes a minute to do if you have a free Anaconda.org account.

1. If you have not done so already, open a free Anaconda.org account and record your new user name and password.
2. Run the command conda install anaconda-client, and then enter your Anaconda.org username and password.

3. Log into your Anaconda.org account with the command:

   anaconda login

4. Upload your package to Anaconda.org:

   anaconda upload ~/miniconda/conda-bld/linux-64/click-7.0-py37_0.tar.bz2

   NOTE:

   Change your path and filename to the path and filename you saved in Building and installing.

   TIP:

   To save time, you can set conda to always upload a successful build to Anaconda.org with the command: conda config --set anaconda_upload yes.

• For more information about all the possible values that can go into the meta.yaml file, see Defining metadata (meta.yaml).

This tutorial describes how to quickly build an R-language package on macOS for an R module that is already available on CRAN.

You will build a simple package that can be installed in any conda environment of the same R version as your root environment. The tutorial will also tell you how to build the dependencies that may arise while building the package.

This tutorial is for macOS users who wish to build an R-language package from CRAN. No prior knowledge of conda-build or conda recipes is required.

Before you start, check the prerequisites.

The conda skeleton command picks up the CRAN package metadata and prepares the conda-build recipe. The final step is to build the package itself and install it into your conda environment.

It is easy to build a skeleton recipe for any R package that is hosted on CRAN. In this section you are going to use conda skeleton to generate a conda recipe, which informs conda-build about where the source files are located and how to build and install the package.

You'll be working with a package that is hosted on CRAN named fansi, a tool that accounts for the effects of ANSI text formatting control sequences.

First, in your user home directory, run the conda skeleton command:

conda skeleton cran fansi

The two arguments to conda skeleton are the type of hosting location, in this case cran, and the name of the package.

This creates a directory named r-fansi and creates 1 skeleton file in that directory: meta.yaml. Many other files can be added there as necessary, such as build.sh and bld.bat, test scripts, or anything else you need to build your software. Use the ls command to verify that this file has been created. The meta.yaml file has been populated with information from the CRAN metadata and in many cases will not need to be edited.

Files in the folder with meta.yaml are collectively referred to as the "conda-build recipe":

• meta.yaml---Contains all the metadata in the recipe. The package name and package version sections are required---everything else is optional.
• bld.bat---Windows commands to build the package.
• build.sh---macOS and Linux commands to build the package.

Now that you have the conda-build recipe ready, you can use conda-build to create the package:

conda-build r-fansi

When conda-build is finished, it displays the exact path and filename of the conda package. See Troubleshooting a sample issue if the conda-build command fails. If you receive an error with SDK on macOS, review our resources for macOS and SDK.

Example file path:

/Users/jsmith/anaconda3/conda-bld/osx-64/r-fansi-0.4.0-r353h46e59ec_0.tar.bz2

NOTE:

Your path and filename will vary depending on your installation and operating system. Save the path and filename information for the next step.

Now you can install your newly built package in your conda environment by using the use-local flag:

conda install --use-local r-fansi

Now verify that fansi installed successfully:

conda list

Scroll through the list until you find r-fansi.

Notice that fansi is coming from the local conda-build channel.

(base) 0561:~ jsmith$ conda list
# packages in environment at /Users/Jsmith/anaconda3:
# Name                    Version                   Build  Channel
qtpy                      1.5.0                    py37_0
r-base                    3.5.1                h539fb6c_1
r-fansi                   0.4.0            r353h46e59ec_0    local

The version of R will be what you have in your base environment.

See Optional---Building for a different R version to set your own R version.

At this point you now have a conda package for fansi that can be installed in any conda environment of its R version.

The fansi package was a simple one that didn’t have dependencies. To build an R package with dependencies, let’s look at the example of janitor. Janitor is a package hosted on CRAN that is used for examining and cleaning up data.

To begin building it, type:

conda skeleton cran janitor

This creates a directory named r-janitor and creates one skeleton file in that directory: meta.yaml. Many other files can be added there as necessary, such as build.sh and bld.bat, test scripts, or anything else you need to build your software. Use the ls command to verify that this file has been created. The meta.yaml file has been populated with information from the CRAN metadata and in many cases will not need to be edited.

Now that you have the conda-build recipe ready, you can use conda-build to create the package:

conda-build r-janitor

What may happen at this point is that you will have dependencies of this package that do not exist as conda packages yet. They need to be turned into conda packages. Use conda skeleton to recursively build out recipes for the packages that it depends on:

conda skeleton cran janitor --recursive

You can manually build each package individually by typing:

conda-build package-name

NOTE:

Replace "package-name" with the name of each package.

Once all of the package dependencies are resolved, you can build the R package by using:

conda-build .

Now you can install your newly built package in your conda environment by using the use-local flag:

conda install --use-local r-janitor

The remaining optional sections show you how to make R packages for other R versions and other architectures and how to upload them to your Anaconda.org account.

By default, conda-build creates packages for the version of R installed in the root environment. To build packages for other versions of R, you use the --R flag followed by a version.

For example, to explicitly build a version of the fansi package for R 3.6.1, use:

conda-build --R 3.6.1 r-fansi

Notice that the file printed at the end of the conda-build output has changed to reflect the requested version of R. Conda install will look in the package directory for the file that matches your current R version.

Example file path:

/Users/jsmith/anaconda3/conda-bld/osx-64/r-fansi-0.4.0-r353h46e59ec_0.tar.bz2

NOTE:

Your path and filename will vary depending on your installation and operating system. Save the path and filename information for the next task.

Anaconda.org is a repository for public or private packages. Uploading to Anaconda.org allows you to easily install your package in any environment with just the conda install command, rather than manually copying or moving the tarball file from one location to another. You can choose to make your files public or private.

For more information about Anaconda.org, see the Anaconda.org documentation.

1. Create a free Anaconda.org account and record your new Anaconda.org username and password.
2. Run conda install anaconda-client and enter your Anaconda.org username and password.
3. Log into your Anaconda.org account from your terminal with the command anaconda login.

Now you can upload the new local packages to Anaconda.org.

anaconda upload /Users/jsmith/anaconda3/conda-bld/osx-64/r-fansi-0.4.0-r353h46e59ec_0.tar.bz2

NOTE:

Change your path and filename to the exact path and filename you saved in Optional---Building for a different R version. Your path and filename will vary depending on your installation and operating system. If you created packages for multiple versions of R, you must use the anaconda upload command to upload each one.

TIP:

If you want to always automatically upload a successful build to Anaconda.org, run: conda config --set anaconda_upload yes

You can log out of your Anaconda.org account with the command:

anaconda logout

For more options, see the full conda skeleton command documentation.

Review recipe concepts for more information about conda-build recipes.

Conda offers you the flexibility of being able to build things that are not Python related. The first 2 sample recipes, boost and libtiff, are examples of non-Python libraries, meaning they do not require Python to run or build.

• boost is an example of a popular programming library and illustrates the use of selectors in a recipe.
• libtiff is another example of a compiled library, which shows how conda can apply patches to source directories before building the package.
• msgpack, blosc, and cytoolz are examples of Python libraries with extensions.
• toolz, sympy, six, and gensim are examples of Python-only libraries.

gensim works on Python 2, and all of the others work on both Python 2 and Python 3.

Recipes are something that you'll rarely get exactly right on the first try. Something about the build will be wrong, and the build will break. Maybe you only notice a problem during tests, but you need more info than you got from the tests running in conda-build. Conda-build 3.17.0 added the subcommand, conda debug, that is designed to facilitate the recipe debugging process.

Fundamentally, debugging is a process of getting into or recreating the environment and set of shell environment variables that conda-build creates during its build or test processes. This has been possible for a very long time---you could observe the build output, figure out where the files from your build were placed, navigate there, and finally, activate the appropriate environment(s). Then you might also need to set some environment variables manually.

What conda debug does is to create environments for you and provide you with a single command line that you can copy/paste to enter a debugging environment.

The conda debug command accepts 1 of 2 kinds of inputs: a recipe folder or a path to a built package.

If a path to a recipe folder is provided, conda debug creates the build and host environments. It provisions any source code that your recipe specifies. It leaves the build-time scripts in the work folder for you.  When complete, conda debug prints something like this:

################################################################################
Build and/or host environments created for debugging.  To enter a debugging environment:

cd /Users/UserName/miniconda3/conda-bld/debug_1542385789430/work && source /Users/UserName/miniconda3/conda-bld/debug_1542385789430/work/build_env_setup.sh

To run your build, you might want to start with running the conda_build.sh file.
################################################################################

If a path to a built package is provided, conda debug creates the test environment. It prepares any test files that the recipe specified. When complete, conda debug prints something like this:

################################################################################
Test environment created for debugging.  To enter a debugging environment:

cd /Users/UserName/miniconda3/conda-bld/conda-build_1542302975704/work && source /Users/UserName/miniconda3/conda-bld/conda-build_1542302975704/work/build_env_setup.sh

To run your tests, you might want to start with running the conda_test_runner.sh file.
################################################################################

Given the output above, you can now enter an environment to start debugging. Copy paste from your terminal and go:

cd /Users/UserName/miniconda3/conda-bld/debug_1542385789430/work && source /Users/UserName/miniconda3/conda-bld/debug_1542385789430/work/build_env_setup.sh

This is where you'll hopefully know what build commands you want to run to help you debug. Every build is different so your experience will vary. However, if you have no idea at all, you could probably start by running the appropriate build or test script, as mentioned in the output. If you do this, remember that these scripts might be written to exit on error, which may close your shell session. It may be wise to only run these scripts in an explicit subshell:

bash conda_build.sh
bash conda_test_runner.sh

Multiple outputs effectively give the recipe many build phases to consider. The --output-id argument is the mechanism to specify which of these should be used to create the debug envs and scripts. The --output-id argument accepts an fnmatch pattern. You can match any part of the output filenames. This really only works for conda packages, not other output types, such as wheels, because conda-build can't currently predict their filenames without actually carrying out a build.

For example, our NumPy recipe has multiple outputs. If we wanted to debug the NumPy-base output, we would specify it with a command like:

conda debug numpy-feedstock --output-id="numpy-base*"

If you have a matrix build, you may need to be more specific:

Specified --output-id matches more than one output (['/Users/msarahan/miniconda3/conda-bld/debug_1542387301945/osx-64/numpy-base-1.14.6-py27h1a60bec_4.tar.bz2', '/Users/msarahan/miniconda3/conda-bld/debug_1542387301945/osx-64/numpy-base-1.14.6-py27h8a80b8c_4.tar.bz2', '/Users/msarahan/miniconda3/conda-bld/debug_1542387301945/osx-64/numpy-base-1.14.6-py36h1a60bec_4.tar.bz2', '/Users/msarahan/miniconda3/conda-bld/debug_1542387301945/osx-64/numpy-base-1.14.6-py36h8a80b8c_4.tar.bz2', '/Users/msarahan/miniconda3/conda-bld/debug_1542387301945/osx-64/numpy-base-1.14.6-py37h1a60bec_4.tar.bz2', '/Users/msarahan/miniconda3/conda-bld/debug_1542387301945/osx-64/numpy-base-1.14.6-py37h8a80b8c_4.tar.bz2']).  Please refine your output id so that only a single output is found.

You could either reduce your matrix by changing your conda_build_config.yaml, or making a simpler one and passing it on the CLI, or by using the CLI to reduce it.

conda debug numpy-feedstock --output-id="numpy-base*" --python=3.6 --variants="{blas_impl: openblas}"

Specified --output-id matches more than one output (['/Users/UserName/miniconda3/conda-bld/debug_1542387443190/osx-64/numpy-base-1.14.6-py36h28eea48_4.tar.bz2', '/Users/UserName/miniconda3/conda-bld/debug_1542387443190/osx-64/numpy-base-1.14.6-py36ha711998_4.tar.bz2']).  Please refine your output id so that only a single output is found.

However, this is still not enough as our matrix includes two BLAS implementations, MKL and OpenBLAS.

Further reduction:

conda debug numpy-feedstock --output-id="numpy-base*" --python=3.6 --variants="{blas_impl: 'openblas'}"

Debugging folders are named in a way that the conda build purge command will find and clean up. If you use the -p/--path CLI argument, conda-build will not detect these and you'll need to manually clean up yourself. conda build purge-all will also remove previously built packages.

You can specify where you want the root of your debugging stuff to go with the -p/--path CLI argument. The way this works is that conda-build treats that as its "croot" where packages get cached as necessary, as well as potentially indexed. When using the --path argument, you may see folders like "osx-64" or other platform subdirs in the path you specify. It is safe to remove them or ignore them.

There are times when you may want to process a single file in different ways at more than 1 step in the render-build-test flow of conda-build. Conda-build sets the CONDA_BUILD_STATE environment variable during each of these phases. The possible values are:

• RENDER---Set during evaluation of the meta.yaml file.
• BUILD---Set during processing of the bld.bat or build.sh script files.
• TEST---Set during the running of any run_test scripts, which also includes any commands defined in meta.yaml in the test/commands section.

The CONDA_BUILD_STATE variable is undefined outside of these locations.

During the build process, the following environment variables are set, on Windows with bld.bat and on macOS and Linux with build.sh. By default, these are the only variables available to your build script. Unless otherwise noted, no variables are inherited from the shell environment in which you invoke conda-build. To override this behavior, see Inherited environment variables.

Unix-style packages on Windows, which are usually statically linked to executables, are built in a special Library directory under the host prefix. The environment variables listed in the following table are defined only on Windows.

The environment variables listed in the following table are defined only on macOS and Linux.

The environment variables listed in the following table are defined only on macOS.

The environment variable listed in the following table is defined only on Linux.

The environment variables listed in the following table are defined when the source is a git repository, specifying the source either with git_url or path.

These can be used in conjunction with templated meta.yaml files to set things---such as the build string---based on the state of the git repository.

The environment variables listed in the following table are defined when the source is a mercurial repository.

Other than those mentioned above, no variables are inherited from the environment in which you invoke conda-build. You can choose to inherit additional environment variables by adding them to meta.yaml:

build:
  script_env:
   - TMPDIR
   - LD_LIBRARY_PATH # [linux]
   - DYLD_LIBRARY_PATH # [osx]

If an inherited variable is missing from your shell environment, it remains unassigned, but a warning is issued noting that it has no value assigned.

Additionally, values can be set by including = followed by the desired value:

build:
  script_env:
   - MY_VAR=some value

WARNING:

Inheriting environment variables can make it difficult for others to reproduce binaries from source with your recipe. Use this feature with caution or explicitly set values using the = syntax.

NOTE:

If you split your build and test phases with --no-test and --test, you need to ensure that the environment variables present at build time and test time match. If you do not, the package hashes may use different values and your package may not be testable because the hashes will differ.

All of the above environment variables are also set during the test process, using the test prefix instead of the build prefix.

If you have software in a Python wheel file and want to use it with conda or install it in a conda environment, there are 3 ways.

The best way is to obtain the source code for the software and build a conda package from the source and not from a wheel. This helps ensure that the new package uses other conda packages to satisfy its dependencies.

The second best way is to build a conda package from the wheel file. This tells conda more about the files present than a pip install. It is also less likely than a pip install to cause errors by overwriting (or "clobbering") files. Building a conda package from the wheel file also has the advantage that any clobbering is more likely to happen at build time and not runtime.

The third way is to use pip to install a wheel file into a conda environment. Some conda users have used this option safely. The first 2 ways are still the safest and most reliable.

To build a conda package from a wheel file, install the .whl file in the conda recipe's bld.bat or build.sh file.

You may download the .whl file in the source section of the conda recipe's meta.yaml file.

You may instead put the URL directly in the pip install command.

EXAMPLE: The conda recipe for TensorFlow has a pip install command in build.sh with the URL of a .whl file. The meta.yaml file does not download or list the .whl file.

NOTE:

You must use the --no-deps option in the pip install command in order to avoid bundling dependencies into your conda-package.

If you run pip install without the --no-deps option, pip will often install dependencies in your conda recipe and those dependencies will become part of your package. This wastes space in the package and increases the risk of file overlap, file clobbering, and broken packages.

The tutorials will guide you through how to build conda packages — whether you're creating a package with compilers, using conda skeleton, creating from scratch, or building R packages using skeleton CRAN.

Conda-build uses recipes to create conda packages. We have guides on debugging conda recipes, sample recipes for you to use, and information on how to build a package without a recipe.

Use our environment variables guide to understand which environment variables are available, set, and inherited, and how they affect different processes.

The user guide includes information about wheel files and how to build conda packages from them.

The build.sh file is the build script for Linux and macOS and bld.bat is the build script for Windows. These scripts contain the logic that carries out your build steps. Traditionally it has also included install steps. With the traditional one-package-per-recipe way of doing things, anything that your build script copies into the $PREFIX or %PREFIX% folder will be included in your output package. For example, this build.sh:

mkdir -p $PREFIX/bin
cp $RECIPE_DIR/my_script_with_recipe.sh $PREFIX/bin/super-cool-script.sh

If you don't care about deploying your package with pip on PyPI, this can save you a lot of time in figuring out the proper way to include additional files with setup.py.

There are many environment variables defined for you to use in build.sh and bld.bat. Please see Environment variables for more information.

As of conda-build 2.1, you can also define multiple output packages. Each package has its own script or list of files to include. The rules for these outputs are documented at Outputs section. When any output is defined, this overrides the default behavior of bundling anything in $PREFIX. So to output multiple packages from a single recipe, remove any installation steps from build.sh or bld.bat and do them instead in your install script(s) for each output.

build.sh and bld.bat are optional. You can instead use the build/script key in your meta.yaml, with each value being either a string command or a list of string commands. Any commands you put there must be able to run on every platform for which you build. For example, you can't use the cp command because cmd.exe won't understand it in Windows.

build.sh is run with bash and bld.bat is run with cmd.exe.

There is some development towards the ability to use bash scripts in Windows, but this is not currently supported. You may write your script as a .sh file, and then call it in your bld.bat file, but there is no way to directly run build.sh on Windows. The conda recipe at https://github.com/AnacondaRecipes/conda-feedstock/tree/master/recipe is an example of this method.

Anaconda 5.0 switched from OS-provided compiler tools to our own toolsets. This allows improved compiler capabilities, including better security and performance. This page describes how to use these tools and enable these benefits.

Before Anaconda 5.0, compilers were installed using system tools such as XCode or yum install gcc. Now there are conda packages for Linux and macOS compilers. Unlike the previous GCC 4.8.5 packages that included GCC, g++, and GFortran all in the same package, these conda packages are split into separate compilers:

macOS:

• clang_osx-64.
• clangxx_osx-64.
• gfortran_osx-64.

Linux:

• gcc_linux-64.
• gxx_linux-64.
• gfortran_linux-64.

A compiler's "build platform" is the platform where the compiler runs and builds the code.

A compiler's "host platform" is the platform where the built code will finally be hosted and run.

Notice that all of these package names end in a platform identifier which specifies the host platform. All compiler packages are specific to both the build platform and the host platform.

The compiler packages can be installed with conda. Because they are designed with (pseudo) cross-compiling in mind, all of the executables in a compiler package are "prefixed." Instead of gcc, the executable name of the compiler you use will be something like x86_64-conda_cos6-linux-gnu-gcc. These full compiler names are shown in the build logs, recording the host platform and helping prevent the common mistake of using the wrong compiler.

Many build tools such as make and CMake search by default for a compiler named simply gcc, so we set environment variables to point these tools to the correct compiler.

We set these variables in conda activate.d scripts, so any environment in which you will use the compilers must first be activated so the scripts will run. Conda-build does this activation for you using activation hooks installed with the compiler packages in CONDA_PREFIX/etc/conda/activate.d, so no additional effort is necessary.

You can activate the root environment with the command conda activate root.

The macOS compilers require the macOS 10.9 SDK or above. The SDK license prevents it from being bundled in the conda package. We know of 2 current sources for the macOS SDKs:

• https://github.com/devernay/xcodelegacy
• https://github.com/phracker/MacOSX-SDKs

We usually install the 10.10 SDK at /opt/MacOSX10.10.sdk but you may install it anywhere. Edit your conda_build_config.yaml file to point to it, like this:

CONDA_BUILD_SYSROOT:
  - /opt/MacOSX10.10.sdk        # [osx]

At Anaconda, we have this configuration setting in a centralized conda_build_config.yaml at the root of our recipe repository. Since we run build commands from that location, the file and the setting are used for all recipes. The conda_build_config.yaml search order is described further at Creating conda-build variant config files.

Build scripts for macOS should make use of the variables MACOSX_DEPLOYMENT_TARGET and CONDA_BUILD_SYSROOT, which are set by conda-build (see Environment variables). These variables should be translated into correct compiler arguments, e.g. for Clang this would be:

clang .. -isysroot ${CONDA_BUILD_SYSROOT} -mmacosx-version-min=${MACOSX_DEPLOYMENT_TARGET} ..

Most build tools, e.g. CMake and distutils (setuptools), will automatically pick up MACOSX_DEPLOYMENT_TARGET but you need to pass CONDA_BUILD_SYSROOT explicitly. For CMake, this can be done with the option -DCMAKE_OSX_SYSROOT=${CONDA_BUILD_SYSROOT}. When building Python extensions with distutils, one should always extend CFLAGS before calling setup.py:

export CFLAGS="${CFLAGS} -i sysroot ${CONDA_BUILD_SYSROOT}"

When building C++ extensions with Cython, CXXFLAGS must be similarly modified.

Some users want to use the latest Anaconda packages but do not yet want to use the Anaconda compilers. To enable this, the latest Python package builds have a default _sysconfigdata file. This file sets the compilers provided by the system, such as gcc and g++, as the default compilers. This way allows legacy recipes to keep working.

Python packages also include an alternative _sysconfigdata file that sets the Anaconda compilers as the default compilers. The Anaconda Python executable itself is made with these Anaconda compilers.

The compiler packages set the environment variable _PYTHON_SYSCONFIGDATA_NAME, which tells Python which _sysconfigdata file to use. This variable is set at activation time using the activation hooks described above.

The new _sysconfigdata customization system is only present in recent versions of the Python package. Conda-build automatically tries to use the latest Python version available in the currently configured channels, which normally gets the latest from the default channel. If you're using something other than conda-build while working with the new compilers, conda does not automatically update Python, so make sure you have the correct _sysconfigdata files by updating your Python package manually.

The Anaconda 5.0 compilers and conda-build 3 are designed to work together.

Conda-build 3 defines a special jinja2 function, compiler(), to make it easy to specify compiler packages dynamically on many platforms. The compiler function takes at least 1 argument, the language of the compiler to use:

requirements:
  build:
    - {{ compiler('c') }}

"Cross-capable" recipes can be used to make packages with a host platform different than the build platform where conda-build runs. To write cross-capable recipes, you may also need to use the "host" section in the requirements section. In this example we set "host" to "zlib" to tell conda-build to use the zlib in the conda environment and not the system zlib. This makes sure conda-build uses the zlib for the host platform and not the zlib for the build platform.

requirements:
  build:
    - {{ compiler('c') }}
  host:
    - zlib

Generally, the build section should include compilers and other build tools and the host section should include everything else, including shared libraries, Python, and Python libraries.

Anaconda's compilers for Linux are built with something called crosstool-ng. They include not only GCC, but also a "sysroot" with glibc, as well as the rest of the toolchain (binutils). Ordinarily, the sysroot is something that your system provides, and it is what establishes the libc compatibility bound for your compiled code. Any compilation that uses a sysroot other than the system sysroot is said to be "cross-compiling." When the target OS and the build OS are the same, it is called a "pseudo-cross-compiler." This is the case for normal builds with Anaconda's compilers on Linux.

Unfortunately, some software tools do not handle sysroots in intuitive ways. CMake is especially bad for this. Even though the compiler itself understands its own sysroot, CMake insists on ignoring that.  We've filed issues at:

• https://gitlab.kitware.com/cmake/cmake/issues/17483

Additionally, this Stack Overflow issue has some more information: https://stackoverflow.com/questions/36195791/cmake-missing-sysroot-when-cross-compiling

In order to teach CMake about the sysroot, you must do additional work. As an example, please see our recipe for libnetcdf at https://github.com/AnacondaRecipes/libnetcdf-feedstock/tree/master/recipe

In particular, you'll need to copy the cross-linux.cmake file there, and reference it in your build.sh file:

CMAKE_PLATFORM_FLAGS+=(-DCMAKE_TOOLCHAIN_FILE="${RECIPE_DIR}/cross-linux.cmake")

cmake -DCMAKE_INSTALL_PREFIX=${PREFIX} \
  ${CMAKE_PLATFORM_FLAGS[@]} \
  ${SRC_DIR}

The compiler packages listed above are small packages that only include the activation scripts and list most of the software they provide as runtime dependencies.

This design is intended to make it easy for you to customize your own compiler packages by copying these recipes and changing the flags. You can then edit the conda_build_config.yaml file to specify your own packages.

We have been careful to select good, general purpose, secure, and fast flags. We have also used them for all packages in Anaconda Distribution 5.0.0, except for some minor customizations in a few recipes. When changing these flags, remember that choosing the wrong flags can reduce security, reduce performance, and cause incompatibilities.

With that warning in mind, let's look at good ways to customize Clang.

1. Download or fork the code from https://github.com/anacondarecipes/aggregate. The Clang package recipe is in the clang folder. The main material is in the llvm-compilers-feedstock folder.

2. Edit clang/recipe/meta.yaml:

   package:
     name: clang_{{ target_platform }}
     version: {{ version }}

   The name here does not matter but the output names below do. Conda-build expects any compiler to follow the BASENAME_PLATFORMNAME pattern, so it is important to keep the {{target_platform}} part of the name.

   {{ version }} is left as an intentionally undefined jinja2 variable. It is set later in conda_build_config.yaml.

3. Before any packaging is done, run the build.sh script: https://github.com/AnacondaRecipes/aggregate/blob/master/clang/build.sh

   In this recipe, values are changed here. Those values are inserted into the activate scripts that are installed later.

