[Python-modules-team] Bug#770977: Bug#770977: Bug#770977: pytest: Fails to build from source

[Simon Chopin]

Quoting Lennart Sorensen (2014-11-26 16:53:44)
> On Tue, Nov 25, 2014 at 04:35:52PM -0500,  wrote:
> > Hmm, interesting, I can try and see if I can reproduce it on more systems.
> > 
> > So the machine I was building on is an IBM x3650 with 2 quad core 2.5GHz
> > xeons.  Kernel is amd64, userspace is i386 wheezy, chroot is i386 jessie
> > buildd variant.
> > 
> > So I just tried some more tests.
> > 
> > jessie-amd64 on sid-amd64: works
> > jessie-i386 on sid-amd64: works
> > jessie-i386 on wheezy-i386(with amd64 kernel): fails
> > Same chroot copied to a sid machine (so 3.16 kernel instead of 3.2 kernel: works
> > 
> > It would appear that pytest fails on python 3.4.2 when the underlying
> > kernel is 3.2 for some reason that I can't even imagine what is.
> > 
> > I will try upgrading the kernel on our build machine to the
> > wheezy-backprot of 3.16 and see if that solves the problem and report
> > back.
> 
> So to confirm I did this:
> 
> Create a VM with a fresh wheezy install.
> Create a jessie chroot with debootstrap.
> Try building pytest.
> It fails.
> Upgrade wheezy kernel to 3.16 from wheezy-backports.
> Try building pytest in chroot again.
> It passes.
> 
> So the problem is caused by running on the 3.2 wheezy kernel.
> 
> So weird as that is, I think the bug can be closed.  No idea if this
> is a bug in the old kernel, or glibc, or python 3.4, with one of them
> making some assumption about kernel behaviour based on the linux-libc-dev
> headers they were built against.  I suppose it means there is a chance a
> system upgrading from wheezy to jessie could have some issues with running
> python3 stuff in some cases until it is rebooted to the jessie kernel,
> although that would probably be recommended to do as soon as possible
> anyhow.

I think we should downgrade it and maybe tag it moreinfo, but it
shouldn't be closed. The case of a Jessie kernel on a machine running a
kernel that is not Jessie's is a valid use case IMHO.

Not that I'm volunteering to track down and fix it!

Cheers,
Simon
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: signature
URL: <http://lists.alioth.debian.org/pipermail/python-modules-team/attachments/20141126/4dd8e57d/attachment.sig>