Issues tab and upstream lhapdf/gettext

[nathanielerowe]

Based on my discussion here it may be beneficial to use the upstream versions of lhapdf and gettext in order to fix some issues with genie. I also do not see an "Issues" tab available in this repository, so I am posting this under discussions instead. If "Discussions" is not the right place to put this, let me know and I can move it.

[greenc-FNAL]

I don't believe there's anything in this fork of Spack that would prevent one using the upstream versions of lhapdf and gettext. In which outside repositories are the non-official ones located?

[nathanielerowe]

There are different recipes for lhapdf here (builtin) and in fnal_art. It was suggested to me that it would be preferable to remove the one in fnal_art so that it wasn't necessary to maintain both. That part is currently in a PR in that repo and requires no changes here.

The gettext in this repo seems to diverge from mainline spack here. This seemed to cause some issues with building lhapdf as a dependency of genie, as changing to the mainline spack build of gettext fixed my issue.

Because of this, I thought it might be a good idea to use the mainline spack version of gettext. I may have misunderstood and thought that lhapdf itself may not have been being maintained to match the mainline spack, but it looks like the package.py files for these match. If updates to mainline spack are for lhapdf are being monitored, I believe it can be excluded from the discussion. Apologies for any confusion around that point.

My issue is then seemingly fixed by a difference in gettext package.py files in the FNALssi fork and the mainline spack

The error associated with this was

make[1]: Entering directory '/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/wrappers'
Making all in python
make[2]: Entering directory '/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/wrappers/python'
cython ./lhapdf.pyx --cplus -2 -o lhapdf.cpp
for i in lhapdf.cpp lhapdf.pyx clhapdf.pxd __init__.py; do \
  orig="$i"; \
  build="$(basename $i)"; \
  if [ ! -e ${build} ]; then cp ${orig} ${build}; \
  echo ${build} >> fix-out-of-source; fi; \
done
touch fix-out-of-source
CC="/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/lib/spack/env/gcc/g++" CXX="/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/lib/spack/env/gcc/g++" CFLAGS="-O3" CXXFLAGS="-O3" CPPFLAGS="-I/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/include -I." LHAPDF_LOCAL=1 /lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/python-venv-1.0-d6kjr2thd5iybzcu3jjk6nfvrip5yrrs/bin/python build.py
/usr/lib64/gcc/x86_64-suse-linux/12/../../../../x86_64-suse-linux/bin/ld: cannot find -lintl: No such file or directory
collect2: error: ld returned 1 exit status
Build command = /lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/lib/spack/env/gcc/g++  -shared -fPIC  -o  lhapdf.so  /var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/wrappers/python/lhapdf.cpp  -I/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/include -I/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/include -I/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/include  -I/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/include   -O3   -L/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/src/.libs  -L/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/lib  -lstdc++ -lLHAPDF  -I/lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/python-3.11.9-yajs77zqrlbvpsyqicvhgxme2ymdz6vj/include/python3.11 /lus/grand/projects/neutrinoGPU/software/NateDev/fermi-spack-dev/spack/opt/spack/linux-sles15-zen3/gcc-12.3.0/python-3.11.9-yajs77zqrlbvpsyqicvhgxme2ymdz6vj/lib/libpython3.11.so -lintl -lpthread -ldl  -lutil -lm
make[2]: *** [Makefile:499: all-local] Error 1
make[2]: Leaving directory '/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/wrappers/python'
make[1]: *** [Makefile:383: all-recursive] Error 1
make[1]: Leaving directory '/var/tmp/pbs.2064452.polaris-pbs-01.hsn.cm.polaris.alcf.anl.gov/nathanielerowe/spack-stage/spack-stage-lhapdf-6.5.4-w4was2bjecpv3kld7sd2gfgub3b5dcjf/spack-src/wrappers'
make: *** [Makefile:520: all-recursive] Error 1

There is nothing preventing the replacement of the gettext package.py with the upstream one locally, I was suggesting possibly replacing the gettext package.py with the upstream one in this repo. Sorry if that point wasn't clear, hope this makes the issue linked a bit easier to follow.

[greenc-FNAL]

That makes sense, and explains why I couldn't find gettext in a non-Spack recipe repository. Sorry for the misunderstanding.

With which branch of FNALssi/spack were you having the gettext problem?

Re the issues tab: the original motivation for omitting the tab was to avoid people filing issues that would be better filed against upstream Spack, but your case is specifically one where having an issues tab for the fork would be a good idea. We will discuss the matter within the FNAL Spack team to understand the best way forward here, and thank you for raising this.

[nathanielerowe]

I was working on the fnal-develop branch. Another thing about this issue is that I was running into it on both the sbndbuild machines (AL9) and ALCF's polaris, I noticed that there seem to be some OS dependent if statements causing issues, which may have been causing the problem, but I am not super sure.

[nathanielerowe]

Another update: using mainline gettext seemed to also fix similar issues I was having with libxpm and krb5.

[nathanielerowe]

@greenc-FNAL Hi, just wanted to follow up on this. Is there a plan to modify gettext to match the upstream version, or are there reasons to allow them to diverge?

[greenc-FNAL]

Prior to removing the unwanted non-official recipes, one could specify builtin.gettext and/or builtin.lhapdf as appropriate in your specs/environments to ensure that the concretizer chooses the desired versions.