   #!/bin/bash
   
   CHOST=${macos_machine}
   
   FINAL_CPPFLAGS="-D_FORTIFY_SOURCE=2 -mmacosx-version-min=${macos_min_version}"
   FINAL_CFLAGS="-march=core2 -mtune=haswell -mssse3 -ftree-vectorize -fPIC -fPIE -fstack-protector-strong -O2 -pipe"
   FINAL_CXXFLAGS="-march=core2 -mtune=haswell -mssse3 -ftree-vectorize -fPIC -fPIE -fstack-protector-strong -O2 -pipe -stdlib=libc++ -fvisibility-inlines-hidden -std=c++14 -fmessage-length=0"
   # These are the LDFLAGS for when the linker is being called directly, without "-Wl,"
   FINAL_LDFLAGS="-pie -headerpad_max_install_names"
   # These are the LDFLAGS for when the linker is being driven by a compiler, with "-Wl,"
   FINAL_LDFLAGS_CC="-Wl,-pie -Wl,-headerpad_max_install_names"
   FINAL_DEBUG_CFLAGS="-Og -g -Wall -Wextra -fcheck=all -fbacktrace -fimplicit-none -fvar-tracking-assignments"
   FINAL_DEBUG_CXXFLAGS="-Og -g -Wall -Wextra -fcheck=all -fbacktrace -fimplicit-none -fvar-tracking-assignments"
   FINAL_DEBUG_FFLAGS="-Og -g -Wall -Wextra -fcheck=all -fbacktrace -fimplicit-none -fvar-tracking-assignments"
   
   find "${RECIPE_DIR}" -name "*activate*.sh" -exec cp {} . \;
   
   find . -name "*activate*.sh" -exec sed -i.bak "s|@CHOST@|${CHOST}|g" "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@CPPFLAGS@|${FINAL_CPPFLAGS}|g"             "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@CFLAGS@|${FINAL_CFLAGS}|g"                 "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@DEBUG_CFLAGS@|${FINAL_DEBUG_CFLAGS}|g"     "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@CXXFLAGS@|${FINAL_CXXFLAGS}|g"             "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@DEBUG_CXXFLAGS@|${FINAL_DEBUG_CXXFLAGS}|g" "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@DEBUG_CXXFLAGS@|${FINAL_DEBUG_CXXFLAGS}|g" "{}" \;
   # find . -name "*activate*.sh" -exec sed -i.bak "s|@FFLAGS@|${FINAL_FFLAGS}|g"                 "{}" \;
   # find . -name "*activate*.sh" -exec sed -i.bak "s|@DEBUG_FFLAGS@|${FINAL_DEBUG_FFLAGS}|g"     "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@LDFLAGS@|${FINAL_LDFLAGS}|g"               "{}" \;
   find . -name "*activate*.sh" -exec sed -i.bak "s|@LDFLAGS_CC@|${FINAL_LDFLAGS_CC}|g"         "{}" \;
   find . -name "*activate*.sh.bak" -exec rm "{}" \;

4. With those changes to the activate scripts in place, it's time to move on to installing things. Look back at the clang folder's meta.yaml. Here's where we change the package name. Notice what comes before the {{ target_platform }}.

   outputs:
     - name: super_duper_clang_{{ target_platform }}
       script: install-clang.sh
       requirements:
         - clang {{ version }}

   The script reference here is another place you might add customization. You'll either change the contents of those install scripts or change the scripts that those install scripts are installing.

   Note that we make the package clang in the main material agree in version with our output version. This is implicitly the same as the top-level recipe. The clang package sets no environment variables at all, so it may be difficult to use directly.

5. Let's examine the script install-clang.sh:

   #!/bin/bash
   
   set -e -x
   
   CHOST=${macos_machine}
   
   mkdir -p "${PREFIX}"/etc/conda/{de,}activate.d/
   cp "${SRC_DIR}"/activate-clang.sh "${PREFIX}"/etc/conda/activate.d/activate_"${PKG_NAME}".sh
   cp "${SRC_DIR}"/deactivate-clang.sh "${PREFIX}"/etc/conda/deactivate.d/deactivate_"${PKG_NAME}".sh
   
   pushd "${PREFIX}"/bin
     ln -s clang ${CHOST}-clang
   popd

   Nothing here is too unusual.

   Activate scripts are named according to our package name so they won't conflict with other activate scripts.

   The symlink for Clang is a Clang implementation detail that sets the host platform.

   We define macos_machine in aggregate's conda_build_config.yaml: https://github.com/AnacondaRecipes/aggregate/blob/master/conda_build_config.yaml#L79

   The activate scripts that are being installed are where we actually set the environment variables. Remember that these have been modified by build.sh.

6. With any of your desired changes in place, go ahead and build the recipe.

   You should end up with a super_duper_clang_osx-64 package. Or, if you're not on macOS and are modifying a different recipe, you should end up with an equivalent package for your platform.

Remember the Jinja2 function, {{ compiler('c') }}? Here's where that comes in. Specific keys in conda_build_config.yaml are named for the language argument to that jinja2 function. In your conda_build_config.yaml, add this:

c_compiler:
  - super_duper_clang

Note that we're not adding the target_platform part, which is separate. You can define that key, too:

c_compiler:
  - super_duper_clang
target_platform:
  - win-64

With those two keys defined, conda-build will try to use a compiler package named super_duper_clang_win-64. That package needs to exist for your native platform. For example, if you're on macOS, your native platform is osx-64.

The package subdirectory for your native platform is the build platform. The build platform and the target_platform can be the same, and they are the same by default, but they can also be different. When they are different, you're cross-compiling.

If you ever needed a different compiler key for the same language, remember that the language key is arbitrary. For example, we might want different compilers for Python and for R within one ecosystem. On Windows, the Python ecosystem uses the Microsoft Visual C compilers, while the R ecosystem uses the Mingw compilers.

Let's start in conda_build_config.yaml:

python_c_compiler:
  - vs2015
r_c_compiler:
  - m2w64-gcc
target_platform:
  - win-64

In Python recipes, you'd have:

requirements:
  build:
    - {{ compiler('python_c') }}

In R recipes, you'd have:

requirements:
  build:
    - {{ compiler('r_c') }}

This example is a little contrived, because the m2w64-gcc_win-64 package is not available. You'd need to create a metapackage m2w64-gcc_win-64 to point at the m2w64-gcc package, which does exist on the msys2 channel on repo.anaconda.com.

For most languages, certainly for "c" and for "cxx", compiling any given program may create a run-time dependence on symbols from the respective standard library. For example, the standard library for C on linux is generally glibc, and a core component of your operating system. Conda is not able to change or supersede this library (it would be too risky to try to). A similar situation exists on MacOS and on Windows.

Compiler packages usually have two ways to deal with this dependence:

• assume the package must be there (like glibc on linux).
• always add a run-time requirement on the respective stdlib (e.g. libcxx on MacOS).

However, even if we assume the package must be there, the information about the glibc version is still a highly relevant piece of information, which is also why it is reflected in the __glibc virtual package.

For example, newer packages may decide over time to increase the lowest version of glibc that they support. We therefore need a way to express this dependence in a way that conda will be able to understand, so that (in conjunction with the __glibc virtual package) the environment resolver will not consider those packages on machines whose glibc version is too old.

The way to do this is to use the Jinja2 function {{ stdlib('c') }}, which matches {{ compiler('c') }} in as many ways as possible. Let's start again with the conda_build_config.yaml:

c_stdlib:
  - sysroot                     # [linux]
  - macosx_deployment_target    # [osx]
c_stdlib_version:
  - 2.17                        # [linux]
  - 10.13                       # [osx]

In the recipe we would then use:

requirements:
  build:
    - {{ compiler('c') }}
    - {{ stdlib('c') }}

This would then express that the resulting package requires sysroot ==2.17 (corresponds to glibc) on linux and macosx_deployment_target ==10.13 on MacOS in the build environment, respectively. How this translates into a run-time dependence can be defined in the metadata of the respective conda (meta-)package which represents the standard library (i.e. those defined under c_stdlib above).

In this example, sysroot 2.17 would generate a run-export on __glibc >=2.17 and macosx_deployment_target 10.13 would similarly generate __osx >=10.13. This way, we enable packages to define their own expectations about the standard library in a unified way, and without implicitly depending on some global assumption about what the lower version on a given platform must be.

In principle, this facility would make it possible to also express the dependence on separate stdlib implementations (like musl instead of glibc), or to remove the need to assume that a C++ compiler always needs to add a run-export on the C++ stdlib -- it could then be left up to packages themselves whether they need {{ stdlib('cxx') }} or not.

You might want to use the Anaconda compilers outside of conda-build so that you use the same versions, flags, and configuration, for maximum compatibility with Anaconda packages (but in a case where you want simple tarballs, for example). In this case, there is a gotcha.

Even if Anaconda compilers are used from outside of conda-build, the GCC specs are customized so that, when linking an executable or a shared library, an RPATH pointing to lib/ inside the current enviroment prefix directory ($CONDA_PREFIX/lib) is added. This is done by changing the link_libgcc: section inside GCC specs file, and this change is done so that LD_LIBRARY_PATH isn't required for basic libraries.

conda-build knows how to make this automatically relocatable, so that this RPATH will be changed to point to the environment where the package is being installed (at installation time, by conda). But if you only pack this binary in a tarball, it will continue containing this hardcoded RPATH to an environment in your machine. In this case, it is recommended to manually remove the RPATH.

All the metadata in the conda-build recipe is specified in the meta.yaml file. See the example below:

{% set version = "1.1.0" %}

package:
  name: imagesize
  version: {{ version }}

source:
  url: https://pypi.io/packages/source/i/imagesize/imagesize-{{ version }}.tar.gz
  sha256: f3832918bc3c66617f92e35f5d70729187676313caa60c187eb0f28b8fe5e3b5

build:
  noarch: python
  number: 0
  script: python -m pip install --no-deps --ignore-installed .

requirements:
  host:
    - python
    - pip
  run:
    - python

test:
  imports:
    - imagesize

about:
  home: https://github.com/shibukawa/imagesize_py
  license: MIT
  summary: 'Getting image size from png/jpeg/jpeg2000/gif file'
  description: |
    This module analyzes jpeg/jpeg2000/png/gif image header and
    return image size.
  dev_url: https://github.com/shibukawa/imagesize_py
  doc_url: https://pypi.python.org/pypi/imagesize
  doc_source_url: https://github.com/shibukawa/imagesize_py/blob/master/README.rst

All sections are optional except for package/name and package/version.

Headers must appear only once. If they appear multiple times, only the last is remembered. For example, the package: header should appear only once in the file.

Specifies package information.

The lower case name of the package. It may contain "-", but no spaces.

package:
  name: bsdiff4

The version number of the package. Use the PEP-386 verlib conventions. Cannot contain "-". YAML interprets version numbers such as 1.0 as floats, meaning that 0.10 will be the same as 0.1. To avoid this, put the version number in quotes so that it is interpreted as a string.

package:
  version: "1.1.4"

NOTE:

Post-build versioning: In some cases, you may not know the version, build number, or build string of the package until after it is built. In these cases, you can perform Templating with Jinja or utilize Git environment variables and Inherited environment variables.

Specifies where the source code of the package is coming from. The source may come from a tarball file, git, hg, or svn. It may be a local path and it may contain patches.

source:
  url: https://pypi.python.org/packages/source/b/bsdiff4/bsdiff4-1.1.4.tar.gz
  md5: 29f6089290505fc1a852e176bd276c43
  sha1: f0a2c9a30073449cfb7d171c57552f3109d93894
  sha256: 5a022ff4c1d1de87232b1c70bde50afbb98212fd246be4a867d8737173cf1f8f

If an extracted archive contains only 1 folder at its top level, its contents will be moved 1 level up, so that the extracted package contents sit in the root of the work folder.

You can also specify multiple URLs for the same source archive. They will be attempted in order, should one fail.

source:
  url:
    - https://archive.linux.duke.edu/cran/src/contrib/ggblanket_6.0.0.tar.gz
    - https://archive.linux.duke.edu/cran/src/contrib/Archive/ggblanket/ggblanket_6.0.0.tar.gz
  sha256: cd2181fe3d3365eaf36ff8bbbc90ea9d76c56d40e63386b4eefa0e3120ec6665

The git_url can also be a relative path to the recipe directory.

source:
  git_url: https://github.com/ilanschnell/bsdiff4.git
  git_rev: 1.1.4 # (Defaults to "HEAD")
  git_depth: 1 # (Defaults to -1/not shallow)

The depth argument relates to the ability to perform a shallow clone. A shallow clone means that you only download part of the history from Git. If you know that you only need the most recent changes, you can say, git_depth: 1, which is faster than cloning the entire repo. The downside to setting it at 1 is that, unless the tag is on that specific commit, then you won't have that tag when you go to reference it in git_rev (for example). If your git_depth is insufficient to capture the tag in git_rev, you'll encounter an error. So in the example above, unless the 1.1.4 is the very head commit and the one that you're going to grab, you may encounter an error.

source:
  hg_url: ssh://hg@bitbucket.org/ilanschnell/bsdiff4
  hg_tag: 1.1.4

source:
  svn_url: https://github.com/ilanschnell/bsdiff
  svn_rev: 1.1.4 # (defaults to head)
  svn_ignore_externals: True # (defaults to False)
  svn_username: username  # Optional, if set must also have svn_password
  svn_password: password  # Optional, if set must also have svn_username

To access a restricted SVN repository, specify both svn_username and svn_password.

CAUTION:

source:
  svn_username: {{ environ["SVN_USERNAME"] }}
  svn_password: {{ environ["SVN_PASSWORD"] }}

If the path is relative, it is taken relative to the recipe directory. The source is copied to the work directory before building.

source:
  path: ../src

If the local path is a git or svn repository, you get the corresponding environment variables defined in your build environment. The only practical difference between git_url or hg_url and path as source arguments is that git_url and hg_url would be clones of a repository, while path would be a copy of the repository. Using path allows you to build packages with unstaged and uncommitted changes in the working directory. git_url can build only up to the latest commit.

Patches may optionally be applied to the source.

source:
  #[source information here]
  patches:
    - my.patch # the patch file is expected to be found in the recipe

Conda-build automatically determines the patch strip level.

Within conda-build's work directory, you may specify a particular folder to place source into. This feature is new in conda-build 3.0. Conda-build will always drop you into the same folder (build folder/work), but it's up to you whether you want your source extracted into that folder, or nested deeper. This feature is particularly useful when dealing with multiple sources, but can apply to recipes with single sources as well.

source:
  #[source information here]
  folder: my-destination/folder

The filename key is fn. It was formerly required with URL source types. It is not required now.

If the fn key is provided, the file is saved on disk with that name. If the fn key is not provided, the file is saved on disk with a name matching the last part of the URL.

For example, http://www.something.com/myfile.zip has an implicit filename of myfile.zip. Users may change this by manually specifying fn.

source:
  url: http://www.something.com/myfile.zip
  fn: otherfilename.zip

Some software is most easily built by aggregating several pieces. For this, conda-build 3.0 has added support for arbitrarily specifying many sources.

The syntax is a list of source dictionaries. Each member of this list follows the same rules as the single source for earlier conda-build versions (listed above). All features for each member are supported.

Example:

source:
  - url: https://package1.com/a.tar.bz2
    folder: stuff
  - url: https://package1.com/b.tar.bz2
    folder: stuff
  - git_url: https://github.com/conda/conda-build
    folder: conda-build

Here, the two URL tarballs will go into one folder, and the git repo is checked out into its own space. Git will not clone into a non-empty folder.

NOTE:

Specifies build information.

Each field that expects a path can also handle a glob pattern. The matching is performed from the top of the build environment, so to match files inside your project you can use a pattern similar to the following one: "**/myproject/**/*.txt". This pattern will match any .txt file found in your project.

NOTE:

Recursive globbing using ** is supported only in conda-build >= 3.0.

The build number should be incremented for new builds of the same version. The number defaults to 0. The build string cannot contain "-". The string defaults to the default conda-build string plus the build number.

build:
  number: 1
  string: abc

A hash will appear when the package is affected by one or more variables from the conda_build_config.yaml file. The hash is made up from the "used" variables - if anything is used, you have a hash. If you don't use these variables then you won't have a hash. There are a few special cases that do not affect the hash, such as Python and R or anything that already had a place in the build string.

The build hash will be added to the build string if these are true for any dependency:

• package is an explicit dependency in build, host, or run deps
• package has a matching entry in conda_build_config.yaml which is a pin to a specific version, not a lower bound
• that package is not ignored by ignore_version

OR

• package uses {{ compiler() }} jinja2 function

You can also influence which variables are considered for the hash with:

build:
  force_use_keys:
    - package_1
  force_ignore_keys:
    - package_2

This will ensure that the value of package_2 will not be considered for the hash, and package_1 will be, regardless of what conda-build discovers is used by its inspection.

This may be useful to further split complex multi-output builds, to ensure each package is built, or to ensure the right package hash when using more complex templating or scripting.

The following example creates a Python entry point named "bsdiff4" that calls bsdiff4.cli.main_bsdiff4().

build:
  entry_points:
    - bsdiff4 = bsdiff4.cli:main_bsdiff4
    - bspatch4 = bsdiff4.cli:main_bspatch4

If osx_is_app is set, entry points use python.app instead of Python in macOS. The default is False.

build:
  osx_is_app: True

Adding track_features to one or more of the options will cause conda to de-prioritize it or “weigh it down.” The lowest priority package is the one that would cause the most track_features to be activated in the environment. The default package among many variants is the one that would cause the least track_features to be activated.

No two packages in a given subdir should ever have the same track_feature.

build:
  track_features:
    - feature2

This is needed for some packages that use features specific to setuptools. The default is False.

build:
  preserve_egg_dir: True

Some packages ship .py files that cannot be compiled, such as those that contain templates. Some packages also ship .py files that should not be compiled yet, because the Python interpreter that will be used is not known at build time. In these cases, conda-build can skip attempting to compile these files. The patterns used in this section do not need the ** to handle recursive paths.

build:
  skip_compile_pyc:
    - "*/templates/*.py"          # These should not (and cannot) be compiled
    - "*/share/plugins/gdb/*.py"  # The python embedded into gdb is unknown

A list of globs for files that should always be copied and never soft linked or hard linked.

build:
  no_link:
    - bin/*.py  # Don't link any .py files in bin/

Used instead of build.sh or bld.bat. For short build scripts, this can be more convenient. You may need to use selectors to use different scripts for different platforms.

build:
  script: python setup.py install --single-version-externally-managed --record=record.txt

Set which RPATHs are used when making executables relocatable on Linux. This is a Linux feature that is ignored on other systems. The default is lib/.

build:
  rpaths:
    - lib/
    - lib/R/lib/

Force files to always be included, even if they are already in the environment from the build dependencies. This may be needed, for example, to create a recipe for conda itself.

build:
  always_include_files:
    - bin/file1
    - bin/file2

Advanced features. You can use the following 4 keys to control relocatability files from the build environment to the installation environment:

• binary_relocation.
• has_prefix_files.
• binary_has_prefix_files.
• ignore_prefix_files.

For more information, see Making packages relocatable.

Whether binary files should be made relocatable using install_name_tool on macOS or patchelf on Linux. The default is True. It also accepts False, which indicates no relocation for any files, or a list of files, which indicates relocation only for listed files.

build:
  binary_relocation: False

Binary files may contain the build prefix and need it replaced with the install prefix at installation time. Conda can automatically identify and register such files. The default is True.

NOTE:

build:
  detect_binary_files_with_prefix: False

Windows handles binary prefix replacement very differently than Unix-like systems such as macOS and Linux. At this time, we are unaware of any executable or library that uses hardcoded embedded paths for locating other libraries or program data on Windows. Instead, Windows follows DLL search path rules or more natively supports relocatability using relative paths. Because of this, conda ignores most prefixes. However, pip creates executables for Python entry points that do use embedded paths on Windows. Conda-build thus detects prefixes in all files and records them by default. If you are getting errors about path length on Windows, you should try to disable detect_binary_files_with_prefix. Newer versions of Conda, such as recent 4.2.x series releases and up, should have no problems here, but earlier versions of conda do erroneously try to apply any binary prefix replacement.

By default, conda-build tries to detect prefixes in all files. You may also elect to specify files with binary prefixes individually. This allows you to specify the type of file as binary, when it may be incorrectly detected as text for some reason. Binary files are those containing NULL bytes.

build:
  binary_has_prefix_files:
    - bin/binaryfile1
    - lib/binaryfile2

Text files---files containing no NULL bytes---may contain the build prefix and need it replaced with the install prefix at installation time. Conda will automatically register such files. Binary files that contain the build prefix are generally handled differently---see Binary has prefix files---but there may be cases where such a binary file needs to be treated as an ordinary text file, in which case they need to be identified.

build:
  has_prefix_files:
    - bin/file1
    - lib/file2

Used to exclude some or all of the files in the build recipe from the list of files that have the build prefix replaced with the install prefix.

To ignore all files in the build recipe, use:

build:
  ignore_prefix_files: True

To specify individual filenames, use:

build:
  ignore_prefix_files:
    - file1

This setting is independent of RPATH replacement. Use the Detect binary files with prefix setting to control that behavior.

Specifies whether conda-build should skip the build of this recipe. Particularly useful for defining recipes that are platform specific. The default is False.

build:
  skip: True  # [not win]

Allows you to specify "no architecture" when building a package, thus making it compatible with all platforms and architectures. Noarch packages can be installed on any platform.

Starting with conda-build 2.1, and conda 4.3, there is a new syntax that supports different languages. Assigning the noarch key as generic tells conda to not try any manipulation of the contents.

build:
  noarch: generic

noarch: generic is most useful for packages such as static javascript assets and source archives. For pure Python packages that can run on any Python version, you can use the noarch: python value instead:

build:
  noarch: python

The legacy syntax for noarch_python is still valid, and should be used when you need to be certain that your package will be installable where conda 4.3 is not yet available. All other forms of noarch packages require conda >=4.3 to install.

build:
  noarch_python: True

WARNING:

At the time of this writing, noarch packages should not make use of preprocess-selectors: noarch packages are built with the directives which evaluate to True in the platform it was built, which probably will result in incorrect/incomplete installation in other platforms.

The full conda-build recipe and rendered meta.yaml file is included in the Package metadata by default. You can disable this with:

build:
  include_recipe: False

Normally the build script in build.sh or bld.bat does not pass through environment variables from the command line. Only environment variables documented in Environment variables are seen by the build script. To "white-list" environment variables that should be passed through to the build script:

build:
  script_env:
    - MYVAR
    - ANOTHER_VAR

If a listed environment variable is missing from the environment seen by the conda-build process itself, a UserWarning is emitted during the build process and the variable remains undefined.

Additionally, values can be set by including = followed by the desired value:

build:
  script_env:
   - MY_VAR=some value

NOTE:

Inheriting environment variables can make it difficult for others to reproduce binaries from source with your recipe. Use this feature with caution or explicitly set values using the = syntax.

NOTE:

If you split your build and test phases with --no-test and --test, you need to ensure that the environment variables present at build time and test time match. If you do not, the package hashes may use different values, and your package may not be testable, because the hashes will differ.

Some build or host Requirements section will impose a runtime requirement. Most commonly this is true for shared libraries (e.g. libpng), which are required for linking at build time, and for resolving the link at run time. With run_exports (new in conda-build 3) such a runtime requirement can be implicitly added by host requirements (e.g. libpng exports libpng), and with run_exports/strong even by build requirements (e.g. GCC exports libgcc).

# meta.yaml of libpng
build:
  run_exports:
    - libpng

Here, because no specific kind of run_exports is specified, libpng's run_exports are considered "weak". This means they will only apply when libpng is in the host section, when they will add their export to the run section. If libpng were listed in the build section, the run_exports would not apply to the run section.

# meta.yaml of gcc compiler
build:
  run_exports:
    strong:
      - libgcc

There is also run_exports/weak which is equivalent to an unspecific kind of run_exports but useful if you want to define both strong and weak run exports.

Strong run_exports are used for things like runtimes, where the same runtime needs to be present in the host and the run environment, and exactly which runtime that should be is determined by what's present in the build section. This mechanism is how we line up appropriate software on Windows, where we must match MSVC versions used across all of the shared libraries in an environment.

# meta.yaml of some package using gcc and libpng
requirements:
  build:
    - gcc            # has a strong run export
  host:
    - libpng         # has a (weak) run export
    # - libgcc       <-- implicitly added by gcc
  run:
    # - libgcc       <-- implicitly added by gcc
    # - libpng       <-- implicitly added by libpng

You can express version constraints directly, or use any of the Jinja2 helper functions listed at Extra Jinja2 functions.

For example, you may use Pinning expressions to obtain flexible version pinning relative to versions present at build time:

build:
  run_exports:
    - {{ pin_subpackage('libpng', max_pin='x.x') }}

With this example, if libpng were version 1.6.34, this pinning expression would evaluate to >=1.6.34,<1.7.

If build and link dependencies need to impose constraints on the run environment but not necessarily pull in additional packages, then this can be done by altering the Run_constrained entries. In addition to weak/strong run_exports which add to the run requirements, weak_constrains and strong_constrains add to the run_constrained requirements. With these, e.g., minimum versions of compatible but not required packages (like optional plugins for the linked dependency, or certain system attributes) can be expressed:

requirements:
  build:
    - build-tool                 # has a strong run_constrained export
  host:
    - link-dependency            # has a weak run_constrained export
  run:
  run_constrained:
    # - system-dependency >=min  <-- implicitly added by build-tool
    # - optional-plugin >=min    <-- implicitly added by link-dependency

Note that run_exports can be specified both in the build section and on a per-output basis for split packages.

run_exports only affects directly named dependencies. For example, if you have a metapackage that includes a compiler that lists run_exports, you also need to define run_exports in the metapackage so that it takes effect when people install your metapackage.  This is important, because if run_exports affected transitive dependencies, you would see many added dependencies to shared libraries where they are not actually direct dependencies. For example, Python uses bzip2, which can use run_exports to make sure that people use a compatible build of bzip2. If people list python as a build time dependency, bzip2 should only be imposed for Python itself and should not be automatically imposed as a runtime dependency for the thing using Python.

The potential downside of this feature is that it takes some control over constraints away from downstream users. If an upstream package has a problematic run_exports constraint, you can ignore it in your recipe by listing the upstream package name in the build/ignore_run_exports section:

build:
  ignore_run_exports:
    - libstdc++

You can also list the package the run_exports constraint is coming from using the build/ignore_run_exports_from section:

build:
  ignore_run_exports_from:
    - {{ compiler('cxx') }}

The pin_depends build key can be used to enforce pinning behavior on the output recipe or built package.

There are 2 possible behaviors:

build:
  pin_depends: record

With a value of record, conda-build will record all requirements exactly as they would be installed in a file called info/requires. These pins will not show up in the output of conda render and they will not affect the actual run dependencies of the output package. It is only adding in this new file.

build:
  pin_depends: strict

With a value of strict, conda-build applies the pins to the actual metadata. This does affect the output of conda render and also affects the end result of the build. The package dependencies will be strictly pinned down to the build string level. This will supersede any dynamic or compatible pinning that conda-build may otherwise be doing.

The overlinking_ignore_patterns key in the build section can be used to ignore patterns of files for the overlinking and overdepending checks. This is sometimes useful to speed up builds that have many files (large repackage jobs) or builds where you know only a small fraction of the files should be checked.

Glob patterns are allowed here, but mind your quoting, especially with leading wildcards.

Use this sparingly, as the overlinking checks generally do prevent you from making mistakes.

build:
  overlinking_ignore_patterns:
    - "bin/*"

The missing_dso_whitelist build key is a list of globs for dynamic shared object (DSO) files that should be ignored when examining linkage information.

During the post-build phase, the shared libraries in the newly created package are examined for linkages which are not provided by the package's requirements or a predefined list of system libraries. If such libraries are detected, either a warning --no-error-overlinking or error --error-overlinking will result.

build:
  missing_dso_whitelist:

These keys allow additions to the list of allowed libraries.

The runpath_whitelist build key is a list of globs for paths which are allowed to appear as runpaths in the package's shared libraries. All other runpaths will cause a warning message to be printed during the build.

build:
  runpath_whitelist:

Specifies the build and runtime requirements. Dependencies of these requirements are included automatically.

Versions for requirements must follow the conda match specification. See Package match specifications.

Tools required to build the package. These packages are run on the build system and include things such as revision control systems (Git, SVN) make tools (GNU make, Autotool, CMake) and compilers (real cross, pseudo-cross, or native when not cross-compiling), and any source pre-processors.

Packages which provide "sysroot" files, like the CDT packages (see below) also belong in the build section.

requirements:
  build:
    - git
    - cmake

This section was added in conda-build 3.0. It represents packages that need to be specific to the target platform when the target platform is not necessarily the same as the native build platform. For example, in order for a recipe to be "cross-capable", shared libraries requirements must be listed in the host section, rather than the build section, so that the shared libraries that get linked are ones for the target platform, rather than the native build platform. You should also include the base interpreter for packages that need one. In other words, a Python package would list python here and an R package would list mro-base or r-base.

requirements:
  build:
    - {{ compiler('c') }}
    - {{ cdt('xorg-x11-proto-devel') }}  # [linux]
  host:
    - python

NOTE:

When both build and host sections are defined, the build section can be thought of as "build tools" - things that run on the native platform, but output results for the target platform. For example, a cross-compiler that runs on linux-64, but targets linux-armv7.

The PREFIX environment variable points to the host prefix. With respect to activation during builds, both the host and build environments are activated. The build prefix is activated after the host prefix so that the build prefix, which always contains native executables for the running platform, has priority over the host prefix, which is not guaranteed to provide native executables (e.g. when cross-compiling).

As of conda-build 3.1.4, the build and host prefixes are always separate when both are defined, or when {{ compiler() }} Jinja2 functions are used. The only time that build and host are merged is when the host section is absent, and no {{ compiler() }} Jinja2 functions are used in meta.yaml. Because these are separate, you may see some build failures when migrating your recipes. For example, let's say you have a recipe to build a Python extension. If you add the compiler Jinja2 functions to the build section, but you do not move your Python dependency from the build section to the host section, your recipe will fail. It will fail because the host environment is where new files are detected, but because you have Python only in the build environment, your extension will be installed into the build environment. No files will be detected. Also, variables such as PYTHON will not be defined when Python is not installed into the host environment.

On Linux, using the compiler packages provided by Anaconda Inc. in the defaults meta-channel can prevent your build system leaking into the built software by using our CDT (Core Dependency Tree) packages for any "system" dependencies. These packages are repackaged libraries and headers from CentOS6 and are unpacked into the sysroot of our pseudo-cross compilers and are found by them automatically.

Note that what qualifies as a "system" dependency is a matter of opinion. The Anaconda Distribution chose not to provide X11 or GL packages, so we use CDT packages for X11. Conda-forge chose to provide X11 and GL packages.

On macOS, you can also use {{ compiler() }} to get compiler packages provided by Anaconda Inc. in the defaults meta-channel. The environment variables MACOSX_DEPLOYMENT_TARGET and CONDA_BUILD_SYSROOT will be set appropriately by conda-build (see Environment variables). CONDA_BUILD_SYSROOT will specify a folder containing a macOS SDK. These settings achieve backwards compatibility while still providing access to C++14 and C++17. Note that conda-build will set CONDA_BUILD_SYSROOT by parsing the conda_build_config.yaml. For more details, see Anaconda compiler tools.

TL;DR: If you use {{ compiler() }} Jinja2 to utilize our new compilers, you must also move anything that is not strictly a build tool into your host dependencies. This includes Python, Python libraries, and any shared libraries that you need to link against in your build. Examples of build tools include any {{ compiler() }}, Make, Autoconf, Perl (for running scripts, not installing Perl software), and Python (for running scripts, not for installing software).

Packages required to run the package. These are the dependencies that are installed automatically whenever the package is installed. Package names should follow the package match specifications.

requirements:
  run:
    - python
    - argparse # [py26]
    - six >=1.8.0

To build a recipe against different versions of NumPy and ensure that each version is part of the package dependencies, list numpy x.x as a requirement in meta.yaml and use conda-build with a NumPy version option such as --numpy 1.7.

The line in the meta.yaml file should literally say numpy x.x and should not have any numbers. If the meta.yaml file uses numpy x.x, it is required to use the --numpy option with conda-build.

requirements:
  run:
    - python
    - numpy x.x

NOTE:

Instead of manually specifying run requirements, since conda-build 3 you can augment the packages used in your build and host sections with run_exports which are then automatically added to the run requirements for you.

Packages that are optional at runtime but must obey the supplied additional constraint if they are installed.

Package names should follow the package match specifications.

requirements:
  run_constrained:
    - optional-subpackage =={{ version }}

For example, let's say we have an environment that has package "a" installed at version 1.0. If we install package "b" that has a run_constrained entry of "a>1.0", then conda would need to upgrade "a" in the environment in order to install "b".

This is especially useful in the context of virtual packages, where the run_constrained dependency is not a package that conda manages, but rather a virtual package that represents a system property that conda can't change. For example, a package on linux may impose a run_constrained dependency on __glibc>=2.12. This is the version bound consistent with CentOS 6. Software built against glibc 2.12 will be compatible with CentOS 6. This run_constrained dependency helps conda tell the user that a given package can't be installed if their system glibc version is too old.

If this section exists or if there is a run_test.[py,pl,sh,bat] file in the recipe, the package is installed into a test environment after the build is finished and the tests are run there.

Test files that are copied from the recipe into the temporary test directory and are needed during testing. If providing a path, forward slashes must be used.

test:
  files:
    - test-data.txt

Test files that are copied from the source work directory into the temporary test directory and are needed during testing.

test:
  source_files:
    - test-data.txt
    - some/directory
    - some/directory/pattern*.sh

This capability was added in conda-build 2.0.

In addition to the runtime requirements, you can specify requirements needed during testing. The runtime requirements that you specified in the "run" section described above are automatically included during testing.

test:
  requires:
    - nose

Commands that are run as part of the test.

test:
  commands:
    - bsdiff4 -h
    - bspatch4 -h

List of Python modules or packages that will be imported in the test environment.

test:
  imports:
    - bsdiff4

This would be equivalent to having a run_test.py with the following:

import bsdiff4

The script run_test.sh---or .bat, .py, or .pl---is run automatically if it is part of the recipe.

NOTE:

Knowing that your software built and ran its tests successfully is necessary, but not sufficient, for keeping whole systems of software running. To have confidence that a new build of a package hasn't broken other downstream software, conda-build supports the notion of downstream testing.

test:
  downstreams:
    - some_downstream_pkg

This is saying "When I build this recipe, after you run my test suite here, also download and run some_downstream_pkg which depends on my package." Conda-build takes care of ensuring that the package you just built gets installed into the environment for testing some_downstream_pkg. If conda-build can't create that environment due to unsatisfiable dependencies, it will skip those downstream tests and warn you. This usually happens when you are building a new version of a package that will require you to rebuild the downstream dependencies.

Downstreams specs are full conda specs, similar to the requirements section. You can put version constraints on your specs in here:

test:
  downstreams:
    - some_downstream_pkg  >=2.0

More than one package can be specified to run downstream tests for:

test:
  downstreams:
    - some_downstream_pkg
    - other_downstream_pkg

However, this does not mean that these packages are tested together. Rather, each of these are tested for satisfiability with your new package, then each of their test suites are run separately with the new package.

Explicitly specifies packaging steps. This section supports multiple outputs, as well as different package output types. The format is a list of mappings. Build strings for subpackages are determined by their runtime dependencies. This support was added in conda-build 2.1.0.

outputs:
  - name: some-subpackage
    version: 1.0
  - name: some-other-subpackage
    version: 2.0

NOTE:

If any output is specified in the outputs section, the default packaging behavior of conda-build is bypassed. In other words, if any subpackage is specified, then you do not get the normal top-level build for this recipe without explicitly defining a subpackage for it. This is an alternative to the existing behavior, not an addition to it. For more information, see Implicit metapackages. Each output may have its own version and requirements. Additionally, subpackages may impose downstream pinning similarly to Pin downstream to help keep your packages aligned.

You can specify files to be included in the package in 1 of 3 ways:

• Explicit file lists.
• Scripts that move files into the build prefix.
• Both of the above

Explicit file lists are relative paths from the root of the build prefix. Explicit file lists support glob expressions. Directory names are also supported, and they recursively include contents.

WARNING:

outputs:
  - name: subpackage-name
    files:
      - a-file
      - a-folder
      - *.some-extension
      - somefolder/*.some-extension

Greater control over file matching may be achieved by defining files as a dictionary separating files to include from those to exclude. When using include/exclude, only files installed by the current recipe are considered. i.e. files in the prefix installed by host dependencies are excluded. include/exclude must not be used simultaneously with glob expressions listed directly in outputs/files. Files matching both include and exclude expressions will be excluded.

outputs:
  - name: subpackage-name
    files:
      include:
        - a-file
        - a-folder
        - *.some-extension
        - somefolder/*.some-extension
      exclude:
        - *.exclude-extension
        - a-folder/**/*.some-extension

Scripts that create or move files into the build prefix can be any kind of script. Known script types need only specify the script name. Currently the list of recognized extensions is py, bat, ps1, and sh.

outputs:
  - name: subpackage-name
    script: move-files.py

The interpreter command must be specified if the file extension is not recognized.

outputs:
  - name: subpackage-name
    script: some-script.extension
    script_interpreter: program plus arguments to run script

For scripts that move or create files, a fresh copy of the working directory is provided at the start of each script execution. This ensures that results between scripts are independent of one another.

NOTE:

When both scripts and files are given, the script is first run and then only the files in the explicit file list are packaged.

Like a top-level recipe, a subpackage may have zero or more dependencies listed as build requirements and zero or more dependencies listed as run requirements.

The dependencies listed as subpackage build requirements are available only during the packaging phase of that subpackage.

A subpackage does not automatically inherit any dependencies from its top-level recipe, so any build or run requirements needed by the subpackage must be explicitly specified.

outputs:
  - name: subpackage-name
    requirements:
      build:
        - some-dep
      run:
        - some-dep

It is also possible for a subpackage requirements section to have a list of dependencies, but no build section or run section. This is the same as having a build section with this dependency list and a run section with the same dependency list.

outputs:
  - name: subpackage-name
    requirements:
      - some-dep

You can also impose runtime dependencies whenever a given (sub)package is installed as a build dependency. For example, if we had an overarching "compilers" package, and within that, had gcc and libgcc outputs, we could force recipes that use GCC to include a matching libgcc runtime requirement:

outputs:
  - name: gcc
    run_exports:
      - libgcc 2.*
  - name: libgcc

See the Export runtime requirements section for additional information.

NOTE:

When viewing the top-level package as a collection of smaller subpackages, it may be convenient to define the top-level package as a composition of several subpackages. If you do this and you do not define a subpackage name that matches the top-level package/name, conda-build creates a metapackage for you. This metapackage has runtime requirements drawn from its dependency subpackages, for the sake of accurate build strings.

EXAMPLE: In this example, a metapackage for subpackage-example will be created. It will have runtime dependencies on subpackage1, subpackage2, some-dep, and some-other-dep.

package:
  name: subpackage-example
  version: 1.0

requirements:
  run:
    - subpackage1
    - subpackage2

outputs:
  - name: subpackage1
    requirements:
      - some-dep
  - name: subpackage2
    requirements:
      - some-other-dep
  - name: subpackage3
    requirements:
      - some-totally-exotic-dep

You can test subpackages independently of the top-level package. Independent test script files for each separate package are specified under the subpackage's test section. These files support the same formats as the top-level run_test.* scripts, which are .py, .pl, .bat, and .sh. These may be extended to support other script types in the future.

outputs:
  - name: subpackage-name
    test:
      script: some-other-script.py

By default, the run_test.* scripts apply only to the top-level package. To apply them also to subpackages, list them explicitly in the script section:

outputs:
  - name: subpackage-name
    test:
      script: run_test.py

Test requirements for subpackages can be specified using the optional test/requires section of subpackage tests. Subpackage tests install their runtime requirements during the test as well.

EXAMPLE: In this example, the test for subpackage-name installs some-test-dep and subpackage-run-req, but not some-top-level-run-req.

requirements:
  run:
    - some-top-level-run-req

outputs:
  - name: subpackage-name
    requirements:
      - subpackage-run-req
    test:
      script: run_test.py
      requires:
        - some-test-dep

Conda-build supports creating packages other than conda packages. Currently that support includes only wheels, but others may come as demand appears. If type is not specified, the default value is conda.

requirements:
  build:
    - wheel

outputs:
  - name: name-of-wheel-package
    type: wheel

Currently you must include the wheel package in your top-level requirements/build section in order to build wheels.

When specifying type, the name field is optional and it defaults to the package/name field for the top-level recipe.

requirements:
  build:
    - wheel

outputs:
  - type: wheel

Conda-build currently knows how to test only conda packages. Conda-build does support using Twine to upload packages to PyPI. See the conda-build help output (conda-build --help) for the list of arguments accepted that will be passed through to Twine.

NOTE:

Specifies identifying information about the package. The information displays in the Anaconda.org channel.

about:
  home: https://github.com/ilanschnell/bsdiff4
  license: BSD 3-Clause
  license_file: LICENSE
  license_family: BSD
  license_url: https://github.com/bacchusrx/bsdiff4/blob/master/LICENSE
  summary: binary diff and patch using the BSDIFF4 format
  description: |
    This module provides an interface to the BSDIFF4 format, command line interfaces
    (bsdiff4, bspatch4) and tests.
  dev_url: https://github.com/ilanschnell/bsdiff4
  doc_url: https://bsdiff4.readthedocs.io
  doc_source_url: https://github.com/ilanschnell/bsdiff4/blob/main/README.rst

Add a file containing the software license to the package metadata. Many licenses require the license statement to be distributed with the package. The filename is relative to the source or recipe directory. The value can be a single filename or a YAML list for multiple license files. Values can also point to directories with license information. Directory entries must end with a / suffix (this is to lessen unintentional inclusion of non-license files; all of the directory's contents will be unconditionally and recursively added).

about:
  license_file:
    - LICENSE
    - vendor-licenses/

Similar to the license file, the user can add prelink message files to the conda package.

about:
  prelink_message:
    - prelink_message_file.txt
    - folder-with-all-prelink-messages/

If the app section is present, the package is an app, meaning that it appears in Anaconda Navigator.

The command that is called to launch the app in Navigator.

app:
  entry: ipython notebook

The icon file contained in the recipe.

app:
  icon: icon_64x64.png

Summary of the package used in Navigator.

app:
  summary:  "The Jupyter Notebook"

If True, installing the app through Navigator installs into its own environment. The default is False.

app:
  own_environment: True

A schema-free area for storing non-conda-specific metadata in standard YAML form.

EXAMPLE: To store recipe maintainer information:

extra:
  maintainers:
   - name of maintainer

Conda-build supports Jinja templating in the meta.yaml file.

EXAMPLE: The following meta.yaml would work with the GIT values defined for Git repositores. The recipe is included at the base directory of the Git repository, so the git_url is ../:

package:
  name: mypkg
  version: {{ GIT_DESCRIBE_TAG }}

build:
  number: {{ GIT_DESCRIBE_NUMBER }}

  # Note that this will override the default build string with the Python
  # and NumPy versions
  string: {{ GIT_BUILD_STR }}

source:
  git_url: ../

Conda-build checks if the Jinja2 variables that you use are defined and produces a clear error if it is not.

You can also use a different syntax for these environment variables that allows default values to be set, although it is somewhat more verbose.

EXAMPLE: A version of the previous example using the syntax that allows defaults:

package:
  name: mypkg
  version: {{ environ.get('GIT_DESCRIBE_TAG', '') }}

build:
  number: {{ environ.get('GIT_DESCRIBE_NUMBER', 0) }}

  # Note that this will override the default build string with the Python
  # and NumPy versions
  string: {{ environ.get('GIT_BUILD_STR', '') }}

source:
  git_url: ../

One further possibility using templating is obtaining data from your downloaded source code.

EXAMPLE: To process a project's setup.py and obtain the version and other metadata:

{% set data = load_setup_py_data() %}

package:
  name: conda-build-test-source-setup-py-data
  version: {{ data.get('version') }}

# source will be downloaded prior to filling in jinja templates
# Example assumes that this folder has setup.py in it
source:
  path_url: ../

These functions are completely compatible with any other variables such as Git and Mercurial.

Extending this arbitrarily to other functions requires that functions be predefined before Jinja processing, which in practice means changing the conda-build source code. See the conda-build issue tracker.

For more information, see the Jinja2 template documentation and the list of available environment variables.

Jinja templates are evaluated during the build process. To retrieve a fully rendered meta.yaml, use the conda render command.

There are several additional functions available to Jinja2, which can be used to load data from other files. These are load_setup_py_data, load_file_regex, load_file_data, and load_str_data.

• load_setup_py_data: Load data from a setup.py file. This can be useful to obtain metadata such as the version from a project's setup.py file. For example:

  {% set data = load_setup_py_data() %}
  {% set version = data.get('version') %}
  package:
    name: foo
    version: {{ version }}

• load_file_regex: Search a file for a regular expression returning the first match as a Python re.Match object.

  For example, using load_file_regex(load_file, regex_pattern, from_recipe_dir=False) -> re.Match | None:

  {% set version_match = load_file_regex(
    load_file="conda_package_streaming/__init__.py",
    regex_pattern='^__version__ = "(.+)"') %}
  {% set version = version_match[1] %}
  
  package:
    version: {{ version }}

• load_file_data: Parse JSON, TOML, or YAML files and load data from them. For example, you can use this to load poetry configurations from pyproject.toml. This is especially useful, as setup.py is no longer the only standard way to define project metadata (see PEP 517 and PEP 518):

  {% set pyproject = load_file_data('pyproject.toml') %}
  {% set poetry = pyproject.get('tool', {}).get('poetry') %}
  package:
    name: {{ poetry.get('name') }}
    version: {{ poetry.get('version') }}

• load_str_data: Load and parse data from a string. This is similar to load_file_data, but it takes a string instead of a file as an argument. This may seem pointless at first, but you can use this to pass more complex data structures by environment variables. For example:

  {% set extra_deps = load_str_data(environ.get("EXTRA_DEPS", []), "json") %}
  requirements:
    run:
      - ...
      {% for dep in extra_deps %}
      - {{ dep }}
      {% endfor %}

  Then you can pass the EXTRA_DEPS environment variable to the build like so:

  EXTRA_DEPS='["foo =1.0", "bar >=2.0"]' conda build path/to/recipe

The functions load_setup_py_data, load_file_regex, and load_file_data all take the parameters from_recipe_dir and recipe_dir. If from_recipe_dir is set to true, then recipe_dir must also be passed. In that case, the file in question will be searched for relative to the recipe directory. Otherwise the file is searched for in the source (after it is downloaded and extracted, if necessary). If the given file is an absolute path, neither of the two directories are searched.

The functions load_file_data and load_str_data also accept *args and **kwargs which are passed verbatim to the function used to parse the file. For JSON this would be json.load; for TOML, toml.load; and for YAML yaml.safe_load.

Besides the default Jinja2 functionality, additional Jinja functions are available during the conda-build process: pin_compatible, pin_subpackage, compiler, and resolved_packages. Please see Extra Jinja2 functions for the definition of the first 3 functions. Definition of resolved_packages is given below:

• resolved_packages('environment_name'): Returns the final list of packages (in the form of package_name version build_string) that are listed in requirements:host or requirements:build. This includes all packages (including the indirect dependencies) that will be installed in the host or build environment. environment_name must be either host or build. This function is useful for creating meta-packages that will want to pin all of their direct and indirect dependencies to their exact match. For example:

  requirements:
    host:
      - curl 7.55.1
    run:
    {% for package in resolved_packages('host') %}
      - {{ package }}
    {% endfor %}

  might render to (depending on package dependencies and the platform):

  requirements:
      host:
          - curl 7.55.1
      run:
          - ca-certificates 2017.08.26 h1d4fec5_0
          - curl 7.55.1 h78862de_4
          - libgcc-ng 7.2.0 h7cc24e2_2
          - libssh2 1.8.0 h9cfc8f7_4
          - openssl 1.0.2n hb7f436b_0
          - zlib 1.2.11 ha838bed_2

  Here, output of resolved_packages was:

  ['ca-certificates 2017.08.26 h1d4fec5_0', 'curl 7.55.1 h78862de_4',
  'libgcc-ng 7.2.0 h7cc24e2_2', 'libssh2 1.8.0 h9cfc8f7_4',
  'openssl 1.0.2n hb7f436b_0', 'zlib 1.2.11 ha838bed_2']

You can add selectors to any line, which are used as part of a preprocessing stage. Before the meta.yaml file is read, each selector is evaluated and if it is False, the line that it is on is removed. A selector has the form # [<selector>] at the end of a line.

source:
  url: http://path/to/unix/source    # [not win]
  url: http://path/to/windows/source # [win]

NOTE:

Preprocessing selectors are evaluated after Jinja templates.

A selector is a valid Python statement that is executed. The following variables are defined. Unless otherwise stated, the variables are booleans.

The use of the Python version selectors, py27, py34, etc. is discouraged in favor of the more general comparison operators.  Additional selectors in this series will not be added to conda-build.

Note that for each subdir with OS and architecture that conda supports, two preprocessing selectors are created for the OS and the architecture separately except when the architecture is not a valid python expression (*-32 and *-64 in particular).

Because the selector is any valid Python expression, complicated logic is possible:

source:
  url: http://path/to/windows/source      # [win]
  url: http://path/to/python2/unix/source # [unix and py2k]
  url: http://path/to/python3/unix/source # [unix and py>=35]

NOTE:

The selectors delete only the line that they are on, so you may need to put the same selector on multiple lines:

source:
  url: http://path/to/windows/source     # [win]
  md5: 30fbf531409a18a48b1be249052e242a  # [win]
  url: http://path/to/unix/source        # [unix]
  md5: 88510902197cba0d1ab4791e0f41a66e  # [unix]

NOTE:

To select multiple operating systems use the or statement. While it might be tempting to use skip: True  # [win and osx], this will only work if the platform is both windows and osx simultaneously (i.e. never).

build:
   skip: True  # [win or osx]

You can add scripts to a recipe. They must be located in the same directory as the meta.yaml file. The following scripts can be added:

• pre-link---Executed before the package is installed. An error is indicated by a nonzero exit and causes conda to stop and causes the installation to fail.
• post-link---Executed after the package is installed. An error is indicated by a nonzero exist and causes installation to fail. If there is an error, conda does not write any package metadata.
• pre-unlink---Executed before the package is removed. An error is indicated by a nonzero exist and causes the removal to fail.

In addition to being co-located with the meta.yaml file, they must be named simply post-link.sh or post-link.bat. Conda-build will rename them to .<name>-<action>.sh (or .bat) where <name> is the package name and <action> is one of the preceeding actions.

These scripts are executed in a subprocess by conda, using %COMSPEC% /c <script> on Windows and /bin/bash <script> on macOS and Linux.

The convention for the path and filenames of these scripts on Windows is:

Scripts/.<name>-<action>.bat

On Linux and macOS the convention is:

bin/.<name>-<action>.sh

The scripts set the following environment variables:

The scripts are:

• Windows:

  • pre-link.bat
  • post-link.bat
  • pre-unlink.bat

• macOS and Linux:

  • pre-link.sh
  • post-link.sh
  • pre-unlink.sh

Post-link and pre-unlink scripts should:

• Be avoided whenever possible.
• Not touch anything other than the files being installed.
• Not write anything to stdout or stderr, unless an error occurs.
• Not depend on any installed or to-be-installed conda packages.
• Depend only on simple system tools such as rm, cp, mv, and ln.

The scripts should not write to stdout or stderr unless an error occurs, but they may write to $PREFIX/.messages.txt, which is shown after conda completes all actions.

Recipes are allowed to have activate scripts which will be sourced or called when the environment is activated. It is generally recommended to avoid using activate scripts when another option is possible because people do not always activate environments the expected way and these packages may then misbehave.

When using them in a recipe, feel free to name them activate.bat, activate.sh, deactivate.bat, and deactivate.sh in the recipe. The installed scripts are recommended to be prefixed by the package name and a separating -.

Below is some sample code for Unix and Windows that will make this install process easier.

In build.sh:

# Copy the [de]activate scripts to $PREFIX/etc/conda/[de]activate.d.
# This will allow them to be run on environment activation.
for CHANGE in "activate" "deactivate"
do
    mkdir -p "${PREFIX}/etc/conda/${CHANGE}.d"
    cp "${RECIPE_DIR}/${CHANGE}.sh" "${PREFIX}/etc/conda/${CHANGE}.d/${PKG_NAME}_${CHANGE}.sh"
done

In build.bat:

setlocal EnableDelayedExpansion

:: Copy the [de]activate scripts to %PREFIX%\etc\conda\[de]activate.d.
:: This will allow them to be run on environment activation.
for %%F in (activate deactivate) DO (
    if not exist %PREFIX%\etc\conda\%%F.d mkdir %PREFIX%\etc\conda\%%F.d
    copy %RECIPE_DIR%\%%F.bat %PREFIX%\etc\conda\%%F.d\%PKG_NAME%_%%F.bat
    :: Copy unix shell activation scripts, needed by Windows Bash users
    copy %RECIPE_DIR%\%%F.sh %PREFIX%\etc\conda\%%F.d\%PKG_NAME%_%%F.sh
)

Often, the most difficult thing about building a conda package is making it relocatable. Relocatable means that the package can be installed into any prefix. Otherwise, the package would be usable only in the environment in which it was built.

Conda-build does the following things automatically to make packages relocatable:

• Binary object files are converted to use relative paths using install_name_tool on macOS and patchelf on Linux.
• Any text file without NULL bytes that contains the build prefix or the placeholder prefix /opt/anaconda1anaconda2anaconda3 is registered in the info/has_prefix file in the package metadata. When conda installs the package, any files in info/has_prefix have the registered prefix replaced with the install prefix. For more information, see Package metadata.
• Any binary file containing the build prefix can automatically be registered in info/has_prefix using build/detect_binary_files_with_prefix in meta.yaml. Alternatively, individual binary files can be registered by listing them in build/binary_has_prefix_files in meta.yaml. The registered files will have their build prefix replaced with the install prefix at install time. This works by padding the install prefix with null terminators, such that the length of the binary file remains the same. The build prefix must therefore be long enough to accommodate any reasonable installation prefix. On macOS and Linux, conda-build pads the build prefix to 255 characters by appending _placeholds to the end of the build directory name.

NOTE:

The prefix length was changed in conda-build 2.0 from 80 characters to 255 characters. Legacy packages with 80-character prefixes must be rebuilt to take advantage of the longer prefix.

• There may be cases where conda identified a file as binary, but it needs to have the build prefix replaced as if it were text---no padding with null terminators. Such files can be listed in build/has_prefix_files in meta.yaml.

A conda package is an archive file that contains:

• Metadata under the info/ directory.
• A collection of files that are installed directly into an install prefix.

There are currently two formats of archives that are supported:

The formats are identical across platforms and operating systems. During the install process, all files are extracted into the install prefix, with the exception of the ones in info/. Installing a conda package into an environment is similar to executing the following commands:

cd <environment prefix>
tar xjf some-package-1.0-0.tar.bz2

Only files, including symbolic links, are part of a conda package. Directories are not included. Directories are created and removed as needed, but you cannot create an empty directory from the tar archive directly.

The info/ directory contains all metadata about a package. Files in this location are not installed under the install prefix. Although you are free to add any file to this directory, conda only inspects the content of the files discussed below.

This file contains basic information about the package, such as name, version, build string, and dependencies. The content of this file is stored in repodata.json, which is the repository index file, hence the name index.json. The JSON object is a dictionary containing the keys shown below. The filename of the conda package is composed of the first 3 values, as in: <name>-<version>-<build>.tar.bz2 or <name>-<version>-<build>.conda.

Lists all files that are part of the package itself, 1 per line. All of these files need to get linked into the environment. Any files in the package that are not listed in this file are not linked when the package is installed. The directory delimiter for the files in info/files should always be "/", even on Windows. This matches the directory delimiter used in the tarball.

Optional file. Lists all files that contain a hard-coded build prefix or placeholder prefix, which needs to be replaced by the install prefix at installation time.

NOTE:

Each line of this file should be either a path, in which case it is considered a text file with the default placeholder /opt/anaconda1anaconda2anaconda3, or a space-separated list of placeholder, mode, and path, where:

• Placeholder is the build or placeholder prefix.
• Mode is either text or binary.
• Path is the relative path of the file to be updated.

EXAMPLE: On Windows:

"Scripts/script1.py"
"C:\Users\username\anaconda\envs\_build" text "Scripts/script2.bat"
"C:/Users/username/anaconda/envs/_build" binary "Scripts/binary"

EXAMPLE: On macOS or Linux:

bin/script.sh
/Users/username/anaconda/envs/_build binary bin/binary
/Users/username/anaconda/envs/_build text share/text

NOTE:

The directory delimiter for the relative path must always be "/", even on Windows. The placeholder may contain either "\" or "/" on Windows, but the replacement prefix will match the delimiter used in the placeholder. The default placeholder /opt/anaconda1anaconda2anaconda3 is an exception, being replaced with the install prefix using the native path delimiter. On Windows, the placeholder and path always appear in quotes to support paths with spaces.

Optional file. The software license for the package.

Optional file. Lists all files that cannot be linked---either soft-linked or hard-linked---into environments and are copied instead.

Optional file. Contains the entries in the About section of the meta.yaml file. The following keys are added to info/about.json if present in the build recipe:

• home.
• dev_url.
• doc_url.
• license_url.
• license.
• summary.
• description.
• license_family.

A directory containing the full contents of the build recipe.

The fully rendered build recipe. See conda render.

This directory is present only when the the include_recipe flag is True in the Build section.

You may optionally execute scripts before and after the link and unlink steps. For more information, see Adding pre-link, post-link, and pre-unlink scripts.

A conda repository---or channel---is a directory tree, usually served over HTTPS, which has platform subdirectories, each of which contains conda packages and a repository index. The index file repodata.json lists all conda packages in the platform subdirectory. Use conda index to create such an index from the conda packages within a directory. It is simple mapping of the full conda package filename to the dictionary object in info/index.json described in Adding pre-link, post-link, and pre-unlink scripts.

In the following example, a repository provides the conda package misc-1.0-np17py27_0.tar.bz2 on 64-bit Linux and 32-bit Windows:

<some path>/linux-64/repodata.json
                     repodata.json.bz2
                     misc-1.0-np17py27_0.tar.bz2
           /win-32/repodata.json
                   repodata.json.bz2
                   misc-1.0-np17py27_0.tar.bz2

NOTE:

Both conda packages have identical filenames and are distinguished only by the repository subdirectory that contains them.

This match specification is not the same as the syntax used at the command line with conda install, such as conda install python=3.4. Internally, conda translates the command line syntax to the spec defined in this section.

EXAMPLE: python=3.4 is translated to python 3.4.*. conda search 'python=3.1' does NOT bring up Python 3.10, only Python 3.1.*.

Package dependencies are specified using a match specification. A match specification is a space-separated string of 1, 2, or 3 parts:

• The first part is always the exact name of the package.
• The second part refers to the version and may contain special characters. See table below.
• The third part is always the exact build string. When there are three parts, the second part must be the exact version.

HINT:

NOTE:

For package match specifications, pre-release versioning is also supported such that >1.0b4 will match 1.0b5 and 1.0rc1 but not 1.0b4 or 1.0a5.

Conda parses the version by splitting it into parts separated by |. If the part begins with <, >, =, or !, it is parsed as a relational operator. Otherwise, it is parsed as a version, possibly containing the "*" operator.

Remember that the version specification cannot contain spaces, as spaces are used to delimit the package, version, and build string in the whole match specification. python >= 2.7 is an invalid match specification. However, "python >= 2.7" (with double or single quotes) is matched as any version of a package named python>=2.7.

The build string constraint "numpy=1.11.2=*nomkl*" matches the NumPy 1.11.2 packages without MKL, but not the normal MKL NumPy 1.11.2 packages.

The build string constraint "numpy=1.11.1|1.11.3=py36_0" matches NumPy 1.11.1 or 1.11.3 built for Python 3.6, but not any versions of NumPy built for Python 3.5 or Python 2.7.

The following are all valid match specifications for numpy-1.8.1-py27_0:

• numpy
• numpy 1.8*
• numpy 1.8.1
• numpy >=1.8
• numpy ==1.8.1
• numpy 1.8|1.8*
• numpy >=1.8,<2
• numpy >=1.8,<2|1.9
• numpy 1.8.1 py27_0
• numpy=1.8.1=py27_0

When using the command line, put double or single quotes around any package version specification that contains the space character or any of the following characters: <, >, *, or |.

Shared libraries are libraries that are loosely coupled to the programs and extensions that depend on them. When loading an executable into memory, an operating system finds all dependent shared libraries and links them to the executable so that it can run.

Windows, macOS, and Linux all provide a way to build executables and libraries that contain links to the shared libraries they depend on, instead of directly linking the libraries themselves.

Unlike macOS and Linux, Windows does not have the concept of embedding links into binaries. Instead, Windows depends primarily on searching directories for matching filenames, as documented in Search Path Used by Windows to Locate a DLL.

There is an alternate configuration, called side-by-side assemblies, that requires specification of DLL versions in either an embedded manifest or an appropriately named XML file alongside the binary in question. Conda does not currently use side-by-side assemblies, but it may turn towards that in the future to resolve complications with multiple versions of the same library on the same system.

For now, most DLLs are installed into (install prefix)\\Library\\bin. This path is added to os.environ["PATH"] for all Python processes, so that DLLs can be located, regardless of the value of the system's PATH environment variable.

NOTE:

In macOS and Linux, dynamic links are discovered in a similar manner to the way that Python modules are discovered via PYTHONPATH, and executables are discovered via PATH. A list of search locations is made, and then the library objects are searched for in the search locations. By default, as well as by design, the system dynamic linker does not have any special preference for the conda environment lib directories.

You can specify both absolute links and relative links. If the links are absolute paths, such as /Users/UserName/my_build_env, the library works only on a system where that exact path exists. Therefore, relative links are preferred in conda packages.

Relative links require a special variable in the link itself:

• On Linux, the $ORIGIN variable allows you to specify "relative to this file as it is being executed".

• On macOS, the variables are:

  • @rpath---Allows you to set relative links from the system load paths.
  • @loader_path---Equivalent to $ORIGIN.
  • @executable_path---Supports the Apple .app directory approach, where libraries know where they live relative to their calling application.

Conda-build uses @loader_path on macOS and $ORIGIN on Linux because we install into a common root directory and can assume that other libraries are also installed into that root. The use of the variables allows you to build relocatable binaries that can be built on one system and sent everywhere.

On Linux, conda-build modifies any shared libraries or generated executables to use a relative dynamic link by calling the patchelf tool. On macOS, the install_name_tool tool is used.

WARNING:

EXAMPLE: You install an old version of libcurl into your conda environment due to some compatibility issues with the code you're using. Then, you set export LD_LIBRARY_PATH=/home/UserName/envs/curl_env/lib. From that point on, every program that you execute in that session will favor this libcurl to your system libcurl because it is now effectively at the "front" of the dynamic load path.

Including conda environment paths in LD_LIBRARY_PATH or DYLD_LIBRARY_PATH is not recommended.

The nature of binary compatibility (and incompatibility) means that we sometimes need to build binary packages (and any package containing binaries) with several variants to support different usage environments. For example, using NumPy's C API means that a package must be used with the same version of NumPy at runtime that was used at build time.

There has been limited support for this for a long time. Including Python in both build and run requirements resulted in a package with Python pinned to the version of Python used at build time, and a corresponding addition to the filename such as "py27". Similar support existed for NumPy with the addition of an x.x pin in the recipe after Conda-build PR 573 was merged. Before conda-build version 3.0, there were also many longstanding proposals for general support (Conda-build issue 1142).

As of conda-build 3.0, a new configuration scheme has been added, dubbed "variants." Conceptually, this decouples pinning values from recipes, replacing them with Jinja2 template variables. It adds support for the notion of "compatible" pinnings to be integrated with ABI compatibility databases, such as ABI Laboratory. Note that the concept of "compatible" pinnings is currently still under heavy development.

Variant input is ultimately a dictionary. These dictionaries are mostly very flat. Keys are made directly available in Jinja2 templates. As a result, keys in the dictionary (and in files read into dictionaries) must be valid jinja2 variable names (no - characters allowed). This example builds Python 2.7 and 3.5 packages in one build command:

conda_build_config.yaml like:

python:
    - 2.7
    - 3.5

meta.yaml contents like:

package:
    name: compiled-code
    version: 1.0

requirements:
    build:
        - python
    run:
        - python

The command to build recipes is unchanged relative to earlier conda-build versions. For example, with our shell in the same folder as meta.yaml and conda_build_config.yaml, we just call the conda build . command.

There are a few characteristic use cases for pinning.  Please consider this a map for the content below.

1. Shared library providing a binary interface. All uses of this library use the binary interface. It is convenient to apply the same pin to all of your builds. Example: boost

   conda_build_config.yaml in your HOME folder:

   boost:
     - 1.61
     - 1.63
   pin_run_as_build:
     boost: x.x

   meta.yaml:

   package:
       name: compiled-code
       version: 1.0
   
   requirements:
       build:
           - boost
       run:
           - boost

   This example demonstrates several features:

   • User-wide configuration with a specifically named config file (conda_build_config.yaml in your home folder). More options below in Creating conda-build variant config files.
   • Building against multiple versions of a single library (set versions installed at build time).
   • Pinning runtime requirements to the version used at build time. More information below at Pinning at the variant level.
   • Specify granularity of pinning. x.x pins major and minor version. More information at Pinning expressions.

2. Python package with externally accessible binary component. Not all uses of this library use the binary interface (some only use pure Python). Example: NumPy.

   conda_build_config.yaml in your recipe folder (alongside meta.yaml):

   numpy:
     - 1.11
     - 1.12

   meta.yaml:

   package:
       name: numpy_using_pythonAPI_thing
       version: 1.0
   
   requirements:
       build:
           - python
           - numpy
       run:
           - python
           - numpy

   This example demonstrates a particular feature: reduction of builds when pins are unnecessary. Since the example recipe above only requires the Python API to NumPy, we will only build the package once and the version of NumPy will not be pinned at runtime to match the compile-time version.  There's more information at Avoiding unnecessary builds.

   For a different package that makes use of the NumPy C API, we will need to actually pin NumPy in this recipe (and only in this recipe, so that other recipes don't unnecessarily build lots of variants).  To pin NumPy, you can use the variant key directly in meta.yaml:

   package:
       name: numpy_using_cAPI_thing
       version: 1.0
   
   requirements:
       build:
           - numpy
       run:
           - numpy

   For legacy compatibility, Python is pinned implicitly without specifying {{ python }} in your recipe. This is generally intractable to extend to all package names, so in general, try to get in the habit of always using the Jinja2 variable substitution for pinning using versions from your conda_build_config.yaml file.

   There are also more flexible ways to pin using the Pinning expressions. See Pinning at the recipe level for examples.

3. One recipe splits into multiple packages, and package dependencies need to be dynamically pinned among one another. Example: GCC/libgcc/libstdc++/gfortran/etc.

   The dynamic pinning is the tricky part. Conda-build provides new ways to refer to other subpackages within a single recipe.

   package:
       name: dynamic_supackage
       version: 1.0
   
   requirements:
       run:
           - {{ pin_subpackage('my_awesome_subpackage') }}
   
   outputs:
     - name: my_awesome_subpackage
       version: 2.0

   By referring to subpackages this way, you don't need to worry about what the end version of my_awesome_subpackage will be. Update it independently and just let conda-build figure it out and keep things consistent. There's more information below in the Referencing subpackages section.

Let's say we have a set of recipes that currently builds a C library, as well as Python and R bindings to that C library. xgboost, a recent machine learning library, is one such example. Under conda-build 2.0 and earlier, you needed to have 3 recipes - 1 for each component. Let's go over some simplified meta.yaml files. First, the C library:

package:
    name: libxgboost
    version: 1.0

Next, the Python bindings:

package:
    name: py-xgboost
    version: 1.0

requirements:
    build:
        - libxgboost  # you probably want to pin the version here, but there's no dynamic way to do it
        - python
    run:
        - libxgboost  # you probably want to pin the version here, but there's no dynamic way to do it
        - python

package:
    name: r-xgboost
    version: 1.0

requirements:
    build:
        - libxgboost  # you probably want to pin the version here, but there's no dynamic way to do it
        - r-base
    run:
        - libxgboost  # you probably want to pin the version here, but there's no dynamic way to do it
        - r-base

To build these, you'd need several conda-build commands, or a tool like conda-build-all to build out the various Python versions. With conda-build 3.0 and split packages from conda-build 2.1, we can simplify this to one coherent recipe that also includes the matrix of all desired Python and R builds.

First, the meta.yaml file:

package:
    name: xgboost
    version: 1.0

outputs:
    - name: libxgboost
    - name: py-xgboost
      requirements:
          - {{ pin_subpackage('libxgboost', exact=True) }}
          - python

    - name: r-xgboost
      requirements:
          - {{ pin_subpackage('libxgboost', exact=True) }}
          - r-base

Next, the conda_build_config.yaml file, specifying our build matrix:

python:
    - 2.7
    - 3.5
    - 3.6
r_base:
    - 3.3.2
    - 3.4.0

With this updated method, you get a complete build matrix: 6 builds total. One libxgboost library, 3 Python versions, and 2 R versions. Additionally, the Python and R packages will have exact pins to the libxgboost package that was built by this recipe.

Variant input files are yaml files. Search order for these files is the following:

1. A file named conda_build_config.yaml in the user's HOME folder (or an arbitrarily named file specified as the value for the conda_build/config_file key in your .condarc file).
2. A file named conda_build_config.yaml in the current working directory.
3. A file named conda_build_config.yaml in the same folder as meta.yaml with your recipe.
4. Any additional files specified on the command line with the --variant-config-files or -m command line flags, which can be passed multiple times for multiple files. The conda build and conda render commands accept these arguments.

Values in files found later in this search order will overwrite and replace the values from earlier files.

NOTE:

conda_build:
  config_file: some/path/to/file

Ultimately, a variant is just a dictionary. This dictionary is provided directly to Jinja2 and you can use any declared key from your variant configuration in your Jinja2 templates. There are two ways that you can feed this information into the API:

1. Pass the variants keyword argument to API functions. Currently, the build, render, get_output_file_paths, and check functions accept this argument. variants should be a dictionary where each value is a list of versions to iterate over. These are aggregated as detailed in the Aggregation of multiple variants section below.
2. Set the variant member of a Config object. This is just a dictionary. The values for fields should be strings or lists of strings, except "extended keys", which are documented in the Extended keys section below.

Again, with meta.yaml contents like:

package:
    name: compiled-code
    version: 1.0

requirements:
    build:
        - python
    run:
        - python

You could supply a variant to build this recipe like so:

variants = {"python": ["2.7", "3.5"]}
api.build(path_to_recipe, variants=variants)

Note that these Jinja2 variable substitutions are not limited to version numbers. You can use them anywhere, for any string value. For example, to build against different MPI implementations:

With meta.yaml contents like:

package:
    name: compiled-code
    version: 1.0

requirements:
    build:
        - {{ mpi }}
    run:
        - {{ mpi }}

You could supply a variant to build this recipe like this (conda_build_config.yaml):

mpi:
    - openmpi  # version spec here is totally valid, and will apply in the recipe
    - mpich  # version spec here is totally valid, and will apply in the recipe

Selectors are valid in conda_build_config.yaml, so you can have one conda_build_config.yaml for multiple platforms:

mpi:
    - openmpi  # [osx]
    - mpich    # [linux]
    - msmpi    # [win]

Jinja is not allowed in conda_build_config.yaml, though. It is the source of information to feed into other Jinja templates, and the buck has to stop somewhere.

A critical part of any build system is ensuring that you can reproduce the same output at some future point in time. This is often essential for troubleshooting bugs. For example, if a package contains only binaries, it is helpful to understand what source code created those binaries, and thus what bugs might be present.

Since conda-build 2.0, conda-build has recorded its rendered meta.yaml files into the info/recipe folder of each package it builds. Conda-build 3.0 is no different in this regard, but the meta.yaml that is recorded is a frozen set of the variables that make up the variant for that build.

NOTE:

There are some special keys that behave differently and can be more nested:

• zip_keys: a list of strings or a list of lists of strings. Strings are keys in variant. These couple groups of keys, so that particular keys are paired, rather than forming a matrix. This is useful, for example, to couple vc version to Python version on Windows. More info below in the Coupling keys section.
• pin_run_as_build: should be a dictionary. Keys are package names. Values are "pinning expressions" - explained in more detail in Customizing compatibility. This is a generalization of the numpy x.x spec, so that you can pin your packages dynamically based on the versions used at build time.
• extend_keys: specifies keys that should be aggregated, and not replaced, by later variants. These are detailed below in the Extended keys section.
• ignore_version: list of package names whose versions should be excluded from meta.yaml's requirements/build when computing hash. Described further in Avoiding unnecessary builds.

Sometimes particular versions need to be tied to other versions. For example, on Windows, we generally follow the upstream Python.org association of Visual Studio compiler version with Python version. Python 2.7 is always compiled with Visual Studio 2008 (also known as MSVC 9). We don't want a conda_build_config.yaml like the following to create a matrix of Python/MSVC versions:

python:
  - 2.7
  - 3.5
vc:
  - 9
  - 14

Instead, we want 2.7 to be associated with 9, and 3.5 to be associated with 14. The zip_keys key in conda_build_config.yaml is the way to achieve this:

python:
  - 2.7
  - 3.5
vc:
  - 9
  - 14
zip_keys:
  - python
  - vc

You can also have nested lists to achieve multiple groups of zip_keys:

zip_keys:
  -
    - python
    - vc
  -
    - numpy
    - blas

The rules for zip_keys are:

1. Every list in a group must be the same length. This is because without equal length, there is no way to associate earlier elements from the shorter list with later elements in the longer list. For example, this is invalid, and will raise an error:

   python:
     - 2.7
     - 3.5
   vc:
     - 9
   zip_keys:
     - python
     - vc

2. zip_keys must be either a list of strings, or a list of lists of strings. You can't mix them. For example, this is an error:

   zip_keys:
     -
       - python
       - vc
     - numpy
     - blas

Rule #1 raises an interesting use case: How does one combine CLI flags like --python with zip_keys? Such a CLI flag will change the variant so that it has only a single entry, but it will not change the vc entry in the variant configuration. We'll end up with mismatched list lengths, and an error. To overcome this, you should instead write a very simple YAML file with all involved keys. Let's call it python27.yaml, to reflect its intent:

python:
  - 2.7
vc:
  - 9

Provide this file as a command-line argument:

conda build recipe -m python27.yaml

You can also specify variants in JSON notation from the CLI as detailed in the CONDA_* variables and command line arguments to conda-build section. For example:

conda build recipe --variants "{'python': ['2.7', '3.5'], 'vc': ['9', '14']}"

To avoid building variants of packages where pinning does not require having different builds, you can use the ignore_version key in your variant. Then all variants are evaluated, but if any hashes are the same, then they are considered duplicates, and are deduplicated. By omitting some packages from the build dependencies, we can avoid creating unnecessarily specific hashes and allow this deduplication.

For example, let's consider a package that uses NumPy in both run and build requirements, and a variant that includes 2 NumPy versions:

variants = [{"numpy": ["1.10", "1.11"], "ignore_version": ["numpy"]}]

meta.yaml:

requirements:
    build:
        - numpy
    run:
        - numpy

Here, the variant says that we'll have two builds - one for each NumPy version. However, since this recipe does not pin NumPy's run requirement (because it doesn't utilize NumPy's C API), it is unnecessary to build it against both NumPy 1.10 and 1.11.

The rendered form of this recipe, with conda-build ignoring NumPy's value in the recipe, is going to be just one build that looks like:

meta.yaml:

requirements:
    build:
        - numpy
    run:
        - numpy

ignore_version is an empty list by default. The actual build performed is probably done with the last 'numpy' list element in the variant, but that's an implementation detail that you should not depend on. The order is considered unspecified behavior because the output should be independent of the input versions.

WARNING:

Any pinning done in the run requirements will affect the hash, and thus builds will be done for each variant in the matrix. Any package that sometimes is used for its compiled interface and sometimes used for only its Python interface may benefit from careful use of ignore_version in the latter case.

NOTE:

To ensure consistency with existing users of conda-build, environment variables such as CONDA_PY behave as they always have, and they overwrite all variants set in files or passed to the API.

The full list of respected environment variables are:

• CONDA_PY
• CONDA_NPY
• CONDA_R
• CONDA_PERL
• CONDA_LUA

CLI flags are also still available. These are sticking around for their usefulness in one-off jobs.

• --python
• --numpy
• --R
• --perl
• --lua

In addition to these traditional options, there's one new flag to specify variants: --variants. This flag accepts a string of JSON-formatted text. For example:

conda build recipe --variants "{python: [2.7, 3.5], vc: [9, 14]}"

The matrix of all variants is first consolidated from several dicts of lists into a single dict of lists, and then transformed in a list of dicts (using the Cartesian product of lists), where each value is a single string from the list of potential values.

For example, general input for variants could be something like:

a = {"python": ["2.7", "3.5"], "numpy": ["1.10", "1.11"]}
# values can be strings or lists.  Strings are converted to one-element lists internally.
b = {"python": ["3.4", "3.5"], "numpy": "1.11"}

Here, let's say b is found after a, and thus has priority over a. Merging these 2 variants yields:

merged = {"python": ["3.4", "3.5"], "numpy": ["1.11"]}

b's values for python have overwritten a's. From here, we compute the Cartesian product of all input variables. The end result is a collection of dicts, each with a string for each value. Output would be something like:

variants = [{"python": "3.4", "numpy": "1.11"}, {"python": "3.5", "numpy": "1.11"}]

conda-build would loop over these variants where appropriate, such as when building, outputting package output names, and so on.

If numpy had had two values instead of one, we'd end up with four output variants: 2 variants for python, times 2 variants for numpy:

variants = [
    {"python": "3.4", "numpy": "1.11"},
    {"python": "3.5", "numpy": "1.11"},
    {"python": "3.4", "numpy": "1.10"},
    {"python": "3.5", "numpy": "1.10"},
]

To establish your initial variant, you may point to an existing conda environment. Conda-build will examine the contents of that environment and pin to the exact requirements that make up that environment.

conda build --bootstrap name_of_env

You may specify either environment name or filesystem path to the environment. Note that specifying environment name does mean depending on conda's environment lookup.

These are not looped over to establish the build matrix. Rather, they are aggregated from all input variants, and each derived variant shares the whole set. These are used internally for tracking which requirements should be pinned, for example, with the pin_run_as_build key. You can add your own extended keys by passing in values for the extend_keys key for any variant.

For example, if you wanted to collect some aggregate trait from multiple conda_build_config.yaml files, you could do something like this:

HOME/conda_build_config.yaml:

some_trait:
  - dog
extend_keys:
  - some_trait

recipe/conda_build_config.yaml:

some_trait:
  - pony
extend_keys:
  - some_trait

Note that both of the conda_build_config.yaml files need to list the trait as an extend_keys entry.  If you list it in only one of them, an error will be raised to avoid confusion with one conda_build_config.yaml file that would add entries to the build matrix, and another which would not. For example, this should raise an error:

some_trait:
  - dog

recipe/conda_build_config.yaml:

some_trait:
  - pony
extend_keys:
  - some_trait

When our two proper YAML config files are combined, ordinarily the recipe-local variant would overwrite the user-wide variant, yielding {'some_trait': 'pony'}. However, with the extend_keys entry, we end up with what we've always wanted: a dog and pony show: {'some_trait': ['dog', 'pony'])}

Again, this is mostly an internal implementation detail - unless you find a use for it. Internally, it is used to aggregate the pin_run_as_build and ignore_version entries from any of your conda_build_config.yaml files.

