conda-forge
Handling packages with invalid `MatchSpec` (for their `dependencies` and `constrains`)
Context
Some packages published on conda-forge like infowillow-1.6.3-pyhd8ed1ab_0.conda use invalid MatchSpec.
This is for instance the case of one of its dependencies which is encoded as:
pillow-heif >=0.10.0,<1.0.0<py312
This MatchSpec is invalid, since the <py312 suffix is not supported by the MatchSpec specification.
Questions
-
Is this a known problem?
-
How should those packages be handled? Must they simply be removed?
-
Might some of such specifications stem from grayskull converting PEP 508 dependency specification customs from PyPI packages?
It doesn't seem to be related to grayskull. I could not find that package either but I did find a reference to it in https://github.com/conda-forge/willow-feedstock/pull/27. It was a human mistake and one should open an issue there. BTW, this repo is not the best place this this report.
I think it is a general question for conda-forge about whether it is ok to remove packages that have non-sensical specs, that could break the parsing of the channel data.
FWIW, this shouldn't be a problem in the future provided conda-build 24.7+ is in use (https://github.com/conda/conda-build/pull/5359).
I think it is a general question for conda-forge about whether it is ok to remove packages that have non-sensical specs, that could break the parsing of the channel data.
This happens all the time. We discussed it in element channel but I'll post here for posterity. The steps to fix it are:
- Fix the feedstock, also open the issue there so the maintainers know of this issue.
- Send a repodatapatch PR to fix the current package
BTW, this repo is not the best place this this report.
Sorry for that. Where shall I report issues in the future?
Sorry for that
Not a problem.
Where shall I report issues in the future?
For feedstock problems? In the feedstock. For general problems we use https://github.com/conda-forge/conda-forge.github.io.