Windows integration tests aborting prematurely due to CTRL-C
As seen in !4647 (comment 1787709942), it appears that TestBuildGitFetchStrategyFallback
does not always run on the Windows integration tests for some reason. This has led to master:broken because TestBuildGitFetchStrategyFallback
was failing with the Windows cmd
shell.
@ajwalker enabled the -debug
flag on splitic
in !4080, and this shows that all the Windows integration jobs are missing some go test
lines, especially on node indexes 2-4.
Only node index 1 for windows 21H2
has an invocation for package gitlab.com/gitlab-org/gitlab-runner/executors/shell
; the rest show no signs at all.
All jobs show the same number of tests, so this makes me wonder if the run grouping is somehow not working right on windows 21H2
?
57 tests for index 1/4, ~11.40 seconds.
56 tests for index 2/4, ~11.20 seconds.
56 tests for index 3/4, ~11.20 seconds.
56 tests for index 4/4, ~11.20 seconds.
I haven't been able to reproduce the issue on a Windows Server 2022 Datacenter node with Go 1.21.7, though.
I added more debugging in !4654 (closed), but ironically in my first pipeline to see the results https://gitlab.com/gitlab-org/gitlab-runner/-/jobs/6260643804 ran TestBuildGitFetchStrategyFallback
just fine.