Pinning expressions are the syntax used to specify how many parts of the version to pin. They are by convention strings containing x characters separated by .. The number of version parts to pin is simply the number of things that are separated by .. For example, "x.x" pins major and minor version. "x" pins only major version.

Wherever pinning expressions are accepted, you can customize both lower and upper bounds.

# produces pins like >=1.11.2,<1.12
variants = [{"numpy": "1.11", "pin_run_as_build": {"numpy": {"max_pin": "x.x"}}}]

Note that the final pin may be more specific than your initial spec. Here, the spec is 1.11, but the produced pin could be 1.11.2, the exact version of NumPy that was used at build time.

# produces pins like >=1.11,<2
variants = [
    {"numpy": "1.11", "pin_run_as_build": {"numpy": {"min_pin": "x.x", "max_pin": "x"}}}
]

Note that for pre-release versions min_pin will be ignored and substituted with the exact input version since pre-releases can never match >=x.x (see Package match specifications for details on pre-release version matching).

Some packages, such as boost, always need to be pinned at runtime to the version that was present at build time. For these cases where the need for pinning is consistent, pinning at the variant level is a good option. Conda-build will automatically pin run requirements to the versions present in the build environment when the following conditions are met:

1. The dependency is listed in the requirements/build section. It can be pinned, but does not need to be.
2. The dependency is listed by name (no pinning) in the requirements/run section.
3. The pin_run_as_build key in the variant has a value that is a dictionary, containing a key that matches the dependency name listed in the run requirements. The value should be a dictionary with up to 4 keys: min_pin, max_pin, lower_bound, upper_bound. The first 2 are pinning expressions. The latter 2 are version numbers, overriding detection of current version.

An example variant/recipe is shown here:

conda_build_config.yaml:

boost: 1.63
pin_run_as_build:
    boost:
      max_pin: x.x

meta.yaml:

requirements:
    build:
        - boost
    run:
        - boost

The result here is that the runtime boost dependency will be pinned to >=(current boost 1.63.x version),<1.64.

More details on the pin_run_as_build function is below in the Extra Jinja2 functions section.

Note that there are some packages that you should not use pin_run_as_build for. Packages that don't always need to be pinned should be pinned on a per-recipe basis (described in the next section). NumPy is an interesting example here. It actually would not make a good case for pinning at the variant level. Because you only need this kind of pinning for recipes that use NumPy's C API, it would actually be better not to pin NumPy with pin_run_as_build. Pinning it is over-constraining your requirements unnecessarily when you are not using NumPy's C API. Instead, we should customize it for each recipe that uses NumPy. See also the Avoiding unnecessary builds section above.

Pinning at the recipe level overrides pinning at the variant level, because run dependencies that have pinning values in meta.yaml (even as Jinja variables) are ignored by the logic handling pin_run_as_build. We expect that pinning at the recipe level will be used when some recipe's pinning is unusually stringent (or loose) relative to some standard pinning from the variant level.

By default, with the pin_compatible('package_name') function, conda-build pins to your current version and less than the next major version. For projects that don't follow the philosophy of semantic versioning, you might want to restrict things more tightly. To do so, you can pass one of two arguments to the pin_compatible function.

variants = [{"numpy": "1.11"}]

meta.yaml:

requirements:
    build:
        - numpy
    run:
        - {{ pin_compatible('numpy', max_pin='x.x') }}

This would yield a pinning of >=1.11.2,<1.12.

The syntax for the min_pin and max_pin is a string pinning expression. Each can be passed independently of the other. An example of specifying both:

variants = [{"numpy": "1.11"}]

meta.yaml:

requirements:
    build:
        - numpy
    run:
        - {{ pin_compatible('numpy', min_pin='x.x', max_pin='x.x') }}

This would yield a pinning of >=1.11,<1.12.

You can also pass the minimum or maximum version directly. These arguments supersede the min_pin and max_pin arguments and are thus mutually exclusive.

variants = [{"numpy": "1.11"}]

meta.yaml:

requirements:
    build:
        - numpy
    run:
        - {{ pin_compatible('numpy', lower_bound='1.10', upper_bound='3.0') }}

This would yield a pinning of >=1.10,<3.0.

As of conda-build 3.0, you can add a file named recipe_append.yaml in the same folder as your meta.yaml file. This file is considered to follow the same rules as meta.yaml, except that selectors and Jinja2 templates are not evaluated. Evaluation of selectors and Jinja2 templates will likely be added in future development.

Any contents in recipe_append.yaml will add to the contents of meta.yaml. List values will be extended and string values will be concatenated. The proposed use case for this is to tweak/extend central recipes, such as those from conda-forge, with additional requirements while minimizing the actual changes to recipe files so as to avoid merge conflicts and source code divergence.

As of conda-build 3.0, you can add a file named recipe_clobber.yaml in the same folder as your meta.yaml file. This file is considered to follow the same rules as meta.yaml, except that selectors and Jinja2 templates are not evaluated. Evaluation of selectors and Jinja2 templates will likely be added in future development.

Any contents in recipe_clobber.yaml will replace the contents of meta.yaml. This can be useful, for example, for replacing the source URL without copying the rest of the recipe into a fork.

With only a few things supported, we could just add things to the filename, such as py27 for Python, or np111 for NumPy. Variants are meant to support the general case, and in the general case this is no longer an option. Instead, used variant keys and values are hashed using the SHA1 algorithm, and that hash is a unique identifier. The information that went into the hash is stored with the package in a file at info/hash_input.json. Packages only have a hash when there are any "used" variables beyond the ones that are already accounted for in the build string (py, np, etc). The takeaway message is that hashes will appear when binary compatibility matters, but not when it doesn't.

Currently, only the first 7 characters of the hash are stored. Output package names will keep the pyXY and npXYY, but may have added the 7-character hash. Your package names will look like:

my-package-1.0-py27h3142afe_0.tar.bz2

As of conda-build 3.1.0, this hashing scheme has been simplified. A hash will be added if all of these are true for any dependency:

• Package is an explicit dependency in build, host, or run deps.
• Package has a matching entry in conda_build_config.yaml which is a pin to a specific version, not a lower bound.
• That package is not ignored by ignore_version.

OR

• Package uses {{ compiler() }} Jinja2 function.

Since conflicts only need to be prevented within one version of a package, we think this will be adequate. If you run into hash collisions with this limited subspace, please file an issue on the conda-build issue tracker.

There is a CLI tool that just pretty-prints this JSON file for easy viewing:

conda inspect hash-inputs <package path>

This produces output such as:

{'python-3.6.4-h6538335_1': {'files': [],
                            'recipe': {'c_compiler': 'vs2015',
                                        'cxx_compiler': 'vs2015'}}}

Two especially common operations when dealing with these API and ABI incompatibilities are ways of specifying such compatibility, and of explicitly expressing the compiler to be used. Three new Jinja2 functions are available when evaluating meta.yaml templates:

• pin_compatible('package_name', min_pin='x.x.x.x.x.x', max_pin='x', lower_bound=None, upper_bound=None): To be used as pin in run and/or test requirements. Takes package name argument. Looks up compatibility of named package installed in the build environment and writes compatible range pin for run and/or test requirements. Defaults to a semver-based assumption: package_name >=(current version),<(next major version). Pass min_pin or max_pin a Pinning expressions . This will be enhanced as time goes on with information from ABI Laboratory.
• pin_subpackage('package_name', min_pin='x.x.x.x.x.x', max_pin='x', exact=False): To be used as pin in run and/or test requirements. Takes package name argument. Used to refer to particular versions of subpackages built by parent recipe as dependencies elsewhere in that recipe. Can use either pinning expressions, or exact (including build string).
• compiler('language'): To be used in build requirements most commonly. Run or test as necessary. Takes language name argument. This is shorthand to facilitate cross-compiler usage. This Jinja2 function ties together 2 variant variables, {language}_compiler and target_platform, and outputs a single compiler package name. For example, this could be used to compile outputs targeting x86_64 and arm in one recipe, with a variant.

There are default "native" compilers that are used when no compiler is specified in any variant. These are defined in conda-build's jinja_context.py file. Most of the time, users will not need to provide compilers in their variants - just leave them empty and conda-build will use the defaults appropriate for your system.

Conda-build 2.1 brought in the ability to build multiple output packages from a single recipe. This is useful in cases where you have a big build that outputs a lot of things at once, but those things really belong in their own packages. For example, building GCC outputs not only GCC, but also GFortran, g++, and runtime libraries for GCC, GFortran, and g++. Each of those should be their own package to make things as clean as possible. Unfortunately, if there are separate recipes to repack the different pieces from a larger, whole package it can be hard to keep them in sync. That's where variants come in. Variants, and more specifically the pin_subpackage(name) function, give you a way to refer to the subpackage with control over how tightly the subpackage version relationship should be in relation to other subpackages or the parent package. The following will output 5 conda packages.

meta.yaml:

package:
  name: subpackage_demo
  version: 1.0

requirements:
  run:
    - {{ pin_subpackage('subpackage_1') }}
    - {{ pin_subpackage('subpackage_2', max_pin='x.x') }}
    - {{ pin_subpackage('subpackage_3', min_pin='x.x', max_pin='x.x') }}
    - {{ pin_subpackage('subpackage_4', exact=True) }}


outputs:
  - name: subpackage_1
    version: 1.0.0
  - name: subpackage_2
    version: 2.0.0
  - name: subpackage_3
    version: 3.0.0
  - name: subpackage_4
    version: 4.0.0

Here, the parent package will have the following different runtime dependencies:

• subpackage_1 >=1.0.0,<2 (default uses min_pin='x.x.x.x.x.x, max_pin='x', pins to major version with default >= current version lower bound)
• subpackage_2 >=2.0.0,<2.1 (more stringent upper bound)
• subpackage_3 >=3.0,<3.1 (less stringent lower bound, more stringent upper bound)
• subpackage_4 4.0.0 h81241af (exact pinning - version plus build string)

On macOS and Linux, we can and do ship GCC packages. These will become even more powerful with variants since you can specify versions of your compiler much more explicitly and build against different versions, or with different flags set in the compiler package's activate.d scripts. On Windows, rather than providing the actual compilers in packages, we still use the compilers that are installed on the system. The analogous compiler packages on Windows run any compiler activation scripts and set compiler flags instead of actually installing anything.

Over time, conda-build will require that all packages explicitly list their compiler requirements this way. This is to both simplify conda-build and improve the tracking of metadata associated with compilers - localize it to compiler packages, even if those packages are doing nothing more than activating an already-installed compiler, such as Visual Studio.

Note also the run_exports key in meta.yaml. This is useful for compiler recipes to impose runtime constraints based on the versions of subpackages created by the compiler recipe. For more information, see the Export runtime requirements section of the meta.yaml docs. Compiler packages provided by Anaconda use the run_exports key extensively. For example, recipes that include the gcc_linux-cos5-x86_64 package as a build time dependency (either directly, or through a {{ compilers('c') }} Jinja2 function) will automatically have a compatible libgcc runtime dependency added.

Usually the newest compilers are the best compilers, but in some special cases you'll need to use older compilers.

For example, NVIDIA's CUDA libraries only support compilers that they have rigorously tested. Often the latest GCC compiler is not supported for use with CUDA. If your recipe needs to use CUDA, you'll need to use an older version of GCC.

There are special keys associated with the compilers. The key name of each special key is the compiler key name plus _version.

For example, if your compiler key is c_compiler, the version key associated with it is c_compiler_version. If you have a recipe for Tensorflow with GPU support, put a conda_build_config.yaml file alongside meta.yaml, with contents like:

c_compiler_version:    # [linux]
    - 5.4              # [linux]
cxx_compiler_version:  # [linux]
    - 5.4              # [linux]

Specify selectors so that this extra version information is not also applied to Windows and macOS. Those platforms have totally different compilers and could have their own versions if necessary.

It is not necessary to specify c_compiler or cxx_compiler because the default value (gcc on Linux) will be used. It is necessary to specify both c and cxx versions, even if they are the same, because they are treated independently.

By placing this file in the recipe, it will apply only to this recipe. All other recipes will default to the latest compiler.

NOTE:

The compiler Jinja2 function is written to support cross-compilers. This depends on setting at least 2 variant keys: (language)_compiler and target_platform. The target platform is appended to the value of (language)_compiler with the _ character. This leads to package names like g++_linux-aarch64. We recommend a convention for naming your compiler packages as: <compiler name>_<target_platform>.

Using a cross-compiler in a recipe would look like the following:

variants = {
    "cxx_compiler": ["g++"],
    "target_platform": ["linux-cos5-x86_64", "linux-aarch64"],
}

and a meta.yaml file:

package:
    name: compiled-code
    version: 1.0

requirements:
    build:
        - {{ compiler('cxx') }}

This assumes that you have created 2 compiler packages named g++_linux-cos5-x86_64 and g++_linux-aarch64 - all conda-build is providing you with is a way to loop over appropriately named cross-compiler toolchains.

The compiler function is also how you could support a non-standard Visual Studio version, such as using VS 2015 to compile Python 2.7 and packages for Python 2.7. To accomplish this, you need to add the {{ compiler('<language>') }} to each recipe that will make up the system.  Environment consistency is maintained through dependencies - thus it is useful to have the runtime be a versioned package with only one version being able to be installed at a time. For example, the vc package, originally created by Conda-Forge, is a versioned package (only one version can be installed at a time), and it installs the correct runtime package. When the compiler package imposes such a runtime dependency, then the resultant ecosystem is self-consistent.

Given these guidelines, consider a system of recipes using a variant like this:

variants = {"cxx_compiler": ["vs2015"]}

The recipes include a compiler meta.yaml like this:

package:
    name: vs2015
    version: 14.0
build:
    run_exports:
        - vc 14

They also include some compiler-using meta.yaml contents like this:

package:
    name: compiled-code
    version: 1.0

requirements:
    build:
        # these are the same (and thus redundant) on windows, but different elsewhere
        - {{ compiler('c') }}
        - {{ compiler('cxx') }}

These recipes will create a system of packages that are all built with the VS 2015 compiler, and which have the vc package matched at version 14, rather than whatever default is associated with the Python version.

Command-line interface (CLI) adds onto conda-build's functionality. CLI provides functions enabling you to convert packages between formats, render recipes, develop an index of packages, and more.

When a package is installed, it can add a shortcut to the Windows Start menu. Conda and conda-build handle this with the package menuinst, which currently supports only Windows. For instructions on using menuinst, see the menuinst wiki.

The easiest way to ensure that a package made with conda constructor does not install any menu shortcuts is to remove menuinst from the list of conda packages included. To do this, add the following to the constuct.yaml file:

exclude:
  - menuinst

Follow these guidelines for submitting or editing conda-build documentation.

Identify who your audience is, their skill level, and how they can use the information.

Match the level of technical language with the audience's level of proficiency. It's better to uderestimate the knowledge of your readers than overestimate it. Limit technical terms to those the user will encounter. If you must define a large number of terms, use a glossary to supplement definitions in the text.

Use the active voice (e.g. Click this) and address users directly (write "you" rather than "the user"). When explaining an action, use the "command" form of the verb: "Choose an option from the menu and press Enter."

See the tutorial template for the format. Provide descriptive task and subtask titles and do not number headings.

This document describes the steps for creating a tutorial for conda-build.

• Copy the template: https://github.com/conda/conda-build/tree/main/docs/source/resources/tutorial-template.rst.
• Replace the italicized text with your tutorial content, following the writing style guide.
• Review other tutorials for additional guidance.
• Contact us at documentation@anaconda.com for help.

Provide descriptions of the tutorial's:

• Purpose.
• Benefits.
• Application.

• Who is your audience?
• What skills or prior knowledge do they need?
• How will they use this tutorial?

Before you start, check the Prerequisites.

Provide descriptions of and links to requisite programs.

If you're using several technical terms that your readers may be unfamiliar with, provide a glossary of key terms.

• Provide descriptive titles for each section.
• Identify the major tasks.
• Separate each major task into subtasks.
• Write a series of steps that walk the user through each subtask.

• Provide links to related content.
• Add final notes for how to expand upon the tutorial.

• Reduce render time when there is a large number of unused variants. (#5392)

• Ensure variables mentioned in script_env are undefined in the multi-output build environment if they are undefined in the environment that conda-build is invoked from. (#5322)
• Variables used in single-line jinja2 for and set statements are now properly included in the variant matrix for some edge cases. (#5447)
• Allow undefined jinja variables when a particular metadata block evaluates as skipped. (#5458)

• Remove conda_build.build.check_external. patchelf is an explicit conda-build dependency on Linux, so it will always be installed. (#5441)
• Remove conda_build.metadata._get_env_path. Use conda.base.context.locate_prefix_by_name instead. (#5441)
• Remove conda_build.build._construct_metadata_for_test_from_recipe. Test built packages instead, not recipes (e.g., conda build --test package instead of conda build --test recipe/). (#5478)

• @beeankha
• @conda-bot
• @jameslamb
• @kenodegard
• @beckermr
• @msarahan
• @minrk
• @dependabot[bot]
• @pre-commit-ci[bot]

• Check for WSL existence before calling os.stat. (#5433 via #5434)

• @kenodegard

• Skip generating repodata.json.bz2 for local index; generate repodata.json only; require conda-package-handling >=2.2.0 matching conda. (#5231)
• Add new include/exclude sections for glob expressions in multi-output outputs/files. (#4196 via #5216)
• Increase performance by using pickle instead of copy.deepcopy. (#5281)
• Report fully rendered recipe to stdout before the build process starts. (#3798 via #5344)
• Validate run_constrained dependencies to prevent faulty specs reaching final repodata. (#5047 via #5359)
• PIP_* env variables are set when building outputs in multi-output recipes. (#3993 via #5368)
• Reduce performance overhead of logging. (#5384)

• Include file path in addition to the content when generating the file hash to avoid unwanted caching during linkage analysis. (#4821)
• Error handling when LIEF fails is now consistent with patchelf. (#5176)
• Ensure cross-building recipes select the correct noarch package variants. (#5341 via #5350)
• On Linux platforms, prefer the sysroot matching the target_platform when cross-compiling (#5403).

• Mark conda_build.build._construct_metadata_for_test_from_recipe as deprecated. Test built packages instead, not recipes (e.g., conda build --test package instead of conda build --test recipe/). (#3192 via #5352)
• Mark conda_build.build.check_external for deprecation. patchelf is an explicit conda-build dependency on Linux so it will always be installed. (#5355)

• Remove the following deprecations:

  • conda_build.config.Config.override_channels (use conda.base.context.context.channels instead). (#5333)
  • conda_build.config.noarch_python_build_age_default. (#5333)
  • conda_build.conda_interface.add_parser_channels (use conda.cli.helpers.add_parser_channels instead). (#5333)
  • conda_build.conda_interface.add_parser_prefix (use conda.cli.helpers.add_parser_prefix instead). (#5333)
  • conda_build.conda_interface.ArgumentParser (use conda.cli.conda_argparse.ArgumentParser instead). (#5333)
  • conda_build.conda_interface.binstar_upload (use conda.base.context.context.binstar_upload instead). (#5333)
  • conda_build.conda_interface.cc_conda_build (use conda.base.context.context.conda_build instead). (#5333)
  • conda_build.conda_interface.cc_platform (use conda.base.context.context.platform instead). (#5333)
  • conda_build.conda_interface.Channel (use conda.models.channel.Channel instead). (#5333)
  • conda_build.conda_interface.Completer. (#5333)
  • conda_build.conda_interface.configparser (use configparser instead). (#5333)
  • conda_build.conda_interface.CondaError (use conda.exceptions.CondaError instead). (#5333)
  • conda_build.conda_interface.CondaHTTPError (use conda.exceptions.CondaHTTPError instead). (#5333)
  • conda_build.conda_interface.CondaSession (use conda.gateways.connection.session.CondaSession instead). (#5333)
  • conda_build.conda_interface.CONDA_VERSION (use conda.__version__ instead). (#5333)
  • conda_build.conda_interface.context (use conda.base.context.context instead). (#5333)
  • conda_build.conda_interface.create_default_packages (use conda.base.context.context.create_default_packages instead). (#5333)
  • conda_build.conda_interface.default_python (use conda.base.context.context.default_python instead). (#5333)
  • conda_build.conda_interface.determine_target_prefix (use conda.base.context.determine_target_prefix instead). (#5333)
  • conda_build.conda_interface.download (use conda.gateways.connection.download.download instead). (#5333)
  • conda_build.conda_interface.env_path_backup_var_exists. (#5333)
  • conda_build.conda_interface.envs_dirs (use conda.base.context.context.envs_dirs instead). (#5333)
  • conda_build.conda_interface.EntityEncoder (use conda.auxlib.entity.EntityEncoder instead). (#5333)
  • conda_build.conda_interface.FileMode (use conda.models.enums.FileMode instead). (#5333)
  • conda_build.conda_interface.get_conda_build_local_url (use conda.models.channel.get_conda_build_local_url instead). (#5333)
  • conda_build.conda_interface.get_conda_channel (use conda.models.channel.Channel.from_value instead). (#5333)
  • conda_build.conda_interface.get_prefix (use conda.base.context.context.target_prefix instead). (#5333)
  • conda_build.conda_interface.get_rc_urls (use conda.base.context.context.channels instead). (#5333)
  • conda_build.conda_interface.human_bytes (use conda.utils.human_bytes instead). (#5333)
  • conda_build.conda_interface.import_module (use importlib.import_module instead). (#5333)
  • conda_build.conda_interface.input (use input instead). (#5333)
  • conda_build.conda_interface.InstalledPackages. (#5333)
  • conda_build.conda_interface.lchmod (use conda.gateways.disk.link.lchmod instead). (#5333)
  • conda_build.conda_interface.LinkError (use conda.exceptions.LinkError instead). (#5333)
  • conda_build.conda_interface.LockError (use conda.exceptions.LockError instead). (#5333)
  • conda_build.conda_interface.MatchSpec (use conda.models.match_spec.MatchSpec instead). (#5333)
  • conda_build.conda_interface.non_x86_linux_machines (use conda.base.context.non_x86_machines instead). (#5333)
  • conda_build.conda_interface.NoPackagesFound (use conda.exceptions.ResolvePackageNotFound instead). (#5333)
  • conda_build.conda_interface.NoPackagesFoundError (use conda.exceptions.NoPackagesFoundError instead). (#5333)
  • conda_build.conda_interface.normalized_version (use conda.models.version.normalized_version instead). (#5333)
  • conda_build.conda_interface.os (use os instead). (#5333)
  • conda_build.conda_interface.PackageRecord (use conda.models.records.PackageRecord instead). (#5333)
  • conda_build.conda_interface.PaddingError (use conda.exceptions.PaddingError instead). (#5333)
  • conda_build.conda_interface.partial (use functools.partial instead). (#5333)
  • conda_build.conda_interface.PathType (use conda.models.enums.PathType instead). (#5333)
  • conda_build.conda_interface.pkgs_dirs (use conda.base.context.context.pkgs_dirs instead). (#5333)
  • conda_build.conda_interface.prefix_placeholder (use conda.base.constants.PREFIX_PLACEHOLDER instead). (#5333)
  • conda_build.conda_interface.ProgressiveFetchExtract (use conda.core.package_cache_data.ProgressiveFetchExtract instead). (#5333)
  • conda_build.conda_interface.reset_context (use conda.base.context.reset_context instead). (#5333)
  • conda_build.conda_interface.Resolve (use conda.resolve.Resolve instead). (#5333)
  • conda_build.conda_interface.rm_rf (use conda_build.utils.rm_rf instead). (#5333)
  • conda_build.conda_interface.root_dir (use conda.base.context.context.root_prefix instead). (#5333)
  • conda_build.conda_interface.root_writable (use conda.base.context.context.root_writable instead). (#5333)
  • conda_build.conda_interface.spec_from_line (use conda.cli.common.spec_from_line instead). (#5333)
  • conda_build.conda_interface.specs_from_args (use conda.cli.common.specs_from_args instead). (#5333)
  • conda_build.conda_interface.specs_from_url (use conda.cli.common.specs_from_url instead). (#5333)
  • conda_build.conda_interface.StringIO (use io.StringIO instead). (#5333)
  • conda_build.conda_interface.subdir (use conda.base.context.context.subdir instead). (#5333)
  • conda_build.conda_interface.symlink_conda. (#5333)
  • conda_build.conda_interface.TemporaryDirectory (use conda.gateways.disk.create.TemporaryDirectory instead). (#5333)
  • conda_build.conda_interface.TmpDownload (use conda.gateways.connection.download.TmpDownload instead). (#5333)
  • conda_build.conda_interface._toposort (use conda.common.toposort._toposort instead). (#5333)
  • conda_build.conda_interface.unix_path_to_win (use conda.utils.unix_path_to_win instead). (#5333)
  • conda_build.conda_interface.untracked (use conda.misc.untracked instead). (#5333)
  • conda_build.conda_interface.Unsatisfiable (use conda.exceptions.UnsatisfiableError instead). (#5333)
  • conda_build.conda_interface.UnsatisfiableError (use conda.exceptions.UnsatisfiableError instead). (#5333)
  • conda_build.conda_interface.url_path (use conda.utils.url_path instead). (#5333)
  • conda_build.conda_interface.VersionOrder (use conda.models.version.VersionOrder instead). (#5333)
  • conda_build.conda_interface.walk_prefix (use conda.misc.walk_prefix instead). (#5333)
  • conda_build.conda_interface.win_path_to_unix (use conda.common.path.win_path_to_unix instead). (#5333)
  • conda_build.index.channel_data; conda_build.index.get_build_index return value for channel_data is now always None. (#5333)
  • conda_build.metadata.check_circular_dependencies (use conda_build.metadata._check_circular_dependencies instead). (#5406)
  • conda_build.metadata.toposort (use conda_build.metadata.toposort_outputs instead). (#5406)
  • conda_build.utils._convert_lists_to_sets (use frozendict.deepfreeze instead). (#5333)
  • conda_build.utils.HashableDict (use frozendict.deepfreeze instead). (#5333)
  • conda_build.utils.represent_hashabledict (use frozendict.deepfreeze instead). (#5333)
  • conda_build.utils.rm_rf(config). (#5333)
  • conda_build.variants.get_vars(loop_only). (#5333)

