Conda-Forge has been providing support for Python 3.7 for 4 years now.

Increasingly projects are moving off it (particularly in the PyData community). With Python 3.11's release coming around the corner (October 3rd), conda-forge plans to drop Python 3.7 support when Python 3.11 comes out. This will lighten the load on conda-forge infrastructure and make room for the new versions the community would like to support.

More details can be found in issue conda-forge-pinning-feedstock#2623. Feedback is welcome there.

Conda-forge has supported PyPy since almost 2.5 years now, and the initial PyPy 3.7 builds have been superseded in almost all aspects by the newer builds for 3.8 & 3.9. We are therefore dropping PyPy 3.7 as a supported python version, and will keep focusing on the more contemporary PyPy builds.

Microsoft has deprecated the Visual Studio (VS) 2017 compiler and removed it from all the CI they control (notably Azure Pipelines & Github Actions). This means that the default toolchain (== C/C++ compiler, linker, standard libraries, and related utilities) of that VS version - vc141 - is getting less and less use in upstream libraries (because public hosted CI doesn't use it anymore by default), and therefore support for it is bitrotting at an accelerating pace. We are therefore planning to move our toolchain on windows to vc142 (the default in VS2019) in two weeks, on 2022-08-25.

This will not affect you as a general user of conda-forge packages on windows; the only impact is that if you are locally compiling against artefacts produced by conda-forge and are still using VS2017 yourself, you will need to upgrade your compiler (VS2019 is a drop-in replacement & ABI-compatible).

Azure is removing their OSX 10.15 VM image and so we are bumping to 11. You will need to rerender your feedstock to get this change. Feedstocks without the new VM image specified will not build after Azure fully removes the old image. Please get in touch with us if you have issues or questions!

After more than six months, the conda-forge team and contributors have managed to update the Qt5 packages to the latest LTS version, 5.15.2. Major changes include separating the package for QtWebEngine (qt-webengine) from the rest of Qt (now in a new package called qt-main). This allows recipes that do not use any of the WebEngine components to depend only on qt-main, reducing the total size of the downloaded binaries. As a result of this, qt will be a metapackage that installs both qt-main and qt-webengine as dependencies. With respect to PyQt, the new packages now are in sync with respect to their corresponding PyPI releases, which means that the pyqt package will only provide the core components of Qt, leaving pyqtwebengine and pyqtcharts as optional packages that extend PyQt by providing the QtWebEngine and QtCharts components, respectively. A migrator will be put in place to help with the transition.

A GitHub action now monitors comments on issues in staged-recipes and will add language and review labels to issues/PRs when a staged-recipes sub-team is mentioned in a comment. It adds the Awaiting author contribution label if a member of staged-recipes removes the review-requested label. Unlike notifications, which are only sent to the users which are members of a team at the time of the mention, labels are persistent and visible to everyone, so they should be very helpful for identifying old PRs that need attention.

Due to technical issues in generating new feedstocks, we have deprecated using CircleCI and Drone.io for builds of new feedstocks. Existing CircleCI builds, if any, should be moved to azure. Existing Drone.io builds can be moved to Travis CI or cross-compiled/emulated builds on azure.

We have begun rolling out packages built for PyPy3.8 and PyPy3.9. This work may take a few weeks. See pypy in the user docs for information on how to set up a PyPy environment. Please report issues to the PyPy developers at https://foss.heptapod.net/pypy/pypy. We are also dropping PyPy3.7 in each feedstock as the newer versions of PyPy are added. New versions of migrated feedstocks will not be built for PyPy3.7 and that version of the python interpreter will not be receiving updates. As usual, you can track the status of the migration on our status page.

Due to changes in how Travis CI tracks open-source build time, we are deprecating using it for the win_*, osx_*, and linux_64 platforms.

Travis CI will be available only for platforms in their partner queues. These platforms currently include ppc64le, aarch64 and s390x.

Rerendering will raise an error if Travis CI is used for a non-partner queue platform in the conda-forge GitHub organization.

We will be migrating the default branches of all feedstocks and other conda-forge repos from master to main. We do expect some minor hiccups while this migration is going on. You will need to change to the main branch from master on any local clones via the following git commands:

git branch -m master main
git fetch origin
git branch -u origin/main main
git remote set-head origin -a

If you encounter any problems, please comment on this Github issue.