Dear Maintainer,

The autopkgtest of your package are failing with netcdf-fortran (4.6.0+ds-1):

 Command   /usr/bin/abinit /tmp/tmp.TWJnew3xcm/fast_t00/t00.abi    > /tmp/tmp.TWJnew3xcm/fast_t00/t00.stdout   2> /tmp/tmp.TWJnew3xcm/fast_t00/t00.stderr
  returned exit_code: 139

 [31m[fast][t00][np=1][run_etime: 0.64 s]: fldiff fatal error:
 file 1 has more significant lines than file 2 (762 > 0).
  [file=t00.abo][0m
 [31m[fast][t00][np=1] Test was not expected to fail but subprocesses returned retcode: 139[0m
 [31m
 Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

 Backtrace for this error:
 #0  0x7f93f79c88c2 in ???
 #1  0x7f93f79c7a55 in ???
 #2  0x7f93f763d93f in ???
 #3  0x7f93f7cb7518 in __netcdf_nc_interfaces_MOD_addcnullchar
 #4  0x7f93f7ca53b0 in nf_def_var_
 #5  0x7f93f7d19b8e in __netcdf_MOD_nf90_def_var_onedim
 #6  0x5601b05e77bf in ???
 #7  0x5601b02d9cd0 in ???
 #8  0x5601b02db88f in ???
 #9  0x5601aff6496f in ???
 #10  0x5601aff638fa in ???
 #11  0x5601af275ffd in ???
 #12  0x5601af27388e in ???
 #13  0x7f93f762881c in __libc_start_main
 #14  0x5601af2738c9 in ???
 #15  0xffffffffffffffff in ???
 Segmentation fault
 [0m
 No YAML Error found in: [fast][t00][np=1]

https://ci.debian.net/data/autopkgtest/testing/amd64/a/abinit/24182934/log.gz

The testsuite does not fail when run as part of the package build.

Kind Regards,

Bas

Installing the dbgsym packages makes the backtrace a little more
informative:

  Backtrace for this error:
  #0  0x7f889dd4b93f in ???
          at ./signal/../sysdeps/unix/sysv/linux/x86_64/sigaction.c:0
  #1  0x7f889e3c5518 in __netcdf_nc_interfaces_MOD_addcnullchar
          at ./fortran/module_netcdf_nc_interfaces.F90:2587
  #2  0x7f889e3b33b0 in nf_def_var_
          at ./fortran/nf_genvar.F90:63
  #3  0x7f889e427b8e in __netcdf_MOD_nf90_def_var_onedim
          at ./fortran/netcdf4_variables.F90:61
  #4  0x559ee1f317bf in __m_nctk_MOD_write_var_netcdf
          at ./shared/common/src/27_toolbox_oop/m_nctk.F90:2941
  #5  0x559ee1c23cd0 in __m_parser_MOD_prttagm
          at ./src/42_parser/m_parser.F90:3538
  #6  0x559ee1c2588f in __m_parser_MOD_prttagm_images
          at ./src/42_parser/m_parser.F90:3668
  #7  0x559ee18ae96f in __m_outvar_a_h_MOD_outvar_a_h
          at ./src/57_iovars/m_outvar_a_h.F90:191
  #8  0x559ee18ad8fa in __m_outvars_MOD_outvars
          at ./src/57_iovars/m_outvars.F90:371
  #9  0x559ee0bbfffd in abinit
          at ./src/98_main/abinit.F90:323
  #10  0x559ee0bbd88e in main
          at ./src/98_main/abinit.F90:88

Rebuilding with netcdf-fortan (4.6.0+ds-1) resolves the issue.

Kind Regards,

Bas

Control: reassign -1 src:netcdf-fortran 4.6.0+ds-1
Control: affects -1 src:abinit

Usually that's a strong indication about ABI breakage in a library.

I know nothing about Fortran ABI rules, but nf90_def_var_oneDim()
gaining 3 (optional) parameters in 4.6.0 is suspiciously close
to #3 in the backtrace.

Please check with upstream whether a soname bump might be necessary
for libnetcdff.

cu
Adrian

reassign -1 src:abinit 9.6.2-1
 >

No other netcdf-fortran rdeps have shown issues which leads me to
suspect it is an issue in abinit.

I only ended up taking care of netcdf-fortran because that was split of
from netcdf which is a dependency of gdal. I don't actually use any of
the netcdf packages, so I'm not willing to spend much time on them.

Kind Regards,

Bas

Hi Bas

You wrote:

Only abinit has a comprehensive autopkgtest.  The other reverse
dependencies, cdftools, etsf-io, ferret-vis, ncl and oasis3 have no
autopkgtests, and pyferret only has a superficial 'import pyferret'
test.

You also wrote:

If no change in abinit was needed, why do you suspect an issue in abinit?

In your request for a binNMU in #1016496, you were asked if libnetcdff
shouldn't have bumped its SONAME.  Someone needs to talk to upstream
to get this unblocked.

Regards
Graham

Control: tags -1 wontfix

Because only that package has an issue. I don't know any fortran, so I
cannot troubleshoot this issue.

I'm not going to be that person. I was hoping that the abinit maintainer
would chime in, but they haven't responded to this issue for almost two
months.

Since no one cares enough about netcdf-fortran, I'll go an have it
removed from the archive as I'm not willing to spend time on issues like
these.

Kind Regards,

Bas

Hi Bas

I'm sure you are aware, but we have a process [1] for passing packages
on to other maintainers.

I suggest speaking to Alastair (now added as a recipient).  He
maintains most of netcdf-fortran's reverse-dependencies.

Regards
Graham


[1] https://wiki.debian.org/Orphaning

Hi Alastair

Thank you for adopting netcdf-fortran!

There has still not been a SONAME bump in libnetcdff7.

Please check that the packages that depend on this library (listed
below) still work.  Unfortunately, only abinit has an autopkgtest, and
it is no longer in testing (for other reasons).

Regards
Graham


Reverse-Build-Depends
=====================
* abinit                        (for libnetcdff-dev)
* cdftools                      (for libnetcdff-dev)
* etsf-io                       (for libnetcdff-dev)
* ferret-vis                    (for libnetcdff-dev)
* ncl                           (for libnetcdff-dev)
* oasis3                        (for libnetcdff-dev)
* pyferret                      (for libnetcdff-dev)