On Tue, 21 Apr 2026 09:01:17 +0200 "Pieter Lenaerts" <[email protected]> wrote:
> On Mon Apr 20, 2026 at 2:29 PM CEST, Jeroen Ploemen wrote:
> > did a review of python-papermill, up for sponsorship in the Python
> > Team:
>
> Hi Jeroen,
>
> Thanks for this detailed review. You may have seen that Bastian has
> uploaded python-papermill in parallel.
>
> I will implement your comments and make them part of a next version.
>
> Some specifics:
>
> > * control: the restriction on pyarrow only deals with i386,
> > presumably because the salsa CI doesn't try any archs other than
> > amd64 and i386. The Debian CI on the other hand runs on more
> > archs; does the current setup survive that, or is this package
> > going to run into trouble elsewhere as well? If so, you might
> > want to switch the current [!i386] to a list of all archs where
> > pyarrow is available.
>
> I should have looked at the apache-arrow debian/control file. [1]
>
> apache-arrow restricts to 64-bit LE and that means I should exclude
> not only i386.
Yeah I noticed bage's upload; apparently my review was overlooked. For
now, the package doesn't build at all because something way down the
dependency chain of pyarrow is missing, but once it does I strongly
suspect it will fail the CI on s390x (which is big endian and thus
doesn't have python3-pyarrow) and not migrate as a result.
In the meantime, please go ahead with the necessary improvements as
this package is likely going to need another upload anyway.
For the pyarrow test patch, you can use something along the lines of:
HAVE_PYARROW = True
try:
import pyarrow
except ImportError:
HAVE_PYARROW = False
pass
...to check for the presence of pyarrow, then use that boolean in
place of the current check for a specific architecture. That way,
the only place you need to maintain a list of architectures is on
the build-dependency in d/control.
pgpociIXwgRUz.pgp
Description: OpenPGP digital signature
