Shourya Shukla <shouryashukla.oo@xxxxxxxxx> writes: >> I think a test that relies on platform-specific error string is a >> bug. It's like expecting an exact string out of strerror(), which >> we had to fix a few times. > >> So I am not sure we would want to butcher compat/mingw.c only to >> match such an expectation by a (buggy) test. > > Alright. That is understandable. What alternative do you suggest? Should > we change the check in the test? A buggy check should of course be changed. It should be sufficient to ensure "git submodule summary" fails, regardless of what exact error message it issues, no? If the command does not exit with non-zero exit status, when it gives a "fatal" error message, that may indicate another bug, though.