• Do not generate conda error reports for common exceptions. (#5264)
• Add CondaBuildUserError exception to replace sys.exit calls. (#5353)
• Updated the CI to download the MacOSX 10.15 SDK. (#5387)

• @beeankha
• @conda-bot
• @carterbox made their first contribution in https://github.com/conda/conda-build/pull/5216
• @dholth
• @isuruf
• @jaimergp
• @kenodegard
• @zklaus made their first contribution in https://github.com/conda/conda-build/pull/5364
• @beckermr
• @minrk
• @timkpaine made their first contribution in https://github.com/conda/conda-build/pull/4821
• @dependabot[bot]
• @pre-commit-ci[bot]

• Fix issue with modifying a frozendict when specifying outputs/files in meta.yaml. (#5342 via #5345)
• Fix excessive memory use in inspect_linkages_lief. (#5267 via #5348)

• Mark conda_build.metadata.toposort as deprecated. Use conda_build.metadata.toposort_outputs instead. (#5342 via #5345)
• Mark conda_build.metadata.check_circular_dependencies as deprecated. Use conda_build.metadata._check_circular_dependencies instead. (#5342 via #5345)

• @beeankha
• @kenodegard
• @mbargull

• Only fetch lfs files for specific git_ref. (#5202)
• Add conda_build.metadata._split_line_selector to cache line-selector parsed text. (#5237)
• Add conda_build.render.open_recipe context manager to detect the recipe type (file/meta.yaml, directory/recipe, or tarball/package) and properly handling any exit/close behavior. (#5238)
• For Windows users, the stub executables used for Python entrypoints in packages are now codesigned. (#5252)
• Require conda >=23.7.0. (#5271)

• Fix all CLI arguments to properly initialize conda.base.context.context with parsed arguments. Fixes issue with arguments not being processed (e.g., --override-channels was previously ignored). (#3693 via #5271)

• Deprecate conda_build.conda_interface.CONDA_VERSION constant. Use conda.__version__ instead. (#5222)
• Deprecate conda_build.conda_interface.binstar_upload constant. Use conda.base.context.context.binstar_upload instead. (#5222)
• Deprecate conda_build.conda_interface.default_python constant. Use conda.base.context.context.default_python instead. (#5222)
• Deprecate conda_build.conda_interface.envs_dirs constant. Use conda.base.context.context.envs_dirs instead. (#5222)
• Deprecate conda_build.conda_interface.pkgs_dirs constant. Use conda.base.context.context.pkgs_dirs instead. (#5222)
• Deprecate conda_build.conda_interface.cc_platform constant. Use conda.base.context.context.platform instead. (#5222)
• Deprecate conda_build.conda_interface.root_dir constant. Use conda.base.context.context.root_prefix instead. (#5222)
• Deprecate conda_build.conda_interface.root_writable constant. Use conda.base.context.context.root_writable instead. (#5222)
• Deprecate conda_build.conda_interface.subdir constant. Use conda.base.context.context.subdir instead. (#5222)
• Deprecate conda_build.conda_interface.create_default_packages constant. Use conda.base.context.context.create_default_packages instead. (#5222)
• Deprecate conda_build.conda_interface.get_rc_urls function. Use conda.base.context.context.channels instead. (#5222)
• Deprecate conda_build.conda_interface.get_prefix function. Use conda.base.context.context.target_prefix instead. (#5222)
• Deprecate conda_build.conda_interface.get_conda_channel function. Use conda.models.channel.Channel.from_value instead. (#5222)
• Deprecate conda_build.conda_interface.reset_context function. Use conda.base.context.reset_context instead. (#5222)
• Deprecate conda_build.conda_interface.context singleton. Use conda.base.context.context instead. (#5251)
• Deprecate conda_build.conda_interface.configparser module. Use configparser instead. (#5251)
• Deprecate conda_build.conda_interface.os module. Use os instead. (#5251)
• Deprecate conda_build.conda_interface.partial function. Use functools.partial instead. (#5251)
• Deprecate conda_build.conda_interface.import_module function. Use importlib.import_module instead. (#5251)
• Deprecate conda_build.conda_interface.determine_target_prefix function. Use conda.base.context.determine_target_prefix instead. (#5251)
• Deprecate conda_build.conda_interface.non_x86_linux_machines constant. Use conda.base.context.non_x86_machines instead. (#5251)
• Deprecate conda_build.conda_interface.ProgressiveFetchExtract class. Use conda.core.package_cache.ProgressiveFetchExtract instead. (#5251)
• Deprecate conda_build.conda_interface.CondaError class. Use conda.exceptions.CondaError instead. (#5251)
• Deprecate conda_build.conda_interface.CondaHTTPError class. Use conda.exceptions.CondaHTTPError instead. (#5251)
• Deprecate conda_build.conda_interface.LinkError class. Use conda.exceptions.LinkError instead. (#5251)
• Deprecate conda_build.conda_interface.LockError class. Use conda.exceptions.LockError instead. (#5251)
• Deprecate conda_build.conda_interface.NoPackagesFoundError class. Use conda.exceptions.NoPackagesFoundError instead. (#5251)
• Deprecate conda_build.conda_interface.PaddingError class. Use conda.exceptions.PaddingError instead. (#5251)
• Deprecate conda_build.conda_interface.UnsatisfiableError class. Use conda.exceptions.UnsatisfiableError instead. (#5251)
• Deprecate conda_build.conda_interface.get_conda_build_local_url class. Use conda.models.channel.get_conda_build_local_url instead. (#5251)
• Deprecate conda_build.config.Config.override_channels. Defer to conda.base.context.context.channels instead. (#5271, #5324)
• Deprecate conda_build.conda_interface._toposort. Use conda.common.toposort._toposort instead. (#5276)
• Deprecate conda_build.conda_interface.add_parser_channels. Use conda.cli.helpers.add_parser_channels instead. (#5276)
• Deprecate conda_build.conda_interface.add_parser_prefix. Use conda.cli.helpers.add_parser_prefix instead. (#5276)
• Deprecate conda_build.conda_interface.ArgumentParser. Use conda.cli.conda_argparse.ArgumentParser instead. (#5276)
• Deprecate conda_build.conda_interface.cc_conda_build. Use conda.base.context.context.conda_build instead. (#5276)
• Deprecate conda_build.conda_interface.Channel. Use conda.models.channel.Channel instead. (#5276)
• Deprecate conda_build.conda_interface.Completer. Unused. (#5276)
• Deprecate conda_build.conda_interface.CondaSession. Use conda.gateways.connection.session.CondaSession instead. (#5276)
• Deprecate conda_build.conda_interface.download. Use conda.gateways.connection.download.download instead. (#5276)
• Deprecate conda_build.conda_interface.EntityEncoder. Use conda.auxlib.entity.EntityEncoder instead. (#5276)
• Deprecate conda_build.conda_interface.env_path_backup_var_exists. Unused. (#5276)
• Deprecate conda_build.conda_interface.FileMode. Use conda.models.enums.FileMode instead. (#5276)
• Deprecate conda_build.conda_interface.human_bytes. Use conda.utils.human_bytes instead. (#5276)
• Deprecate conda_build.conda_interface.input. Use input instead. (#5276)
• Deprecate conda_build.conda_interface.InstalledPackages. Unused. (#5276)
• Deprecate conda_build.conda_interface.lchmod. Use conda.gateways.disk.link.lchmod instead. (#5276)
• Deprecate conda_build.conda_interface.MatchSpec. Use conda.models.match_spec.MatchSpec instead. (#5276)
• Deprecate conda_build.conda_interface.NoPackagesFound. Use conda.exceptions.ResolvePackageNotFound instead. (#5276)
• Deprecate conda_build.conda_interface.normalized_version. Use conda.models.version.normalized_version instead. (#5276)
• Deprecate conda_build.conda_interface.PackageRecord. Use conda.models.records.PackageRecord instead. (#5276)
• Deprecate conda_build.conda_interface.PathType. Use conda.models.enums.PathType instead. (#5276)
• Deprecate conda_build.conda_interface.prefix_placeholder. Use conda.base.constants.PREFIX_PLACEHOLDER instead. (#5276)
• Deprecate conda_build.conda_interface.Resolve. Use conda.resolve.Resolve instead. (#5276)
• Deprecate conda_build.conda_interface.rm_rf. Use conda_build.utils.rm_rf instead. (#5276)
• Deprecate conda_build.conda_interface.spec_from_line. Use conda.cli.common.spec_from_line instead. (#5276)
• Deprecate conda_build.conda_interface.specs_from_args. Use conda.cli.common.specs_from_args instead. (#5276)
• Deprecate conda_build.conda_interface.specs_from_url. Use conda.cli.common.specs_from_url instead. (#5276)
• Deprecate conda_build.conda_interface.StringIO. Use io.StringIO instead. (#5276)
• Deprecate conda_build.conda_interface.symlink_conda. Unused. (#5276)
• Deprecate conda_build.conda_interface.TempDirectory. Use conda.gateways.disk.create.TemporaryDirectory instead. (#5276)
• Deprecate conda_build.conda_interface.TmpDownload. Use conda.gateways.connection.download.TmpDownload instead. (#5276)
• Deprecate conda_build.conda_interface.unix_path_to_win. Use conda.utils.unix_path_to_win instead. (#5276)
• Deprecate conda_build.conda_interface.Unsatisfiable. Use conda.exceptions.UnsatisfiableError instead. (#5276)
• Deprecate conda_build.conda_interface.untracked. Use conda.misc.untracked instead. (#5276)
• Deprecate conda_build.conda_interface.url_path. Use conda.utils.url_path instead. (#5276)
• Deprecate conda_build.conda_interface.VersionOrder. Use conda.models.version.VersionOrder instead. (#5276)
• Deprecate conda_build.conda_interface.walk_prefix. Use conda.misc.walk_prefix instead. (#5276)
• Deprecate conda_build.conda_interface.win_path_to_unix. Use conda.common.path.win_path_to_unix instead. (#5276)
• Deprecate conda_build.variants.get_vars(loop_only). (#5280)
• Deprecate conda_build.utils.HashableDict. Use frozendict.deepfreeze instead. (#5284)
• Deprecate conda_build.utils._convert_lists_to_sets. Use frozendict.deepfreeze instead. (#5284)
• Deprecate conda_build.utils.represent_hashabledict. Use frozendict.deepfreeze instead. (#5284)
• Deprecate conda_build.config.noarch_python_build_age_default. (#5298)
• Postpone conda_build.index.channel_data deprecation. (#5299)
• Remove conda_build.api.get_output_file_path. Use conda_build.api.get_output_file_paths instead. (#5299)
• Remove conda_build.bdist_conda. (#5299)
• Remove conda_build.build.have_prefix_files. (#5299)
• Remove conda_build.conda_interface.get_index. Use conda.core.index.get_index instead. (#5299)
• Remove conda_build.conda_interface.get_version_from_git_tag. Use conda_build.environ.get_version_from_git_tag instead. (#5299)
• Remove conda_build.conda_interface.handle_proxy_407. Handled by conda.gateways.connection.session.CondaSession. (#5299)
• Remove conda_build.conda_interface.hashsum_file. Use conda.gateways.disk.read.compute_sum instead. (#5299)
• Remove conda_build.conda_interface.md5_file. Use conda.gateways.disk.read.compute_sum(path, 'md5') instead. (#5299)
• Remove conda_build.environ._load_all_json. (#5299)
• Remove conda_build.environ._load_json. (#5299)
• Remove conda_build.environ.cached_actions. (#5299)
• Remove conda_build.environ.Environment. Use conda.core.prefix_data.PrefixData instead. (#5299)
• Remove conda_build.environ.InvalidEnvironment. (#5299)
• Remove conda_build.environ.LINK_ACTION. (#5299)
• Remove conda_build.environ.PREFIX_ACTION. (#5299)
• Remove conda_build.index._apply_instructions. Use conda_index._apply_instructions instead. (#5299)
• Remove conda_build.index.DummyExecutor. (#5299)
• Remove conda_build.index.LOCK_TIMEOUT_SECS. (#5299)
• Remove conda_build.index.LOCKFILE_NAME. (#5299)
• Remove conda_build.index.MAX_THREADS_DEFAULT. (#5299)

• Enable CodSpeed benchmarks for select tests. (#5233)

• @beeankha
• @conda-bot
• @jaimergp
• @Callek made their first contribution in https://github.com/conda/conda-build/pull/5252
• @kenodegard
• @mbargull
• @Tobias-Fischer made their first contribution in https://github.com/conda/conda-build/pull/5202
• @ytausch made their first contribution in https://github.com/conda/conda-build/pull/5214
• @dependabot[bot]
• @pre-commit-ci[bot]

• Add compatibility for LIEF=0.14. (#5227 via #5228)

• Fix stdlib being recognized in variant hash inputs. (#5190 via #5195)

• Mark conda_build.bdist_conda module as pending deprecation. (#5196)
• Mark conda_build.build.have_prefix_files as deprecated. (#5199)
• Mark conda_build.conda_interface.handle_proxy_407 as deprecated. Handled by conda.gateways.connection.session.CondaSession. (#5203)
• Mark conda_build.conda_interface.hashsum_file as deprecated. Use conda.gateways.disk.read.compute_sum instead. (#5203)
• Mark conda_build.conda_interface.md5_file as deprecated. Use conda.gateways.disk.read.compute_sum(path, 'md5') instead. (#5203)
• Mark conda_build.environ.PREFIX_ACTION as deprecated. (#5203)
• Mark conda_build.environ.LINK_ACTION as deprecated. (#5203)
• Mark conda_build.environ.cache_actions as deprecated. (#5203)
• Mark conda_build.index.DummyExecutor as deprecated. (#5203)
• Mark conda_build.index.MAX_THREADS_DEFAULT as deprecated. (#5203)
• Mark conda_build.index.LOCK_TIMEOUT_SECS as deprecated. (#5203)
• Mark conda_build.index.LOCKFILE_NAME as deprecated. (#5203)
• Postpone conda_build.index.channel_data deprecation. (#5203)
• Rename conda_build.environ.create_env('specs_or_actions' -> 'specs_or_precs'). (#5203)
• Rename `conda_build.environ._execute_actions(‘actions’ -> ‘precs’). (#5203)
• Rename `conda_build.environ._display_actions(‘actions’ -> ‘precs’). (#5203)
• Rename conda_build.inspect.check_install('platform' -> 'subdir'). (#5203)
• Rename conda_build.render.execute_download_actions('actions' -> 'precs'). (#5203)
• Rename conda_build.render.get_upstream_pins('actions' -> 'precs'). (#5203)
• Remove conda_build.cli.main_render.execute(print_results). (#5203)
• Remove conda_build.conda_interface.Dist. (#5203)
• Remove conda_build.conda_interface.display_actions. (#5203)
• Remove conda_build.conda_interface.execute_actions. (#5203)
• Remove conda_build.conda_interface.execute_plan. (#5203)
• Remove conda_build.conda_interface.install_actions. (#5203)
• Remove conda_build.conda_interface.linked. (#5203)
• Remove conda_build.conda_interface.linked_data. (#5203)
• Remove conda_build.conda_interface.package_cache. (#5203)
• Remove conda_build.environ.get_install_actions. Use conda_build.environ.get_package_records instead. (#5203)
• Remove conda_build.index._determine_namespace. (#5203)
• Remove conda_build.index._make_seconds. (#5203)
• Remove conda_build.index.REPODATA_VERSION. (#5203)
• Remove conda_build.index.CHANNELDATA_VERSION. (#5203)
• Remove conda_build.index.REPODATA_JSON_FN. (#5203)
• Remove conda_build.index.REPODATA_FROM_PKGS_JSON_FN. (#5203)
• Remove conda_build.index.CHANNELDATA_FIELDS. (#5203)
• Remove conda_build.index._clear_newline_chars. (#5203)
• Remove conda_build.index._get_jinja2_environment. (#5203)
• Remove conda_build.index._maybe_write. (#5203)
• Remove conda_build.index._make_build_string. (#5203)
• Remove conda_build.index._warn_on_missing_dependencies. (#5203)
• Remove conda_build.index._cache_post_install_details. (#5203)
• Remove conda_build.index._cache_recipe. (#5203)
• Remove conda_build.index._cache_run_exports. (#5203)
• Remove conda_build.index._cache_icon. (#5203)
• Remove conda_build.index._make_subdir_index_html. (#5203)
• Remove conda_build.index._make_channeldata_index_html. (#5203)
• Remove conda_build.index._get_source_repo_git_info. (#5203)
• Remove conda_build.index._cache_info_file. (#5203)
• Remove conda_build.index._alternate_file_extension. (#5203)
• Remove conda_build.index._get_resolve_object. (#5203)
• Remove conda_build.index._get_newest_versions. (#5203)
• Remove conda_build.index._add_missing_deps. (#5203)
• Remove conda_build.index._add_prev_ver_for_features. (#5203)
• Remove conda_build.index._shard_newest_packages. (#5203)
• Remove conda_build.index._build_current_repodata. (#5203)
• Remove conda_build.index.ChannelIndex. (#5203)
• Remove conda_build.inspect.check_install('prepend'). (#5203)
• Remove conda_build.inspect.check_install('minimal_hint'). (#5203)
• Remove conda_build.noarch_python.ISWIN. Use conda_build.utils.on_win instead. (#5203)
• Remove conda_build.noarch_python._force_dir. Use os.makedirs(exist_ok=True) instead. (#5203)
• Remove conda_build.noarch_python._error_exit. (#5203)
• Remove conda_build.render.actions_to_pins. (#5203)
• Remove conda_build.utils.linked_data_no_multichannels. (#5203)
• Mark conda_build.api.get_output_file_path as deprecated. Use conda_build.api.get_output_file_paths instead. (#5208)
• Mark conda_build.environ.Environment as deprecated. Use conda.core.prefix_data.PrefixData instead. (#5219)
• Mark conda_build.conda_interface.get_version_from_git_tag as deprecated. Use conda_build.environ.get_version_from_git_tag instead. (#5221)

• Update advice for installing conda-build into base environment. (#5223)

• Add a check to print an additional warning and return an empty string when bits is “arm64” in msvc_env_cmd. (#4867)

• @beeankha
• @conda-bot
• @dholth
• @finnagin made their first contribution in https://github.com/conda/conda-build/pull/4867
• @kathatherine
• @kenodegard
• @mbargull
• @minrk
• @ryanskeith
• @travishathaway
• @pre-commit-ci[bot]

• Fix rpaths patcher being run on symbolic links. (#5179 via #5181)
• Fix corrupted package cache for outputs in subpackage tests. (#5184)

• @mbargull

• Fix nonzero exitcode on success. (#5167 via #5169)

• @kenodegard

• Update conda inspect channels to use updated solver/transaction logic. (#5033)
• Relax script_env error in outputs when variable referenced in script_env is not defined. This unifies current behavior with the top-level build. (#5105)
• Add support for Python 3.12. (#4997 via #4998)
• Adopt calender versioning (CalVer) per CEP-8 for consistency with conda. (#4975)
• Adopt expedited CEP-9 deprecation policy. (#5064)

• Mark conda inspect channels --test-installable as pending deprecation. (#5033)
• Mark conda_build.inspect_pkg.check_install(package) as pending deprecation in favor of conda_build.inspect_pkg.check_install(subdir). (#5033)
• Mark conda_build.inspect_pkg.check_install(prepend) as pending deprecation. (#5033)
• Mark conda_build.inspect_pkg.check_install(minimal_hint) as pending deprecation. (#5033)
• Mark conda_build.conda_interface.Dist as pending deprecation. (#5074)
• Mark conda_build.conda_interface.display_actions as pending deprecation. (#5074)
• Mark conda_build.conda_interface.execute_actions as pending deprecation. (#5074)
• Mark conda_build.conda_interface.execute_plan as pending deprecation. (#5074)
• Mark conda_build.conda_interface.get_index as pending deprecation. (#5074)
• Mark conda_build.conda_interface.install_actions as pending deprecation. (#5074)
• Mark conda_build.conda_interface.linked as pending deprecation. (#5074)
• Mark conda_build.conda_interface.linked_data as pending deprecation. (#5074)
• Mark conda_build.utils.linked_data_no_multichannels as pending deprecation. (#5074)
• Mark conda_build.environ.get_install_actions as pending deprecation in favor of conda_build.environ.get_package_records. (#5152)
• Mark conda_build.environ.create_env(specs_or_actions) as pending deprecation in favor of conda_build.environ.create_env(specs_or_precs). (#5152)
• Mark conda_build.index.channel_data as pending deprecation. (#5152)
• Mark conda_build.index._determine_namespace as pending deprecation. (#5152)
• Mark conda_build.index._make_seconds as pending deprecation. (#5152)
• Mark conda_build.index.REPODATA_VERSION as pending deprecation. (#5152)
• Mark conda_build.index.CHANNELDATA_VERSION as pending deprecation. (#5152)
• Mark conda_build.index.REPODATA_JSON_FN as pending deprecation. (#5152)
• Mark conda_build.index.REPODATA_FROM_PKGS_JSON_FN as pending deprecation. (#5152)
• Mark conda_build.index.CHANNELDATA_FIELDS as pending deprecation. (#5152)
• Mark conda_build.index._clear_newline_chars as pending deprecation. (#5152)
• Mark conda_build.index._apply_instructions as pending deprecation. (#5152)
• Mark conda_build.index._get_jinja2_environment as pending deprecation. (#5152)
• Mark conda_build.index._maybe_write as pending deprecation. (#5152)
• Mark conda_build.index._maybe_build_string as pending deprecation. (#5152)
• Mark conda_build.index._warn_on_missing_dependencies as pending deprecation. (#5152)
• Mark conda_build.index._cache_post_install_details as pending deprecation. (#5152)
• Mark conda_build.index._cache_recipe as pending deprecation. (#5152)
• Mark conda_build.index._cache_run_exports as pending deprecation. (#5152)
• Mark conda_build.index._cache_icon as pending deprecation. (#5152)
• Mark conda_build.index._make_subdir_index_html as pending deprecation. (#5152)
• Mark conda_build.index._make_channeldata_index_html as pending deprecation. (#5152)
• Mark conda_build.index._get_source_repo_git_info as pending deprecation. (#5152)
• Mark conda_build.index._cache_info_file as pending deprecation. (#5152)
• Mark conda_build.index._alternate_file_extension as pending deprecation. (#5152)
• Mark conda_build.index._get_resolve_object as pending deprecation. (#5152)
• Mark conda_build.index._get_newest_versions as pending deprecation. (#5152)
• Mark conda_build.index._add_missing_deps as pending deprecation. (#5152)
• Mark conda_build.index._add_prev_ver_for_features as pending deprecation. (#5152)
• Mark conda_build.index._shard_newest_packages as pending deprecation. (#5152)
• Mark conda_build.index._build_current_repodata as pending deprecation. (#5152)
• Mark conda_build.index.ChannelIndex as pending deprecation. (#5152)
• Mark conda_build.render.actions_to_pins as pending deprecation. (#5152)
• Mark conda_build.render.execute_download_actions(actions) as pending deprecation in favor of conda_build.render.execute_download_actions(precs). (#5152)
• Mark conda_build.render.get_upstream_pins(actions) as pending deprecation in favor of conda_build.render.get_upstream_pins(precs). (#5152)
• Remove conda_build.api.update_index. (#5151)
• Remove conda_build.cli.main_build.main. (#5151)
• Remove conda_build.cli.main_convert.main. (#5151)
• Remove conda_build.cli.main_debug.main. (#5151)
• Remove conda_build.cli.main_develop.main. (#5151)
• Remove conda_build.cli.main_index. (#5151)
• Remove conda_build.cli.main_inspect.main. (#5151)
• Remove conda_build.cli.main_metapackage.main. (#5151)
• Remove conda_build.cli.main_render.main. (#5151)
• Remove conda_build.cli.main_skeleton.main. (#5151)
• Remove conda_build.conda_interface.IndexRecord. (#5151)
• Remove conda_build.conda_interface.CrossPlatformStLink. (#5151)
• Remove conda_build.conda_interface.SignatureError. (#5151)
• Remove conda_build.conda_interface.which_package. (#5151)
• Remove conda_build.conda_interface.which_prefix. (#5151)
• Remove conda_build.conda_interface.get_installed_version. (#5151)
• Remove conda_build.config.python2_fs_encode. (#5151)
• Remove conda_build.config._ensure_dir. (#5151)
• Remove conda_build.config.Config.CONDA_LUA. (#5151)
• Remove conda_build.config.Config.CONDA_PY. (#5151)
• Remove conda_build.config.Config.CONDA_NPY. (#5151)
• Remove conda_build.config.Config.CONDA_PERL. (#5151)
• Remove conda_build.config.Config.CONDA_R. (#5151)
• Remove conda_build.environ.clean_pkg_cache. (#5151)
• Remove conda_build.index.update_index. (#5151)
• Remove conda_build.inspect_pkg.dist_files. (#5151)
• Remove conda_build.inspect_pkg.which_package(avoid_canonical_channel_name). (#5151)
• Remove conda_build.inspect_pkg._installed. (#5151)
• Remove conda_build.metadata.Metadata.name(fail_ok). (#5151)
• Remove conda_build.os_utils.ldd.get_package_files. (#5151)
• Remove conda_build.os_utils.liefldd.is_string. (#5151)
• Remove conda_build.os_utils.liefldd.codefile_type_liefldd. (#5151)
• Remove conda_build.os_utils.liefldd.codefile_type_pyldd. (#5151)
• Remove conda_build.os_utils.liefldd.codefile_type. (#5151)
• Remove conda_build.os_utils.pyldd.mach_o_change. (#5151)
• Remove conda_build.os_utils.pyldd.is_codefile. (#5151)
• Remove conda_build.os_utils.pyldd.codefile_type. (#5151)
• Remove conda_build.os_utils.pyldd.inspect_rpaths. (#5151)
• Remove conda_build.os_utils.pyldd.get_runpaths. (#5151)
• Remove conda_build.os_utils.pyldd.otool_sys. (#5151)
• Remove conda_build.os_utils.pyldd.ldd_sys. (#5151)
• Remove conda_build.plugin.index. (#5151)
• Remove conda_build.post.determine_package_nature. (#5151)
• Remove conda_build.post.library_nature(subdir). (#5151)
• Remove conda_build.post.library_nature(bldpkgs_dirs). (#5151)
• Remove conda_build.post.library_nature(output_folder). (#5151)
• Remove conda_build.post.library_nature(channel_urls). (#5151)
• Remove conda_build.post.dists_from_names. (#5151)
• Remove conda_build.post.FakeDist. (#5151)
• Remove conda_build.post._get_fake_pkg_dist. (#5151)
• Remove conda_build.utils.relative. (#5151)
• Remove conda_build.utils.samefile. (#5151)

• Add GoatCounter (https://www.goatcounter.com/) as an analytics tool. (#5093)

• Remove unused Allure test report collection. (#5113)
• Remove dependency on conda.plan. (#5074)
• Remove almost all dependency on conda.models.dist. (#5074)
• Replace usage of legacy conda.models.dist.Dist with conda.models.records.PackageRecord. (#5074)

• @conda-bot
• @dholth
• @jaimergp
• @jezdez
• @johnnynunez
• @kenodegard
• @msarahan
• @travishathaway
• @pre-commit-ci[bot]

• Fix linking check regressions by restoring pre-3.28 behavior for conda_build.inspect_pkg.which_package. (#5141)

• @mbargull

• Update conda_build.os_utils.liefldd.ensure_binary to handle None inputs. (#5123 via #5124)
• Update conda_build.inspect_pkg.which_package to use a cached mapping of paths to packages (first call: O(n), subsequent calls: O(1)) instead of relying on Path.samefile comparisons (O(n * m)). (#5126 via #5130)

• @kenodegard

• Update conda_build.metadata.MetaData.get_section to consistently return lists for “source” and “outputs”. (#5111 via #5112)

• Resolve duplicate package record issue in conda_build.inspect_pkg.which_package. (#5106 via #5108)
• Ensure conda_build.post._lookup_in_prefix_packages displays str(PackageRecord) instead of repr(PackageRecord). (#5106 via #5108)
• Fix finalization of recipes with multiple sources. (#5111 via #5112)
• Improve handling by catching the more general ImportError instead of ModuleNotFoundError to cover cases involving menuinst 1.x. (#5116)

• @jaimergp
• @kenodegard

• Relax conda_build.metadata.MetaData.version checks when outputs have been defined. (#5096)
• Remove lief from pyproject.toml since it causes pip check to fail. To be re-added in the future after an update to py-lief package. (#5099)

• @dholth
• @kenodegard

In the upcoming January 2024 release of conda-build, significant changes are underway. We’re set to transition to the CalVer versioning system. Additionally, we’ll be formally embracing CEP 8 to manage our release schedule. Moreover, an expedited version of CEP 9 will be adopted for deprecation handling, omitting the pending deprecation phase and streamlining the period from deprecation to removal to a mere 2 months.

• Add stblib jinja function similar to compiler to explicitly define sysroot dependencies. (#4999)
• Utilize conda-known subdirs for selector definitions, enabling conda_build to support new architectures with only an updated conda version. New OS support requires additional information for proper conda_build functionality, including UNIX-like platform designation, shared library prefix, and binary archive format for the platform. (#5009)
• Eliminate unnecessary cache clearing from conda_build.build.test. (#5031)
• Consolidate which_package implementations and replace conda.models.dist.Dist usage with conda.models.records.PrefixRecords. (#5041)

• Display package file name in get_hash_input. (#5021)
• Fall back to solved record filename to locate the downloaded tarball in get_upstream_pins. (#4991 via #5037)
• Prevent overwriting of variants in high priority cbc.yaml entries when absent in lower priority cbc.yamls. (#5039)
• Correct the check for a missing anaconda-client to display a useful error message. (#5050)
• Fix conda_index.index verbose DEBUG/INFO message logging. (#5066)

• Mark conda_build.environ.clean_pkg_cache for pending deprecation. (#5031)
• Mark conda_build.conda_interface.IndexRecord for pending deprecation. Use conda.models.records.PackageRecord instead. (#5032)
• Mark conda_build.os_utils.pyldd.is_string for pending deprecation. Use isinstance(value, str) instead. (#5040)
• Mark conda_build.os_utils.pyldd.is_codefile for pending deprecation. Use conda_build.os_utils.pyldd.codefile_class instead. (#5040)
• Mark conda_build.os_utils.pyldd.codefile_type for pending deprecation. Use conda_build.os_utils.pyldd.codefile_class instead. (#5040)
• Mark conda_build.inspect_pkg.dist_files for pending deprecation. (#5041)
• Mark conda_build.inspect_pkg.which_package(avoid_canonical_channel_name) for pending deprecation. (#5041)
• Mark conda_build.inspect_pkg._installed for pending deprecation. (#5041)
• Mark conda_build.os_utils.ldd.get_package_files for pending deprecation. (#5041)
• Mark conda_build.os_utils.pyldd.mach_o_change for pending deprecation. (#5041)
• Mark conda_build.os_utils.pyldd.inspect_rpath for pending deprecation. (#5041)
• Mark conda_build.os_utils.pyldd.get_runpaths for pending deprecation. (#5041)
• Mark conda_build.os_utils.pyldd.otool_sys for pending deprecation. (#5041)
• Mark conda_build.os_utils.pyldd.ldd_sys for pending deprecation. (#5041)
• Mark conda_build.post.determine_package_nature for pending deprecation. Use conda_build.post.get_dsos and conda_build.post.get_run_exports instead. (#5041)
• Mark conda_build.post.library_nature(subdir, bldpkgs_dirs, output_folder, channel_urls) for pending deprecation. (#5041)
• Mark conda_build.post.dist_from_names for pending deprecation. Query conda.core.prefix_data.PrefixData instead. (#5041)
• Mark conda_build.post.FakeDist for pending deprecation. Use conda.models.records.PrefixRecord instead. (#5041)
• Mark conda_build.post._get_fake_pkg_dist for pending deprecation. Use conda.models.records.PrefixRecord instead. (#5041)
• Mark conda_build.utils.relative for pending deprecation. Use os.path.relpath or pathlib.Path.relative_to instead. (#5042)

• Incorporate the conda-sphinx-theme into conda-build documentation. (#5067)
• Update certain pages to remove redundant TOC entries. (#5067)

• Implement Ruff linter in pre-commit configuration. (#5015)
• Replace black with ruff format in pre-commit setup. (#5052)
• Identify Unicode tests as incompatible with libmamba. (#5059)

• @conda-bot
• @danpetry made their first contribution in https://github.com/conda/conda-build/pull/5039
• @duncanmmacleod
• @h-vetinari made their first contribution in https://github.com/conda/conda-build/pull/4999
• @isuruf
• @jaimergp
• @jakirkham
• @kenodegard
• @mbargull
• @travishathaway
• @pre-commit-ci[bot]

• Remove glob2 dependency. As of Python 3.5, the ‘**’, operator was available to glob when using recursive=True. Builtin glob is also much faster. (#5005)
• Handle emscripten-wasm32 and wasi-wasm32 platforms. (#4813)

• Delay imports in conda command plugin until the command is used, avoiding import-time side effects. (#4949)

• When templating new recipes from a PyPI package, the build script {{ PYTHON }} -m pip install . -vv is deprecated in favor of {{ PYTHON }} -m pip install . -vv --no-deps --no-build-isolation. (#4960)

• Document ~= (compatibility release) match spec. (#4553)
• Clarify that the build prefix is activated after the host prefix. (#4942)
• Add explanation that conda-build should be run from the base environment. (#4995)

• @beeankha
• @conda-bot
• @dholth
• @DaveKaretnyk made their first contribution in https://github.com/conda/conda-build/pull/5004
• @boldorider4 made their first contribution in https://github.com/conda/conda-build/pull/4960
• @jaimergp
• @jezdez
• @jugmac00
• @kenodegard
• @ryanskeith
• @scdub made their first contribution in https://github.com/conda/conda-build/pull/4965
• @wolfv made their first contribution in https://github.com/conda/conda-build/pull/4813
• @dependabot[bot]
• @pre-commit-ci[bot]

• Delay imports in conda command plugin until the command is used, avoiding import-time side effects including unwanted logging configuration. (#4949)

• @beeankha
• @conda-bot
• @dholth
• @jezdez
• @kenodegard
• @pre-commit-ci[bot]

• Add pip to env-doc make command so function works correctly (pip is no longer added by default with the python conda package). (#4633)
• Log extra-meta data to make it easier to verify that the right extra-meta data is burned into packages (also helps to co-relate packages and their build-log). The feature was first introduced in #4303 and is now improved via the logging call. (#4901)
• Implement subcommands as conda plugins. (#4921)

• Fix handling of unknown binaries with newer (py)lief versions. (#4900)
• Disable LIEF logging to remove “Unknown format” warning message. (#4850)
• Revert enable_static default value in conda_build.config to remove “Failed to get_static_lib_exports” warning messages. (#4850)
• Avoid duplicate logging by not propagating the top-level conda-build logger. (#4903)
• Fix git cloning for repositories with submodules containing local relative paths. (#4914)

• Mark executable invocations (e.g., conda-build) as pending deprecation. (#4921)
• Mark module based invocations (e.g., python -m conda_build.cli.main_build) as pending deprecation. (#4921)

• Update pkg-spec docs to mention .conda package format. (#4633)
• Drop unnecessary Jinja package name variables from variants.rst docs file. (#4834)

• Drop duplicate get_summary call in conda_build.skeletons.pypi. (#3998)
• Fix failing resolved_packages test due to recent OpenSSL 3.0.8 release to defaults. (#4912)

• @beeankha
• @conda-bot
• @dbast
• @jaimergp
• @jakirkham
• @josegonzalez made their first contribution in https://github.com/conda/conda-build/pull/3998
• @katietz
• @kenodegard
• @rfezzani made their first contribution in https://github.com/conda/conda-build/pull/4850
• @ryanskeith
• @sven6002
• @dependabot[bot]
• @pre-commit-ci[bot]

• Noarch packages that use virtual packages have the virtual packages added to the hash contents of the package. This facilitates the building of noarch packages multiple times for different platforms with platform specific dependencies. (#4606)
• Add support for svn source credentials (svn_username and svn_password). (#4692)
• Depend on standalone conda-index instead of bundled indexing code. (#4828)
• Switch from setup.py to pyproject.toml and use Hatchling for our build system. (#4840)
• Add Python 3.11 support. (#4852)

• Ensure tests/commands are also run in the presence of run_test.* (#4429)
• Require the source when rendering a recipe that uses the load_file_data function. (#4817)
• Download packages during build into the correct subdir folder. (#4832)
• Use a unique subdir variable name when rebuilding the index for multi-output builds. (#4862)

• Inline conda index logic is pending deprecation. conda-build still provides conda-index a.k.a. conda index CLI, but uses standalone conda-index during builds. (#4828)
• Prefer the standalone conda-index package, instead of conda-build index or conda index, to use faster indexing code. (#4828)
• Mark conda_build.metadata.ns_cfg as pending deprecation. Use conda_build.get_selectors.get_selectors instead. (#4837)
• Mark conda_build.config.python2_fs_encode as pending deprecation. (#4843)
• Mark conda_build.config._ensure_dir as pending deprecation. Use stdlib’s pathlib.Path.mkdir(exist_ok=True) or os.makedirs(exist_ok=True) instead. (#4843)

• Format with black and replaced pre-commit’s darker hook with black. (#4836)
• Format with isort and add pre-commit isort hook. (#4836)
• Minor code simplification for conda_build.index.ChannelIndex._ensuredirs. (#4843)
• Enable xattr test on macOS. (#4845)

• @beeankha
• @conda-bot
• @dholth
• @duncanmmacleod
• @ffirmanff made their first contribution in https://github.com/conda/conda-build/pull/4692
• @isuruf
• @jezdez
• @jakirkham
• @jjhelmus
• @kenodegard
• @rishabh11336 made their first contribution in https://github.com/conda/conda-build/pull/4782
• @ryanskeith made their first contribution in https://github.com/conda/conda-build/pull/4843
• @pre-commit-ci[bot]

• Fix the failing git clone when source has LFS files. (#4318)
• Fix many false-positives during the detection of Perl core modules in conda skeleton cpan. (#4592)
• conda skeleton cpan now correctly adds a C compiler as dependency if the distribution contains an .xs file. (#4599)
• Install downstream packages in correct subdir. (#4763, #4803)
• Update supported Python version in setup.py. (#4804)

• Removed conda <4.13 logic. (#4677)
• conda_build.conda_interface.CrossPlatformStLink is pending deprecation in favor of using os.stat().st_nlink. (#4728)
• Drop Python 3.7 support. (#4796)

• Updated broken links to example conda recipes and updated link to the now archived conda-recipes, with additional links to AnacondaRecipes aggregated feedstocks and conda-forge feedstocks. (#4580)
• Replaced two instances of “Anaconda Cloud” with “anaconda.org”. (#4719)

• Update test matrix to run tests on all supported Python versions on Linux. Only run tests on lower & upper Python bounds for Windows and macOS. (#4691)
• Re-enable code coverage reporting to codecov. (#4767)
• Eliminate test setup’s manual clone of https://github.com/conda/conda_build_test_recipe in favor of a session fixture. (#4781)
• Use tomllib (Python 3.11+) or tomli for .toml support. (#4783)

• @beeankha
• @conda-bot
• @dbast
• @dholth
• @ernstluring made their first contribution in https://github.com/conda/conda-build/pull/4318
• @xileF1337 made their first contribution in https://github.com/conda/conda-build/pull/4592
• @jezdez
• @jakirkham
• @johnnynunez made their first contribution in https://github.com/conda/conda-build/pull/4804
• @kathatherine
• @kenodegard
• @minrk
• @peetw made their first contribution in https://github.com/conda/conda-build/pull/4662
• @sven6002 made their first contribution in https://github.com/conda/conda-build/pull/4621
• @tttc3 made their first contribution in https://github.com/conda/conda-build/pull/4580
• @dependabot[bot]
• @pre-commit-ci[bot]

• Change Zstd default compression to 19. (#4663)
• Fix build/host environment activation broken in >=3.23.0,<=3.23.2. (#4665)
• Add PREFIX/bin to PATH on Windows and remove PREFIX root from PATH on Unix. (#4665)

• Skip test suite for non-code changes. (#4664)

• @jakirkham
• @kenodegard
• @mbargull

• conda-build CLI overrode condarc’s zstd_compression_level with the default value. (#4650)

• @kenodegard
• @mbargull
• @pre-commit-ci[bot]

• Fixed regression when writing link JSON files introduced in #4603. (#4636)
• Update conda dependency. (#4635)

• @jezdez
• @kenodegard

• Outputs now support both script and files arguments. When both script and an explicit file list are given, the script is run first and then the files given in the explicit file list are packaged. (#4281)
• Add overlinking_ignore_patterns build parameter to speed up recipes where it is not helpful. (#4576)
• Add win-arm64 as a recognized platform (subdir). (#4579)
• Add opt-in environment variable to run conda in isolated mode (python -I -m conda) when invoked from conda-build. This is necessary to fix an issue when packaging conda itself. Alternative solutions (see #4628) are under investigation, so the current implementation will likely change. (#4604, #4625)
• Refactored conda_build.convert.update_lib_contents to use pathlib.Path. Mark test_cli.test_convert as xfail on Windows (something with the GitHub Windows Runner makes this particularly flaky). (#4619, #4626)

• Drop cytoolz dependency. (#4556)
• Removed internal usage of the memoized decorator. (#4593, #4615)
• Remove py2 compatibility code and use of six.py code exported from conda. (#4603)
• conda_build.environ.system_vars is marked as pending deprecation. Use conda_build.environ.os_vars instead. (#4615)
• Conda 4.2 introduced a private conda env that appears to have been a testing environment. This is no longer used and is only included in conda-build as an informational status. Removing in accordance with the feature pending deprecation starting with conda 22.11.0. (#4629)

• Add s390x selector missing from docs. (#4550)
• Update “Channels and generating an index” docs to specify additional platforms. (#4602)
• Link to https://packaging.python.org/en/latest/tutorials/packaging-projects/ over setuptools for how to pack a project. (#4632)

• Renamed canary recipe from conda.recipe to recipe. (#4584)
• Introduce canary and review builds as part of GitHub CI. (#4608, #4613)
• Removed unused AppVeyor files. (#4562)
• Removed unused .checkignore. (#4564)
• Removed unused .editorconfig. (#4564)
• Removed unused .gitmodules. (#4564)
• Removed unused .lgtm.yml. (#4564)
• Removed unused .binstar.yml. (#4616)

• @chenghlee
• @conda-bot
• @dholth
• @tnabtaf made their first contribution in https://github.com/conda/conda-build/pull/4602
• @erykoff made their first contribution in https://github.com/conda/conda-build/pull/4603
• @isuruf
• @jaimergp made their first contribution in https://github.com/conda/conda-build/pull/4604
• @jezdez
• @kenodegard
• @mariusvniekerk
• @msarahan
• @skupr-anaconda made their first contribution in https://github.com/conda/conda-build/pull/4550
• @pre-commit-ci[bot]
• @brettcannon made their first contribution in https://github.com/conda/conda-build/pull/4632

• Created function load_file_data available in Jinja templates for meta.yaml (#4465, #4480)
• Created function load_str_data available in Jinja templates for meta.yaml (#4465, #4480)
• Support using --zstd-compression-level to control the compression of v2 style conda packages. (#4467)

• When building with Python 3.10, STDLIB_DIR and SP_DIR now refer to python3.10, not the symlink python3.1. (#4479)
• Reduce verbosity of urllib3 on the default log level. (#4517)
• Fixed issue identifying DSOs from sysroots when cross-compiling. (#4529)

• Improved documentation for load_setup_py_data (#4465, #4480)
• Added documentation for load_file_regex (#4465, #4480)
• Fix prerequisites for build tutorials link. (#4478)
• Link in contributing docs. (#4532)

• Fix patch tests. (#4495)
• Added patch/m2-patch as a hard dependency. (#4495)
• Update “Artistic-2.0” license test to use a valid package. (#4516)
• Rename master branch to main. (#4515, #4531)

• @abrahammurciano made their first contribution in #4465/#4480
• @chrisburr
• @conda-bot
• @duncanmmacleod
• @jezdez
• @jakirkham
• @jugmac00 made their first contribution in #4478
• @kathatherine made their first contribution in #4515
• @kenodegard
• @stuarteberg
• @teake
• @travishathaway
• @pre-commit-ci[bot]

• Replace is_dir with scandir for channel_root. (#4273)
• Remove rpaths in PREFIX/../ that doesn’t start with PREFIX. This includes BUILD_PREFIX, SRC_DIR. Previously it was only BUILD_PREFIX. (#4287)
• Add entry_points to outputs in FIELDS schema. (#4389)
• Support for setuptools 61+. (#4430)
• Use set membership for faster indexing. (#4459)

• Remove rpaths that occur multiple times. (#4287)
• Enable bdist_conda via entry_point mechanism supported also by setuptools >=60.0.0. Usable via from setuptools import setup and setup(distclass=conda_build.bdist_conda.CondaDistribution, ...). (#4368)
• Patch setuptools’s vendored distutils.core as well. (#4434)
• Resolve conda debug failure when a trailing slash path is provided. (#4448)
• Fix import error caused by conda 4.13.0’s removal of Python 2.7 code. (#4482)

• Usage of bdist_conda via from distutils.core import setup and distclass=distutils.command.bdist_conda.CondaDistribution, as that only works for setuptools <60.0.0. (#4368)
• Remove Python 2.7 imports removed in conda 4.13.0. (#4482)

• Move Windows tests from Azure to GitHub Actions. (#4353, #4436)
• Add pyupgrade to pre-commit. (#4374)
• Move MacOS tests from Azure to GitHub Actions. (#4412, #4436, #4455)
• Update Makefile for easier testing. (#4425)
• Remove unused Travis CI configs. (#4438)
• Add MyST to the documentation system and render release notes correctly. (#4483)

• @beeankha
• @conda-bot
• @dbast
• @dholth made their first contribution in https://github.com/conda/conda-build/pull/4273
• @isuruf
• @jezdez
• @jakirkham
• @kenodegard
• @remkade made their first contribution in https://github.com/conda/conda-build/pull/4425
• @rchord made their first contribution in https://github.com/conda/conda-build/pull/4353
• @travishathaway made their first contribution in https://github.com/conda/conda-build/pull/4448
• @wimglenn made their first contribution in https://github.com/conda/conda-build/pull/4434
• @pre-commit-ci[bot]

• Adds --extra-meta key=value option which allows users to save any specified extra metadata to about.json to e.g. store the repo-url, git-sha1 or the CI run-id a package was built from.

• Old work directories will be preserved when croot and build_id are set manually

• Separate contributor related documentation into dedicated file
• Migrating to github actions for tests
• Synced file(s) with conda/infra

• @Lnaden
• @jezdez
• @cjmartian
• @beeankha
• @pre-commit
• @conda-bot

• Handle an import from the vendored auxlib library in Conda 4.11.0 better.

• cran-skeleton: more unit tests

• @kenodegard
• @jdblischak
• @jezdez
• @pre-commit

• Add limited support for platform_system/sys_platform env markers in PyPI skeleton
• cran-skeleton: Adds a flag --no-comments to remove the instructional comments from the recipe files.

• When checking for circular dependencies in cross-compiling mode, build requirements are ignored now.
• Make sure symlinked directories are found in always_include_files
• Fix pinning expressions for prerelease builds

• @isuruf
• @mbargull
• @kenodegard
• @jdblischak
• @dbast
• @jezdez
• @ChristofKaufmann

• Revert “Consider any file containing .yaml in its name as maybe a recipe file” (#4235)
• Support setting build/script_env values containing “=” (#4211)
• Drop Python 2.7 support from setup.py (#4202)
• Make variant configuration error message more informative (#4198)
• Ensure file globs are always sorted (#4186)
• Add preliminary support for prelink_message files in conda packages (#4203)

• Do not munge rpath for non Mach-O files on macOS (#4238)
• Fix Windows test file extension reported by conda-debug (#4224)

• Document build/script_env recipe option (#4211)
• Clarify wording about selecting multiple operating systems (#4139)

• @chrisburr
• @gabm
• @isuruf
• @jacobtylerwalls
• @katietz
• @kenodegard
• @marcelotrevisani
• @xhochy

• Add new centos 7 distribution cleof to rpm skeleton for s390x.  (#4181)

• Fixed bug where symlinks in symlinks caused conda build to exit.  (#4180)

• @mingwandroid
• @beckermr
• @katietz
• @beckermr

• Fix stupid error in prefix replacement  (#4177)

• @mingwandroid

• Fix noarch: python version from version-age determination  (#4174)

• @mingwandroid

• activate_in_script defaults to true  (#4120)
• Add Setting and build/noarch_python_build_age and fix tests not finding packages  (#4120)
• Allow directories as license_file source  (#4153)
• Consider any file containing .yaml in its name as maybe a recipe file  (#4120)
• Add weak_constrains and strong_constrains run_exports types  (#4125)
• Issue a single command for the upload command  (#4120)
• Print hash_inputs after upload info  (#4120)
• Add cross-r-base for cross compiling
• Add --build-id-pat option
• macOS: Delete build_prefix rpaths
• Use smarter build_number
• Combine default_structs with FIELDS
• Fix conda render indent from 4 to 2
• macOS: arm64 ci/test-suite setup
• Removing more conda-forge testing deps
• Variants: Be more informative
• more verbosity in tests
• Use MacOSX10.10.sdk, not MacOS10.9.sdk in tests  (#4120)
• Warn when files have been removed from the prefix  (#4120)

• Add conda-verify to install_conda_build_test_deps  (#4120)
• Add flaky to testing dependencies  (#4138)
• Fix tests not finding packages
• Avoid writing to the package cache in package_has_file (collisions)  (#4120)
• Change package_has_file to refresh if out of date  (#4120)
• Ensure ~/.condarc does not leak into testing_config  (#4120)
• Fix applying patches to read-only files  (#4140)
• Fix auth in aboutjson  (#4137)
• Fix skeleton URLs for CentOS 6 (EOL) and various CI fixes  (#4154)
• Fix typo in cran skeleton  (#4143)
• Force channel_targets to be considered used  (#4120)
• Fix printing bytes-like object is required, not 'str' when applying patches  (#4118)
• Set “platform” in index.json to the target platform for cross-platform builds  (#4124)
• Reduce get_rpaths_raw/patchelf disagree warnings  (#4131)
• LIEF: Allow parsing static libs to fail  (#4149)
• pass cache_dir to api.build  (#4120)
• Fix symlinks to directories
• Make post-link run_export/library_nature determination less work when CONDA_OFFLINE=1
• Remove Python 2.7 from CI matrix
• Fix test_pypi_installer_metadata (builds against python 3.9 not 3.7)
• tests: Fix test_render_with_python_arg_reduces_subspace
• tests: Update python 3 from 3.5/6 to 3.9 in many
• Set numpy default to 1.16
• tests: Fix pins for numpy_used
• tests: CI: Win: Circumvent delayed expansion
• Install patch or m2-patch, write .sh files as binary, more Win tests
• tests: Avoid issue with coverage==5.0 on Win+Py2.7
• Assume non-revisible patches
• Add flaky marker and --strict-markers to setup.cfg
• Don’t sort recipes
• Use extra R_ARGS and fix them
• shell check fix

• @mingwandroid
• @isuruf
• @mbargull
• @njalerikson
• @cjmartian
• @chrisburr
• @hugobuddel
• @kurtschelfthout

• A new feature build/ignore_run_exports_from which will ignore run_exports
• coming from a package listed in build/ignore_run_exports_from.  (#4114)

• Respect PEP440 ~= ‘Compatible release clause’  (#4113)
• Detect amalgamated patches  (#4099)
• Handle realpath properly in unsafe patch check  (#4099)
• Force channel_targets to be considered used  (#4099)
• Look for git in build_prefix in git_info  (#4099)
• Fall back to shutil.copy if shutil.copy2 fails when copying patches  (#4099)
• Fix indexing by file  (#4111)
• Helper functions to extract keys  (#4088)
• Simplify find_config_files call  (#4086)
• Refactor dict_of_lists_to_lists_of_dict  (#4075)

• @mingwandroid
• @isuruf
• @njalerikson
• @cjmartian
• @njalerikson

• Make stats output more easily human-readable  (#4069)
• Prefer meta.yaml build/error_overlinking and error_overdepending  (#4074)
• Cleanup variant processing code  (#4075)
• Add --file option to indexing  (#4076)

• Remove old rpath when loader_path is used  (#4080)
• Fix MACOSX_DEPLOYMENT_TARGET default for osx-arm64  (#4091)
• Rewrite apply_patch again  (#4092)
• Add a .* to conditional_regex  (#4092)

• @isuruf
• @njalerikson
• @cjmartian
• @mingwandroid

• Use CONDA_PACKAGE_EXTENSIONS (#4053)
• raise runtimeerror instead of calling sys.exit (#4062)
• Refactor conda_build.build.get_all_replacements (#4055)

• Do not clobber config argument in conda_build.build.build_tree (#4066)
• Use --dry-run to test that a patch applies. Fixes bug 4054 (#4067)
• Include target_platform in package build string hash (#4065)
• Fix post linking for SDKs with tapi-tbd-v4 (MacOS 11.0 and upwards) (#4048)

• Run bash with -e in outputs too #4033
• Add target to recognized fields in outputs #4034
• Various overlinking fixes for Windows #4036
• variants: remove hard-coded default path for CONDA_BUILD_SYSROOT

• @mingwandroid
• @isuruf
• @mbargull

• enable Python 3.8 on Azure Pipelines (#3841)
• which_package can be passed avoid_canonical_channel_name (#3952)
• make life easier (less shell exit-y) for those who source test scripts (#3952)
• move old host env instead of deleting it when --keep-old-work (#3952)
• convert info.d/.yaml to info/.json (#3952)
• allow manual specification of which binary files to prefix replace (#3952)
• filter out ‘.AppleDouble’ folders from recipe searches (#3952)
• re-wrote apply_patch() to be more robust (#3952)
• many fixes for DSO post-processing (#3952, #3953)
• add support for (limited) tbd parsing (#3953)
• Make sure packages in current repo data w/ features have versions without features (#3957)
• Check all sysroot locations for DSOs (#3969)
• More helpful error message if an empty string is passed as the hash (‘md5’, ‘sha1’ or ‘sha256’ fields) (#3971)
• the GIT_DESCRIBE_HASH variable will be available regardless of whether the sources of the recipe have a git tag or not (#3982)
• add apple silicon support (#4004, #4015)
• set build_platform for aid in cross compiling (#4005)
• import macho on non apple system for cross compiling (#4025)
• Add ccache as a jinja 2 function (#4026)
• Improve cpan skeleton (#4026)
• Retry moving host prefix due to Windows file locking (#4026)
• Rename ccache method from mklink to sylinks (#4028)

• conda_build.metadata: fixed typos in FIELDS (#3866)
• add spaces in CRAN templates (fixes #3943) (#3944)
• raise valid CalledProcessException in macho.otool (#3952)
• cache local_output_folder too for get_build_index (#3952)
• fix relocations when cross compiling (#3995)
• use host_platform instead of sys.platform to facilitate cross compiling (#3997)
• Fix parsing UnsatisfiableError from conda>=4.7.8 (#4001)
• allow packages to depend on themselves when cross compiling (#4011)
• set the correct SHLIB_EXT when cross compiling (#4013, #4021)
• inspect linkages with pyldd when not DLL/EXE files (#4019)
• Respect no_rewrite_stdout_env on Windows (#4026)
• Prefix replacement fixes (#4026)
• Use git am -3 when applying patches (#4026)
• Fix env_var=val assertion (#4026)
• Use exit /B from patch files (#4026)

• extend docs o generating the index (#3877)
• add details to documentation of run_constrained (#3878)
• remove documentation on bdist_conda and environment variables (#3879)
• update cli help information for conda index (#3931)
• Clarify how to install conda-build (#3976)
• Add note for local package install deps (#3980)
• Clarify multiple OS selection (#3984)
• add aarch64 selector to the docs (#4003)
• add docs on build_platform and arm64 (#4020)

• Enable s390x support (#3949, #4030)
• Add xfail test for non-utf-8 charsets (#3972)
• Improve testing on CI (#3987, #4017, #4027)
• Allow python=3.8 for pypi skeletons (#4014)

• load log prior to calling warn method (#3925)
• test suite fixes and prefix replacement fixes (#3932)

• Enable ppc64 support (#3921)

• Update cli help information for conda index (#3931)

• @beenje
• @jjhelmus
• @mingwandroid

• Show a warning instead of failing if a Mach-O file is prouduced by a build running on a platform other than macOS (#3912)
• Revert #3893, restores behavior of build/binary_has_prefix_files to that found in 3.18.12 (#3916)

• clarified ‘deletes the build environment’ in concepts/recipe.rst (#3901)

• @jjhelmus
• @timsnyder
• @chrisburr

• Fix issues with PREFIX detection in Windows #3899

• Change the CI trigger #3904

• @mingwandroid
• @marcelotrevisani
• @jjhelmus

• Keep python pinning in hashing if there is a space #3895
• ci launcher supporting python d shebangs on Windows #3894
• Allow build/binary_has_prefix_files to specify a list of files #3893

• Use patchelf to set RPATH by default #3897

• @isuruf
• @jjhelmus
• @mingwandroid

• Keep python pinning in hashing if there is a space #3895
• ci launcher supporting python d shebangs on Windows #3894
• Allow build/binary_has_prefix_files to specify a list of files #3893
• Use patchelf to set RPATH by default #3897
• Prevent non-atomic writes to repodata JSON files #3833
• Audited and updated all docs with formatting, grammar, and accuracy errors.
• Docs: Removed deprecated page on features
• Fixed issue where symlinks to files that do not exist break conda build #3840

• @bdice
• @beckermr
• @chrisburr
• @csoja
• @guidara
• @isuruf
• @jakirkham
• @jjhelmus
• @marcelotrevisani
• @mcg1969
• @mingwandroid
• @msarahan
• @rrigdon
• @saraedum
• @sscherfke
• @zeehio

• Update build.sh files of skeletons to be shellcheck clean including test to lint future updates.
• Corrected documentation on subpackage test requirements.
• Do not move work dir to work/work/
• fixed a missing .lower() on two tar_xf related util functions
• Fix has_prefix detection for Windows.
• conda_build.inspect_pkg: optimise use of fnmatch
• Do not consider .ignore files when searching with ripgrep
• Remove N*N os.lstat calls in build_info_files_json_v1

• @msarahan
• @rrigdon
• @marcelotrevisani
• @rrigdon
• @soapy1
• @dbast
• @duncanmmacleod
• @beckermr
• @seanyen
• @AndrewAnnex
• @183amir
• @njzjz

• Added the error message when an invalid pip dependency version expression is used
• Conda skeleton pypi quoting just version, summary and description or attributes with special characters
• Set up CI Azure pipeline for Linux
• Update cran skeleton to match supported optional licenses for license file derivation.
• Migrate Unittests to PyTest
• Update script command on conda skeleton pypi to use {{ PYTHON }} -m pip install . -vv
• Add a warning when a received a file on RECIPE_PATH
• Refactored the skeletons/pypi.py get_package_metadata to be more modular
• added --suppress-variables switch to hide environment variables from console output

• Fixed build of ‘.conda’ packages enabled via conda config --set conda_build.pkg_format 2
• Workaround for future deprecations of the SafeConfigParser and readfp of the same module.

• Remove bzip2 package from build toolkit description.

• @msarahan
• @jakirkham
• @marcelotrevisani
• @duncanmmacleod
• @kinow
• @saraedum
• @jjhelmus
• @rrigdon
• @mingwandroid
• @asford
• @timsnyder
• @mcg1969
• @kaitietz
• @stuarteberg
• @isuruf
• @dbast
• @Bezier89

• add --use-channeldata argument to conda render/build.
• Extract the part in the skeletons pypi responsible to get the package metadata to a free function.
• Creat unittests for the get_package_metadata (skeletons/pypi.py) and for the new functions.

• Limit threads to 61 on Windows.
• Do not use channeldata for run_exports unless --use-channeldata specified.
• Finalize top-level metadata if not present as an output.

• Add 3.18.7 release notes

• Add disable_pip to FIELDS

• @rrigdon
• @jjhelmus
• @rrigdon
• @Bezier89
• @jakirkham
• @marcelotrevisani

• license_file can optionally be a yaml list

• fix readup of existing index.json in cache while extracting
• fix spurious post build errors/warning message
• merge channeldata from all urls

• @msarahan
• @rrigdon
• @jjhelmus
• @isuruf
• @ddamiani

• Update authorship for 3.18.7
• Add note on single threading for indexing during build
• Add in fallback for run_exports when channeldata not available
• Make pins for current_repodata additive - always newest, and pins are additions to that
• Limit indexing in build to using one thread
• Speed up by allowing empty run_exports entries in channeldata be valid results
• Bump conda-package-handling to 1.3+
• Add test for run_exports without channeldata
• Fallback to file-based run_exports if channeldata has no results
• Add Mozilla as valid license family
• Add in fallback for run_exports when channeldata not available
• Updated tutorials and resource documentation

• Flake8 and test fixes from pytest deprecations
• Fix in render.py::_read_specs_from_package
• Fix for pkg_loc
• Fix conda debug output being suppressed

• @msarahan
• @rrigdon
• @rrigdon
• @scopatz
• @mbargull
• @jakirkham
• @oleksandr-pavlyk

• package sha256 sums are includex in index.html

• fix bug where package filenames were not included in the index.html

• @rrigdon
• @jjhelmus

• fix one more keyerror with missing timestamp data
• when indexing, allow .tar.bz2 files to use .conda cache, but not vice versa.  This acts as a sanity check on the .conda files.
• add build/rpaths_patcher to meta.yaml, to allow switching between lief and patchelf for binary mangling

• @mingwandroid
• @msarahan
• @csosborn

• channeldata reworked a bit to try to capture any available run_exports for all versions available

• make “timestamp” an optional field in conda index operations

• @msarahan

• Make VS2017 default Visual Studio
• Add hook for customizing the behavior of conda render
• Drop /usr from CDT skeleton path
• Update cran skeleton to use m2w64 compilers for windows instead of toolchain. The linter is telling since long: Using toolchain directly in this manner is deprecated.

• Update cran skeleton to not use toolchain for win
• fix package_has_file so it supports .conda files (use cph)
• fix package_has_file function for .conda format
• fix off-by-one path trimming in prefix_files
• disable overlinking checks when no files in the package have any shared library linkage
• try to avoid finalizing top-level metadata twice
• try to address permission errors on Appveyor and Azure by falling back to copy and warning (not erroring) if removing a file after copying fails
• reduce the files inspected/loaded for channeldata, so that indexing goes faster

• The repodata2.json file is no longer created as part of indexing.  It was not used by anything.  It has been removed as an optimization.  Its purpose was to explore namespaces, and we’ll bring its functionality back when we address that fully.

• @mingwandroid
• @msarahan
• @rrigdon
• @rrigdon
• @soapy1
• @mariusvniekerk
• @jakirkham
• @dbast
• @duncanmmacleod

• speed up post-link checks
• fix activation not running during tests
• improve indexing to show status better, and fix bug where size/hashes were being mixed up between .tar.bz2 and .conda files

• @mingwandroid
• @msarahan
• @rrigdon

• rearrange steps in index.py to optimize away unnecessary work
• restore parallel extract and hash in index operations

• @msarahan

• Set R_USER environment variable when building R packages
• Make Centos 7 default cdt distribution for linux-aarch64
• Bump default python3 version to 3.7 for CI
• Build docs if any docs related file changes
• Add support for conda pkgv2 (.conda) format
• add creation of current_repodata.json - like repodata.json, but only has the newest version of each file
• change repodata layout to support .conda files.  They live under the “packages.conda” key and have similar subkeys to their .tar.bz2 counterparts.
• Always show display actions, regardless of verbosity level
• Ignore registry autorun for all cmd.exe invocations
• Relax default pinning on r-base for benefit of noarch R packages
• Make conda index produce repodata_from_packages.json{,.bz2} which contains unpatched metadata
• Use a shorter environment prefix when testing on unix-like platforms
• Prevent pip from clobbering conda installed python packages by populating .dist_info INSTALLER file

• Allow build/missing_dso_whitelist section to be empty
• Make conda-debug honor custom channels passed using -c
• Do not attempt linkages inspection via lief if not installed
• Fix all lief related regressions brought in v3.17.x
• Fix ZeroDivisionError in ELF sections that have zero entries
• binary_has_prefix_files and text_has_prefix_files now override the automatically detected prefix replacement mode
• Handle special characters properly in pypi conda skeleton
• Build recipes in order of dependencies when passed to CB as directories
• Fix run_test script name for recipes with multiple outputs
• Fix recursion error with subpackages and build_id
• Avoid mutating global variable to fix tests on Windows
• Update CRAN license test case (replace r-ruchardet with r-udpipe)
• Update utils.filter_files to filter out generated .conda_trash files
• Replace stdlib glob with utils.glob. Latter supports recursion (**)

• Updated Sphinx theme to make notes and warnings more visible
• Added tutorial on building R-language packages using skeleton CRAN
• Add 37 to the list of valid values for CONDA_PY
• Corrected argparse rendering error
• Added tutorials section, reorganized content, and added a Windows tutorial
• Added Concepts section, removed extraneous content
• Added release notes section
• Reorganized sections
• Clarify to use ‘where’ on Windows and ‘which’ on Linux to inspect files in PATH
• Add RPATH information to compiler-tools documentation
• Improve the documentation on how to use the macOS SDK in build scripts.
• Document conda build purge-all.
• Fix user-guide index
• Add example for meta.yaml
• Updated theme
• Reorganized conda-build topics, updated link-scripts

• @mingwandroid
• @msarahan
• @rrigdon
• @jjhelmus
• @nehaljwani
• @scopatz
• @Bezier89
• @rrigdon
• @isuruf
• @teake
• @jdblischak
• @bilderbuchi
• @soapy1
• @ESSS
• @tjd2002
• @tovrstra
• @chrisburr
• @katietz
• @hrzafer
• @zdog234
• @gabrielcnr
• @saraedum
• @uilianries
• @theultimate1
• @scw
• @spalmrot-tic

• provide fallback from libarchive back to python tarfile handling for handling tarfiles containing symlinks on windows

• Rever support added for releasing conda-build

• @msarahan
• @jjhelmus
• @scopatz
• @rrigdon
• @ax3l
• @rrigdon

• respect context.offline setting  #3328
• don’t write bytecode when building noarch: python packages  #3330
• escape path separator in repl  #3336
• remove deprecated sudo statement from travis CI configuration  #3338
• fix running of test scripts in outputs  #3343
• allow overriding one key of zip_keys as long as length of group agrees  #3344
• fix compatibility with conda 4.6.0+  #3346
• update centos 7 skeleton (CDT) URL  #3350

• @iainsgillis
• @isuruf
• @jjhelmus
• @nsoranzo
• @msarahan
• @qwhelan

• don’t raise when recipe text can’t be extracted if manual build string is already set  #3326

• @msarahan

• fix pip build isolation / fix absence of “falsey” env vars.  Ignore only if empty string or None.  #3319
• pass-through VS20XYINSTALLDIR var (used by intel compiler to locate VS2017 installation)  #3322

• @jjhelmus
• @msarahan

• fix python-3 only JSON decode error handling (make py27 compatible)  #3307
• fix too much caching in getting used vars from meta.yaml leading to inaccurate hash contents  #3311
• fix merge of build/host not being recognized before an rm_rf call utilized that info  #3311

• @Lnaden
• @msarahan

• ignore non-native binaries in lief for now.  Cross-platform inspection still theoretically possible using subdir parameter.  #3306

• @msarahanl

• fix to ignore unsatisfiable pin_compatible calls for packages in other outputs  #3277
• add license files to CRAN recipes generated by conda skeleton  #3284
• restrict py-lief to running on linux/macos only for now  #3291,
• fix lief operation on files that are missing dynamic section (e.g. go binaries)  #3292
• expand instructions on how to setup a dev env for conda-build  #3296
• fix file= keyword being passed to a logger call #3298
• add test for standalone DLLs with py-lief, don’t error out on them  #3301
• rename windows build script runner to avoid confusion with existing bld.bat files in root dir  #3303
• fix file URL handling on Windows  #3303
• use conda’s download function rather than requests directly, so that conda’s proxy settings are respected  #3303
• silence patch output when output verbosity is False  #3305

• @bergtholdt
• @dsludwig
• @jdblischak
• @msarahan
• @nehaljwani
• @sodre

• omit LIEF depedency on Windows until it is better tested #3288
• activate host environment #3288
• allow calls to nm to fail #3290

• @jjhelmus
• @msarahan
• @nehaljwani

• tell pip to not go find things on PyPI (turn off downloading)  #3109
• new “conda debug” command for creating build/host or test envs for working on recipes  #3237
• new package check: “overdepending” - warns or errors out when your run dependencies include unnecessary shared library packages  #3237
• utilize LIEF for analyzing shared object data, extending capabilities beyond pyldd  #3237

• avoid discarding build string during pin_run_as_build and ensure_valid_spec  #3264
• fix conda index’s handling of packages where ‘depends’ key doesn’t exist  #3270
• fix bad inversion assumption about pip’s PIP_NO_DEPENDENCIES setting  #3271
• fix regex to allow for whitespace after hyphens in outputs section  #3274, #3275
• handle unicode decode fails in output rewriting  #3279
• fix merge of repodata patches that have keys that don’t exist in repo  #3280

• @bergtholdt
• @isuruf
• @minrk
• @msarahan
• @mingwandroid
• @nehaljwani

• rewrite long prefix paths as $PREFIX, etc. for more readable build logs  #3258
• make the --output-folder switch configurable in condarc  #3265
• make the --long-test-prefix switch configurable in condarc, fix logic error with that argument  #3266

• improve robustness of indexing in face of corrupt package data #3238
• change timeouts to 900 instead of 90  #3239
• add activation to wheel bundling script  #3240
• fix PermissionError import from utils, undefined on py2.7  #3247
• fix outputs with custom build string getting hash incorrectly  #3250
• fix tests not being run on windows  #3257

• @Bezier89
• @gabm
• @isuruf
• @minrk
• @msarahan
• @teake
• @tomashek
• @tschoonj

• Remove noarch binary file check (do this in conda-verify instead)  #3212
• Fix utf-8 conversion of check_output_env  #3213
• fix thread count when indexing causing oversubscription  #3217
• fix behavior of try_acquire_locks during lock contention  #3224
• fix test env creation improperly prioritizing local channel  #3229
• don’t check for error when removing conda-init (in conda recipe for this repo)  #3230
• add r-impl to R package template generator  #3232
• fix creation of unix and win shell script files for noarch packages  #3236
• fix path of python interpreter used for noarch packages being tested on win, when created on linux/mac  #3236

• @alexandersturm
• @Bezier89
• @dsludwig
• @mandeep
• @mingwandroid
• @msarahan
• @rchateauneu
• @soapy1

• expand ~ in source paths  #3206
• Use binsort when available to sort file list in tar archives, to optimize compressibility  #3210
• allow meta.yaml’s build/rpaths key to function on macOS, not just linux  #3206

• improve docs on behavior of channel arguments  #3197
• remove mention in docs about building .RPM and .DEB files.  #3199
• fix dist-info errors where dist-info files didn’t match the package name  #3206
• fix some hard-coded .tar.bz2 references, to support other compression formats more readily  #3206
• batch calls to compiling .pyc files to avoid problems with maximum command length  #3206
• use ensure_list in processing files to be extracted  #3210
• fix KeyError that happened when a jinja2 rendering error occurred, which hid the rendering error  3211

• @mingwandroid
• @msarahan
• @stas00
• @teake

• incorporate libarchive to support more compression formats (adds libarchive as a package dep)  #3163
• add build/ignore_verify_codes key to allow recipes to ignore specific conda-verify error codes  #3179

• fix JSON string encoding error in index cache reading  #3156
• restore --variants CLI flag for specifying variants  #3168
• handle empty build section in output  #3175
• add mro-base_impl as something that causes mro build strings  #3163
• fix skeleton PyPI inappropriately dropping package case (needed for URLs)  #3163
• fix packages from earlier loops with multiple outputs being removed prior to later loops  #3185
• fix conda-index not removing entries from index that no longer exist on disk  #3186 #3188
• clean up tempfiles after indexing  #3187
• fix indexing of specific subdirs  #3190
• fix pypi skeleton when python constraint has no operator  #3191
• fix issues testing packages and recipe folders when done separately from build  #3192
• fix source looking for patches in wrong folder when dealing with outputs  #3194

• @dpryan79
• @gabm
• @mbargull
• @mingwandroid
• @msarahan
• @nehaljwani

• sort “removed” fns in index repodata.json  #3154
• fix deps being merged instead of clobbered  #3154
• Handle corrupt packages during indexing better  #3154

• @msarahan

• add CLI flag (–strict-verify) to allow erroring out when conda-verify fails a package  #3135
• output text stating that the license file has been successfully found and included with a package  #3152

• allow channel auth when checking if a package is built  #3133
• If local git cache can’t be updated, delete it and barf (for user to re-run)  #3136
• clean up duplicate pip requirements produced by skeleton  #3138
• replace recipe_log.txt file with recipe_log.json file (for passing recipe history along with package)  #3139
• fix decoding to str before passing package contents to JSON loading  #3140
• fix loss of “removed” section of index with every other indexing operation  #3144
• fix update_index used in tests to index channel, not subdir  #3145
• fix ELF sections not included in memory image of process being loaded by pyldd and giving misleading results  #3148
• fix index operations outputting debug log messages  #3151
• fix private channels showing 404 errors during test phase  #3153

• @Bezier89
• @gabm
• @jakirkham
• @jjhelmus
• @kalefranz
• @msarahan
• @stuarteberg
• @teake

• fix recipe_log.txt not being filtered from info/files  #3134

• @msarahan

• add support for index patch instructions as tarballs containing subdirs  #3129
• add progress bars for indexing (using tqdm)  #3130

• fix log messages being deduplicated too much  #3130
• handle permission errors with moving files in indexing more gracefully  #3132

• @msarahan

• add support for a “recipe log” file.  This will be used at Anaconda to capture the commit activity of a given recipe, which will be published in the RSS feed.  #3123

• fix indexing of noarch subdir as done by conda-forge  #3120
• decode cached index files to utf-8 before reading JSON  #3121
• try to address unicode problems in run_exports handling  #3121
• skip over index metadata files when they are not present in a package  #3125

• @msarahan

• detect and fall back to old update_index behavior (new is channel-wide; old is specific subdir)  #3117
• fix CONDA_BUILD_STATE not being set when load_setup_py_data gets run  #3117
• fix channel_name as CLI argument for conda index.  It can’t be positional.  #3318

• @msarahan

• refactor indexing to cache more efficiently  #3091
• add tags, identifiers, and keywords to about section.  Tie them into channeldata.json.  #3091
• filter .la files from packages by default  #3102
• memoize read_meta_file  #3108

• fix --check for optionally iterable fields  #3098
• fix permission problems prior to fixing shebangs  #3101
• do not disable pip’s cache directory.  Redirect it instead.  #3104
• fix usage of config in load_setup_py_data  #3110
• show logger message when default numpy is used, to communicate what’s happening to the user  #3110

• drop python 3.4, add 3.6, 3.7 to skeleton pypi  #3103

• @jjhelmus
• @kalefranz
• @msarahan
• @nehaljwani
• @nsoranzo
• @ocefpaf
• @teake

• add run_exports and aggregated post-install metadata indexing outputs  #3060
• allow whitelisting runpath entries  #3072
• consider *_compiler_version entries when looping over variants (allow *_compiler_version to be a used variable that affects the hash)  #3084

• fix cached git info for variants  #3082
• fix linux temporary channel not being added at test time, leaving package unresolvable  #3088

• @msarahan
• @teake

• add the “extra” field of a package’s meta.yaml file into the output package’s info/about.json file  #3048
• add option to omit local channel in is_package_build (used by c3i)  #3051
• add pip env vars to prevent it from pulling in external dependencies when used in build scripts  #3053

• fix local channel always being top priority.  Allow user-defined channel orders where local is lower than remotes.  #3049
• Fix conda-verify import error warning showing up in --output text  #3052
• fix RPM skeleton test (point to newer CentOS repo) #3054
• fix test/files and test/source_files looking in the wrong place for info/recipe/parent contents (subpackages)  #3061

• @Bezier89
• @jakirkham
• @mikecormier
• @mingwandroid
• @msarahan

• Allow user-specified channels to come ahead of local channel  #3038
• Add schema for outputs section in FIELDS; provide method for getting rendered recipe text (to support conda-verify)  #3041
• Enable conda-verify by default when it is importable, but only print warnings by default, instead of exiting  #3042
• Add --label CLI argument to allow specifying label for uploading packages to  #3043

• fix apply_selectors, leading to excessive detection of used variables  #3040

• @CJ-Wright
• @msarahan
• @speleo3

• improve environment marker support for pypi skeleton  #2972
• apply selectors before checking requirements, to better understand per-platform used vars  #2973
• Handle conda UnsatisfiableError causing packages to be moved to broken folder without tests actually being run on them  #2974 #2975
• use tempfiles when writing index to reduce risk of corrupt index  #2978
• handle conda index recipe info for older versions of conda-build  #2979
• allow empty missing_dso_whitelist in build section  #2983
• fix host_reqs referring to a detached list, leading to requirements/host not being modified by run_exports  #2987
• fix for bypassing MITM proxies based on SSL_NO_VERIFY env var  #2991
• add missing_dso_whitelist to FIELDS  #2994
• Don’t skip logic in pyldd based on CB verbosity (–quiet)  #2999
• Convert empty git refs to HEAD, so that git_url behavior is more predictable  #3003
• set NPY_DISTUTILS_APPEND_FLAGS=1 so the compiler package flags are respected  #3015
• fix script file renaming when converting package from win to unix  #3014
• allow fn to be omitted when using multiple url sources  #3021
• fix default config settings being shared across Config instances  #3022
• force text interpretation of CRAN DESCRIPTION files  #3020
• fix is_no_link to honor patterns  #3023
• fix test/requires being ignored when --no-copy-test-source-files is specified  #3027
• fix up dependencies in CRAN skeleton output  #3030 #3032

• change skeleton pypi to generate recipes that use pip for install step.  Remove description.  #2972
• Set environment variable to disable pip environment isolation to prevent problems  #2972
• support multiple exclusive_config_files  #3022

• Fix search order for conda_build_config.yaml  #3029

• @isuruf
• @j-hartshorn
• @kalefranz
• @mandeep
• @mbargull
• @mcg1969
• @mingwandroid
• @minrk
• @msarahan
• @nehaljwani
• @ocefpaf
• @tjd2002