Bug#1089841: python-escript: Drop line with true in debian/rules

Santiago Vila sanvila at debian.org
Fri Dec 13 17:34:54 GMT 2024


Package: src:python-escript
Version: 5.6-10
Severity: important

Hello.

Before uploading this release today I noticed that the "|| true"
at the end of this debian/rules line:

(scons $(SFLAGS) cc_optim='$(OPT)'  build_dir=$(BUILD3) verbose=on prefix=$(WORK3) options_file=debian/sid_options.py docs ) || true

was necessary to avoid this weird build error:

*** The MPI_Comm_size() function was called before MPI_INIT was invoked.
*** This is disallowed by the MPI standard.
*** Your MPI job will now abort.
[test-1:780231] Local abort before MPI_INIT completed completed successfully, but am not able to agg
regate error messages, and not able to guarantee that all other processes were killed!
scons: *** [debian/tmp3/posix/doc/sphinx_api/stuff] Error 14
scons: building terminated because of errors.


After I did the upload for 5.6-10 (and also after the mirror pulse sync
at 14:00 UTC today), I am currently unable to reproduce that,
but maybe it was random.

At the same time, in the chroot I had to test the build, apt-get said
today:

The following packages were automatically installed and are no longer required:
   libhdf5-103-1t64 libhdf5-cpp-103-1t64 libhdf5-fortran-102t64 libhdf5-hl-100t64
   libhdf5-hl-cpp-100t64 libhdf5-hl-fortran-100t64 libhdf5-openmpi-103-1t64

Well, I can see those packages in the build logs of previous failed build logs
for python-escript, so I wonder if the reason for the weird build failure
was that those build-dependencies (now gone) were buggy.

(btw: Release 5.6-10 dropped no longer required BD on libnetcdf-dev, but
this does not seem to be related).


To summarize: I'd like to see the "|| true" removed, but I'm not confident enough
to assure that it's no longer needed.

If you (Alastair) are preparing release 6.0, and you have never seen the error
in such version, maybe we could drop the "|| true" with release 6.0.

Thanks.



More information about the debian-science-maintainers mailing list