Hi Étienne, Am Sun, Jan 29, 2023 at 09:19:44PM +0100 schrieb Étienne Mollier: > The uname machine name may not have been armv7l on the test bed, > leading to the present crash. This should read like the below > and I consider pushing a "fix" in that direction tomorrow: > > @pytest.mark.skipif(os.uname().machine[0:3] == 'arm', > reason="fails on armel only, see #1029593.") > > Of course it would be even better to fix the root cause on > armel.
Sounds sensible, thanks for working on this. > Besides, I notice there is a regression in dipy[4], which may be > caused by the nibabel version bump. Since we are in transission > freeze, I suspect the package needs to be reverted to version 4. > That being said the error message in amd64 doesn't make it very > obvious the problem comes from nibabel, so I'm having a closer > look before stating anything definitive. As far as I understood the transition is only for SONAME bumps which is not the case for Python modules (even if I personally would think that bumping versions of widely used modules has features of a transition in terms of potential breakages). In the dipy case I think we are safe since I realised this issue and in turn also upgraded dipy to version 1.6 which seems to build and test nicely. We just should observe its migration closely. Kind regards Andreas. > [2]: > https://ci.debian.net/data/autopkgtest/unstable/armel/n/nibabel/30782893/log.gz > [3]: https://wiki.debian.org/ArchitectureSpecificsMemo > [4]: > https://ci.debian.net/data/autopkgtest/testing/amd64/d/dipy/30816896/log.gz > > Thanks Graham for having kept an eye on this, For sure as always thanks to Graham Andreas. > Sorry for my mess, > -- > Étienne Mollier <emoll...@emlwks999.eu> > Fingerprint: 8f91 b227 c7d6 f2b1 948c 8236 793c f67e 8f0d 11da > Sent from /dev/tty1, please excuse my verbosity. > On air: Lee Abraham - Days Gone By -- http://fam-tille.de