Update GAP_pkg_juliainterface_jll

[fingolfin]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 74.53%. Comparing base (0f83ffd) to head (bd3cd82).
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1139      +/-   ##
==========================================
- Coverage   82.74%   74.53%   -8.21%     
==========================================
  Files          50       55       +5     
  Lines        3824     4716     +892     
==========================================
+ Hits         3164     3515     +351     
- Misses        660     1201     +541     

see 9 files with indirect coverage changes

[lgoettgens]

The ctbllib CI job has a segfault, but that should be unrelated to this change.

[fingolfin]

@lgoettgens what confuses me are the failures in Julia nightly which again report worldage failures. I wonder if that means something went wrong compiling the JLL that caused it to no pick up JULIA_VERSION_MINOR and JULIA_VERSION_MAJOR correctly... huh

.

[lgoettgens]

Indeed, the scheduled CI from this morning (https://github.com/oscar-system/GAP.jl/actions/runs/13105678001/job/36560073838) was completely green, with only the world-age warnings from Documenter. So in some sense, this change makes it worse than before.

[fingolfin]

I can reproduce this locally so it is not something weird in the CI.

But I also used a disassembler to look at the JuliaInterface.so in the JLL -- and thus verified that its Func_JuliaGetGlobalVariableByModule really does contain the new code which updates the world age.

However comparing it with a locally built JuliaInterface.so it appears the offset to world age is different. Conclusion: libjulia_jll needs to be updated with a recent Julia master snapshot, then we can rebuild GAP_pkg_juliainterface_jll and hopefully it'll work as desired.

[lgoettgens]

julia nightly now refers to 1.13-DEV, for which we don't have a libjulia yet (x-ref JuliaPackaging/Yggdrasil#10429). After #1141 is merged, we can at least see CI running on 1.12-DEV again.

[lgoettgens]

The 1.12 tests only show warnings about docmeta, i.e. JuliaDocs/Documenter.jl#2621.

I am looking into the 1.13 failure, this probably has to do with Aqua.jl.

[lgoettgens]

This looks good now.
Two things to note:

• The name of the 1.11 CI job seems to have changed. @fingolfin could you please update the list of required CI jobs to reflect that?
• There are still some errors in nightly everytime we try to access singular (ref https://github.com/oscar-system/GAP.jl/actions/runs/13178758178/job/36784187046?pr=1139#step:11:314). This will probably fix itself once libsingular_julia_jll gets rebuild with julia 1.13 support.