Carlo Marcelo Arenas Belón <carenas@xxxxxxxxx> writes: > On Fri, May 15, 2020 at 05:38:21PM +0200, Alban Gruin wrote: >> >> Thinking about it, I think it would make sense to put this kind of >> information in a diagnostic line, as we already do for the code of a >> failed test. >> >> Instead of this: >> >> > not ok 1 - plain(error:t/./t0001-init.sh:33) >> > # >> > # false && >> > # git init plain && >> > # check_config plain/.git false unset >> > # >> >> Something like this: >> >> > not ok 1 - plain >> > # (error:t/./t0001-init.sh:33) >> > # >> > # false && >> > # git init plain && >> > # check_config plain/.git false unset >> > # > > indeed, and that is why I mentioned this as a workaround only. > > to move to a format like the one you suggest, it might be better to > do it by also moving to TAP13[1] (which allows all that metadata) > > but of course that would also require us to do more changes to the > integration with GitHub's problem matchers and probably a lot more > changes I am not even aware of. At this late stage in the cycle, would it be a safer change to revert the whole thing, I wonder, rather than piling fixes on top of fixes to the initial breakage? 303775a2 (t/test_lib: avoid naked bash arrays in file_lineno, 2020-05-07) 662f9cf1 (tests: when run in Bash, annotate test failures with file name/line number, 2020-04-11) I'll prepare a topic to revert these two directly on top of -rc0 and see how well it goes.
