[Debian-med-packaging] Bug#1086912: Bug#1086912: phyml: depends on openmpi on 32-bit architectures

[Étienne Mollier]

Hi Emilio,

pochu at debian.org, on 2024-11-07:
> phyml depends on openmpi on 32-bit architectures, which has been
> removed in openmpi 5. Thus this package needs to stop depending on it
> on those architectures.

Thanks for the report, in the case of phyml, there is probably a
bit of work involved.

> There are a couple of options:
> - build-dep on mpi-default-dev instead of openmpi-dev, so that the
>   MPICH implementation can be used on those architectures.

I tried to check whether build-depending on mpi-default-dev for
phyml was a viable option, but it turned out to fail to
configure properly:

	checking for mpi-c... no
	configure: error: Package requirements (mpi-c) were not met:
	
	Package 'mpi-c', required by 'virtual:world', not found
	
	Consider adjusting the PKG_CONFIG_PATH environment variable if you
	installed software in a non-standard prefix.
	
	Alternatively, you may set the environment variables MPI_C_CFLAGS
	and MPI_C_LIBS to avoid the need to call pkg-config.
	See the pkg-config man page for more details.

If someone knows this may be resolved while sticking to mpich,
please shout.  Resolving the issue on 32-bit would require less
paperwork than the alternative.

> - remove the package from 32-bit architectures (armel, armhf and i386)
>   by filing an RM bug against ftp.debian.org

This is potentially the only viable option.  It is worth noting
that a number packages depend on phyml though, mostly arch-all
packages:

  * bioperl-run
  * python-biopython (including binary package python3-biopython
    which is arch-any)
  * prottest
  * jmodeltest

I'm notably concerned that 32-bit support removal could have to
be reflected on packages that depend on python3-biopython: there
is a whole ecosystem of scripts down there for running all sort
of processing for biology purpose.  I would expect a number of
changes to arch-all packages so their newly failing test suites
or build issues or installability issues don't prevent migration
to testing.

If no one thinks of something to resolve the configure issue
with mpich, then I consider starting with removals requests and
adjustments this weekend.

Have a nice day,  :)
-- 
  .''`.  Étienne Mollier <emollier at debian.org>
 : :' :  pgp: 8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
 `. `'   sent from /dev/pts/1, please excuse my verbosity
   `-    on air: t - In Abeyance
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20241107/fccd1fd1/attachment.sig>