Open daler opened 2 years ago
Re unitig-caller:
The build failure is unrelated to Python 3.10. Instead the problem is the SeqAn3 version.
The last time this package was successfully built was in February 2022, when it was built against SeqAn3 3.1.0. Then in June SeqAn3 3.2.0 was released. The new SeqAn3 version requires C++20, but more importantly, the current unitig-caller source code is not compatible with some SeqAn3 3.2 API changes: see bacpop/unitig-caller#13.
We could work around that by requiring seqan3 >=3.1.0, <3.2.0
. But first we should see what upstream's response to the filed issue is.
Overview
As part of the recent migration that re-built many packages for Python 3.10, we encountered issues with recipes that either haven't been re-built in a while or that specifically had issues with Python 3.10. These were dealt with in one of two ways. If there was an obvious specific issue (e.g., an issue unique to Python 3.10 or unique to Mac OSX, for example), then a preprocessing selector with
skip: True
was used. E.g.:Otherwise, the recipe was added to the
build-fail-blacklist
at the top of the repo.If we didn't do this, then it would hold up the building of other packages, and further delay the process of migrating to Python 3.10.
What we need help with
Pick a package from the list below. Try to get it working.
If you can get it working, great -- create a PR that passes, check the box here, and edit the issue description to include the PR (see example below).
If you can't get it working, then add a description somewhere -- either in the blacklist or in the recipe itself. Open a PR with those changes (which will probably just be comments/documentation). And similarly edit this issue with the PR and cross it off the list.
On your PR, add the
addressed-bulk-skip
label so we can easily filter and keep track of which PRs are specifically addressing this issue. You can reference this issue in your PR as well if you want.While editing this issue, a package that has been addressed should look like this in the issue text entry box:
which will show up like this:
Recently blacklisted packages
These packages were blacklisted in the latest re-build. Remove them from the
build-fail-blacklist
, and then try building them.auspice has error about symlink of python3 within node dir; nextstrain depends on auspice
These have a sha256 mismatch, so they need review to figure out if that is expected.
These depend on last, which needs sha256 review above:
Compiler errors for python-bioexp; python-hivclustering and hivtrace depend on python-bioext:
conda_build.source:Error: HTTP 503 SERVICE TEMPORARILY UNAVAILABLE for url <https://abibuilder.informatik.uni-tuebingen.de/archive/openms/OpenMSInstaller/release/2.8.0/OpenMS-2.8.0-src.tar.gz>
compiler errors
Genomedata seems unable to find hdf5; segtools and segway depend on genomedata.
dlopen error upon import
Solver seems to want alignlib-lite for py27, even when building for py310. Recipe currently skipped for py2k, py39, and py36 so maybe not the first time it's been problematic.
source no longer available
misc
Building tries to install 1.1.27 for some reason (despite meta.yaml specifying 1.1.25), but 1.1.27 URL does not exist:
Recipes that need to be migrated to conda-forge
These should be moved to conda-forge rather than be re-built on bioconda as they are legacy packages from back when only bioconda supported perl. As these are not directly related to biomedical research, we want to move them to conda-forge.
Note that perl-net-ssleay is a major dependency (see reverse-dependency list below) so it should be done first.
Work on these after perl packages migrated
These need perl-net-ssleay to be migrated in conda-forge before they can be removed from the blacklist to attempt building.
This needs perl-svg in conda-forge:
Recipes that are conditionally skipped
Rather than completely blacklist a recipe, some were able to be selectively skipped with a preprocessing selector. Edit the preprocessing selector on the
skip: True
line to see if any restrictions can be removed: