trixi-framework / trixi-framework/Trixi.jl
Investigate why MPI tests on Windows fail and fix them
- Dominant language
- Julia
- Stars
- 731
- Forks
- 167
- Avg merge
- 2d 18h
- Merged PRs (30d)
- 25
Description
In #1398 we disable a number of tests for MPI on Windows. These were failing recurrently but intermittently when executed on the GitHub Actions runners. We also couldn't find an underlying cause for the failures, but the intermittent recurrence suggests a non-deterministic issue such as out-of-memory errors or actual out-of-bounds errors. Since this caused virtually every PR's CI test to fail, it was not really useful, and thus in #1398 the offending tests were turned off.
However, this means that at the moment we do not test the `P4estMesh` at all in a parallel environment on Windows. Since the underlying cause is still unknown, it would be good to investigate this as soon as possible to at least enable some `P4estMesh` testing on Windows again...
Contributor guide
Research direction
Review the MPI tests disabled in #1398 and the GitHub Actions Windows configuration, focusing on the P4estMesh parallel test path. Reproduce the intermittent failure on a Windows runner and determine whether it is an out-of-memory or out-of-bounds issue. Done means the cause is addressed and at least some P4estMesh MPI testing is reliably enabled on Windows.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- julia
- Domain
- ci-cd, distributed-systems, operating-systems, testing
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 30/100