Closed h-vetinari closed 2 years ago
Which is surprising, because there hasn't been any (visible) change in the artefacts.
Also seeing this in https://github.com/conda-forge/pyppmd-feedstock/pull/22
This is being discussed on gitter at: https://gitter.im/conda-forge/conda-forge.github.io?at=62b468f2904f20479a190b5b
The cause is a combination of:
@chrisburr Great, thanks for analyzing and fixing this so quickly!
Confirming as fixed by https://github.com/conda-forge/cross-python-feedstock/pull/44
@chrisburr and others, I'm still seeing this in: https://github.com/conda-forge/keyring-feedstock/pull/88
SafetyError: The package for pypy3.7 located at /home/conda/feedstock_root/build_artifacts/pkg_cache/pypy3.7-7.3.7-h7cff6a8_2
appears to be corrupted. The path 'bin/pypy3'
has an incorrect size.
reported size: 6152 bytes
actual size: 2653 bytes
Am I doing something weird there or is this issue not completely fixed yet?
To answer my own question, it seems like migration to pypy38 got closed without merging on that feedstock and should probably fix the problem.
We had a similar error in https://github.com/conda-forge/hpp-fcl-feedstock/pull/65 as well.
@chrisburr, are you able to apply the fix that you had for pypy 3.8 and 3.9 to pypy 3.7 for feedstocks that have not yet been able to migrate?
PyPy 3.7 isn't supported by conda-forge anymore so I can't easily.
It's probably possible to hack together something to get conda-forge/cross-python-feedstock
to build for PyPy 3.7 again if someone is sufficiently motivated though I'm not going to be able to dedicate time to it. The best solution is probably to remove the PyPy 3.7 migration from feedstocks that haven't been able to accept the PyPy 3.8+3.9 migration.
The best solution is probably to remove the PyPy 3.7 migration from feedstocks that haven't been able to accept the PyPy 3.8+3.9 migration.
Yes, I think that does seem like the right thing to do. @traversaro, is that an option for you all?
Yes, I suggested something similar in the PR, as I do not think there are a lot of pypy37 + aarch/ppc64 users at all.
The following is appearing in https://github.com/conda-forge/numpy-feedstock/pull/272, even after a restart: