(Q) How to address
permission denied (publickey) in circle builds?
When you see this error in a circle build:Permission denied (publickey). fatal: Could not read from remote repository. Please make sure you have the correct access rights and the repository exists. Exited with code 128
Open a new issue with a comment @conda-forge-admin, please update circle. Once our web services have updated the circle configuration, restart the build.
(Q) Why does conda-build ignore the
py37 selector in meta.yaml?
conda-build has changed the selector syntax. You are now encouraged to use
py==<version>, instead of
py<version>. While the legacy selectors
py36are still valid, selectors
py37and higher are no longer valid.
Please use the new syntax
py==37to avoid confusion.
(Q) What do build numbers above 1000 signify? How do I treat them?
TL;DR: there is no need for build numbers larger than 1000 anymore.
When you update a feedstock that still uses build numbers > 1000, the following rules apply:
When you increase the version, reset the build number back to 0 (e.g.
1005 -> 0).
When the version stays the same and you need to upload a new package, increase the build number by 1 (e.g.
1005 -> 1006).
Backstory: Build numbers of 1000 and larger are a relic from the compiler migration, where a build number offset of 1000 signified that a package was migrated to the new compilers. Since the completion of the compiler migration, this offsetting is not needed anymore. However, we cannot simply subtract the offset without updating the version, due to higher build numbers being preferred by the solver. Therefore build numbers above 1000 will gradually vanish as packages get updated to newer versions.
(Q) How to fix CMake not finding MSBuild.exe on Azure Windows builds?
NMake Makefiles JOMas the CMake generator (
cmake -G"Ninja"), and add
Sadly in the Azure Windows images, MSBuild.exe is not correctly setup for CMake builds with the
Visual Studiogenerators. To work around this, you can use a different CMake generator, e.g.
cmake -G"NMake Makefiles JOM". These two are preferred because they allow for concurrent builds in contrast to e.g. only using
cmake -G"NMake Makefiles"
(Q) Why does my new version appear on Anaconda Cloud, but is not installable with conda?
For certain, high-traffic channels (main & conda-forge), Anaconda uses a CDN to decrease costs. The CDN is only reindexed every 20 minutes, however Anaconda.org uses the original channel that the CDN mirrors. Therefore, packages will show up on the Anaconda Cloud about 20 to 40 minutes before they are downloadable via conda. You can use
conda search <pkg>to see if the package is installable, because this command reads from the CDN.
(Q) How can I make local debugging faster?
If you prefer to debug your recipes locally and not use the provided scripts but instead your own setup, you may also use the mamba solver through
mambabuild. It not only has a faster solve speed but also prints better error messages that make debugging simpler.
To do this, first install the solver and then build the recipe like you normally would
conda install boa -c conda-forge
conda mambabuild myrecipe
For more details visit this page.
(Q) I am seeing
Importing conda-verify failed. error message during build. What do I do?
Importing conda-verify failed. Please be sure to test your packages. conda install conda-verify to make this message go away.
You are seeing this error message because by default, conda-build uses conda-verify to ensure that your recipe and package meet some minimum sanity checks. This message can be safely ignored as conda-forge doesn’t use conda-verify.
(Q) When the bot creates a pull request to a feedstock to update the version, should I approve the pull request and wait with merging until everybody else that is a code owner has approved the PR?
There is no need to approve the PR. Every maintainer can verify and merge the bot PR without waiting on the approval of the other maintainers.
(Q) How to fix “build-locally.py fails with exit code 139”?
With Linux Kernel 4.11 there were some changes in the
vsyscalllinking. Depending on your distribution this may cause the above error. You can fix that on Debian by editing
GRUB_CMDLINE_LINUX_DEFAULT="vsyscall=emulate"in this file. Afterwards, you need to run
update-gruband reboot your system. On other Linux distributions the fix is similar but you need to edit a different configuration file to change the Linux kernel cmdline. This workaround is only needed for images based on CentOS 6 (
cos6). You could also workaround this by forcing the CentOS 7 based images using
The exit code 139 itself actually is the general exit code for a segmentation fault. This could also mean that you have run into a different issue but the above issue is the most likely one with our CentOS 6-based images.
(Q) Is it necessary for me to be an upstream maintainer of the package I submit to Conda-forge?
Everybody can submit a package to Conda-forge, irrespective of whether they maintain the upstream version or not. Additionally, it’s not required but considered good practice to inform the upstream of a new package and invite them to be maintainers as well.
(Q) How do I fix the
libGL.so.1 import error?
Error:ImportError: libGL.so.1: cannot open shared object file: No such file or directory
To fix the error, create a yum_requirements.txt file and add mesa-libGL.
(Q) How can I fix the
The Qt platform plugin "xcb" could not be loaded error during testing?
When testing packages that have a dependency on
pyqt, the following error might occur under linux:qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found. This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem. Available platform plugins are: eglfs, minimal, minimalegl, offscreen, vnc, webgl, xcb.
This comes from the CI environment being headless and can be fixed by adding the
QT_QPA_PLATFORM=offscreenenvironment variable. The variable can either be added directly to the test command or provided in the meta.yaml like so:build: script_env: - QT_QPA_PLATFORM=offscreen
(Q) How can I contact conda-forge/core?
When in an issue or PR, you can contact conda-forge/core by simply mentioning
@conda-forge/corein a comment. If you don’t receive an an answer after a couple of days, feel free to reach out to us via the public gitter channel.
Due to a GitHub limitation, this is disabled for new members. In that case, you can ping core using the bot command @conda-forge-admin, please ping conda-forge/core.
In case your issue is longer or you would like to contact us privately, feel free to reach out to us via the options listed here.
(Q) A feedstock has been abandoned and I would like to take over maintenance.
A feedstock is generally considered abandoned when the maintainer isn’t around anymore and doesn’t merge new PRs or answer any issues. If that is the case, you can add yourself to the team by using the @conda-forge-admin, please add user @username command. If the maintainer doesn’t merge it after roughly a week, contact conda-forge/core to have it merged. Once added, you have full rights to the feedstock and can continue its maintenance.
Even if the maintainer isn’t active anymore, we generally like to keep them in the list of maintainers and not remove them, in case they want to take up maintenance at a later date.