Fix: Automatically finalize t8code objects when Trixi.jl shuts down

[jmark]

This PR provides a solution for automatically finalizing all active t8code related objects before Trixi.jl resp. MPI shuts down. Before this PR t8code objects had to be finalized explicitly by hand before shutting down Trixi or nasty segfaults occurred.

This PR is related this discussion: DLR-AMR/t8code#1295.

This PR depends on DLR-AMR/T8code.jl#75 and DLR-AMR/T8code.jl#76.

[github-actions]

Review checklist

This checklist is meant to assist creators of PRs (to let them know what reviewers will typically look for) and reviewers (to guide them in a structured review process). Items do not need to be checked explicitly for a PR to be eligible for merging.

Purpose and scope

• The PR has a single goal that is clear from the PR title and/or description.
• All code changes represent a single set of modifications that logically belong together.
• No more than 500 lines of code are changed or there is no obvious way to split the PR into multiple PRs.

Code quality

• The code can be understood easily.
• Newly introduced names for variables etc. are self-descriptive and consistent with existing naming conventions.
• There are no redundancies that can be removed by simple modularization/refactoring.
• There are no leftover debug statements or commented code sections.
• The code adheres to our conventions and style guide, and to the Julia guidelines.

Documentation

• New functions and types are documented with a docstring or top-level comment.
• Relevant publications are referenced in docstrings (see example for formatting).
• Inline comments are used to document longer or unusual code sections.
• Comments describe intent ("why?") and not just functionality ("what?").
• If the PR introduces a significant change or new feature, it is documented in NEWS.md with its PR number.

Testing

• The PR passes all tests.
• New or modified lines of code are covered by tests.
• New or modified tests run in less then 10 seconds.

Performance

• There are no type instabilities or memory allocations in performance-critical parts.
• If the PR intent is to improve performance, before/after time measurements are posted in the PR.

Verification

• The correctness of the code was verified using appropriate tests.
• If new equations/methods are added, a convergence test has been run and the results
  are posted in the PR.

Created with ❤️ by the Trixi.jl community.

[andrewwinters5000]

Great that you determined an "under-the-hood" way to handle this finalization process!

[codecov]

Codecov Report

Attention: Patch coverage is 91.30435% with 2 lines in your changes missing coverage. Please review.

Project coverage is 96.39%. Comparing base (2edf9cd) to head (5458fb8).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/auxiliary/t8code.jl	50.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2172      +/-   ##
==========================================
- Coverage   96.39%   96.39%   -0.00%     
==========================================
  Files         483      483              
  Lines       38325    38333       +8     
==========================================
+ Hits        36941    36948       +7     
- Misses       1384     1385       +1     

Flag	Coverage Δ
unittests	96.39% <91.30%> (-<0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jmark]

Although this fix works nicely in serial Julia sessions it is flawed for MPI-based parallel execution. First, the instances of the garbage collector on each Julia process seem to work independently in terms of when and what to free. This leads to stalling MPI processes waiting for other processes who are not finalizing the t8code object at the moment. A similar effect can happen for the last finalization call while shutting down since the Set data structure most probably does not retain the order of insertion. Th latter is crucial to have for multiple MPI processes so that the objects sharing the same MPI arrays are freed.

Phew ...

[jmark]

@andrewwinters5000 @benegee @sloede @ranocha @JoshuaLampert Ready for review.

[JoshuaLampert]

Only a minor thing, but I'm wondering if we should add something like Base.pointer(fw::ForestWrapper) = fw.pointer in T8code.jl to avoid explicitly accessing the field here. (Sorry I missed that in the review in T8code.jl.)
Otherwise this looks good to me.

[jmark]

Only a minor thing, but I'm wondering if we should add something like Base.pointer(fw::ForestWrapper) = fw.pointer in T8code.jl to avoid explicitly accessing the field here. (Sorry I missed that in the review in T8code.jl.) Otherwise this looks good to me.

@JoshuaLampert I already thought about doing this similar to PointerWrapper in P4est.jl. But it didn't work right away and I opted for the stupid way of writing it out explicitly.

I then found out that you can actually mess with unsafe_convert and dispatch on it. So I did: DLR-AMR/T8code.jl#76

[jmark]

Thanks a lot! The segfault https://github.com/trixi-framework/Trixi.jl/actions/runs/12052335616/job/33605413710?pr=2172#step:7:1119 happened in two consecutive CI runs, so it looks like it is a real issue.

It seems so. Pretty tricky to narrow down the cause for that. On my machine the forest wrapper test runs just fine.

[ranocha]

Do you run them in interactive mode locally?

[jmark]

Do you run them in interactive mode locally?

Both! Tested in interactive session and in "script" mode. Works as expected on my local machine.

[JoshuaLampert]

Thanks!

[jmark]

All green, besides CodeCov. Is there anything else to do?

[sloede]

Good to go from my side, but I'd like to have @ranocha sign off as well.

Thanks for all your effort @jmark!

[ranocha]

Thanks!