Repository navigation
Various AIX tests pass on Node 22, fail on later supported versions #61657
Description
Activity
A note on
mainon test-osuosl-aix72-ppc64_be_4:tools/test.py --repeat 2000 async-hooks/test-emit-after-on-destroyedhad two failures out of 2000, both near the end of the run. Both occurred in this line. The test code is the same on v22.x and main. and the failure is that it gets aSIGILLon that line.For wpt/test-streams it gave an error 1 time out of 200 -
CRASHED (Signal: 5)sayingCheck failed: ptr == address.- changed the title
[-]Various AIX tests pass on Node 22, fail on later versions[/-][+]Various AIX tests pass on Node 22, fail on later supported versions[/+]on Feb 4, 2026 It looks like
wpt/test-streamsstarted failing between v24.2.0 and v24.3.0 (although with a low failure rate that's not guaranteed to be true!)
Still tracking down test-emit-after-on-destroyed but it looks it might have been ok in v24.4.0 but fails in v24.8.0 (Will update this comment as more diagnoses occurs)Test with a few commits in the v24 stream between v24.5.1 and v24.6.0
commit comment emit failures test-streams 8686b8037a6 src: remove unused DSAKeyExportJob 0/50000 0/200 eecd3272a62 worker: add name for worker 8/50000 2/200 3c88b769bb9 lib: restructure assert to become a class 2780/50000 0/200 f834a6be591 deps: update undici to 7.13.0 2728/50000 1/200 38aedfbf736 crypto: support ML-DSA KeyObject, sign, and verify Similar to last row Canceled main doc: fix grammatical error in README.md 40/50000 0/200 main exc.worker v26 without worker name change 53/50000 1/200 main exc.assert v26 without assert restruction 45/50000 1/200 test-emit-after-destroyedseemed to get particularly bad at 3c88b76lib: restructure assert to become a classand then started being 'mostly ok' (flaky) at commit a6a31cb467edoc: compress Web Cryptography Algorithm matrixin v24.7.0](3c88b769bb9) where it seems to have been been low level flaky since.This issue has been marked as stale due to 90 days of inactivity.
It will be automatically closed in 30 days if no further activity occurs. If this is still relevant, please leave a comment or update it to keep it open.Still valid
A sibling issue to #61441
After clearing up some of the flakiness and machine-specific problems that were occurring on AIX it became noticable that there were some failures that were NOT failing when run against Node 22.
This issue is to continue the analysis and determine if the failures are from changes in the test cases or functional changes in the Node that caused the intermittent failures in later releases.
The
RUN_TESTSparameter of the stress test that I was using to experiment with issequential/test-watch-mode-restart-esm-loading-error wpt/test-streams async-hooks/test-emit-after-on-destroyed node-api/test_threadsafe_function parallel/test-repl-preview- Thetest-emit-after-destroyedone is a reasonably common flake on AIX and has been marked as such.wpt/test-streamsis the other one that shows a slightly lower failure rate (around 1%) on the later versions.Sample runs showing the differences - each of these ran 100 times on each of four AIX machines: