Jon Turney wrote:
On 08/07/2025 12:06, Christian Franke wrote:
Jon Turney wrote:
[...]
It seems that the 'filerace' test (new?) doesn't work reliably in
the CI environment.
This (new!) test never failed during many runs I did locally before
tagging it as WORKS. There are also occasional failures of 'flock'
and 'fork' at GH.
Today I could reproduce one hang of filerace when the number of cores
is closer to the VM behind GH actions.
$ cygstress -r 100 -c 16,18,20,22 filerace flock fork
...
>>> FAILURE: 11:58:32.68: filerace (exit status 0, command hangs,
processes left, files left in '/tmp/stress-ng.410.141.d')
...
>>> SUMMARY:
>>> FAILURE: filerace: 1 of 100 test(s) failed
>>> SUCCESS: flock: all 100 test(s) succeeded
>>> SUCCESS: fork: all 100 test(s) succeeded
Would it be possible for you to take a look?
Yes.
Thank you.
Should I push a new script version which excludes this test for now?
I think that would be a good idea - keeping it green so we stand a
chance to notice any other problems
Done:
https://cygwin.com/git/?p=newlib-cygwin.git;a=commit;h=ace88c7ba17a
https://github.com/cygwin/cygwin/actions/runs/16170259198
The 'filerace' test actually succeeds but may take much longer than the
requested 5 seconds.
The also disabled 'flock' and 'fork' tests need further investigation...
--
Regards,
Christian
