Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> writes: > A re-roll of my improvements my series to simplify the CI setup a lot > (see diffstat), much of it was dealing with constraints that went away > with Travis et al. CI for this series (OSX runners failing for > unrelated reasons): > > https://github.com/avar/git/actions/runs/2040223909 > > For a much more detailed summary of how the output looks before/after > see v1[]. > > This series heavily conflicts with Johannes's > js/ci-github-workflow-markup in "seen", but in the v1 I suggested > basing that series on top of this one, because it can benefit a lot > from these simplifications. > > I'll reply to this series with a proposed rebasing of that series on > top of this one, which allows for removing almost all of its changes > to "ci/" with no harm to its end-goals, i.e. the splitting up of > "make" and "make test" output is something it'll get for free from > this series. A sample run for this can be seen at https://github.com/git/git/runs/5715128999?check_suite_focus=true With the output, I can point at a specific line, e.g. https://github.com/git/git/runs/5715128999?check_suite_focus=true#step:7:1150 The URLs with Dscho's one that correspond to the above two are https://github.com/git/git/runs/5699946885?check_suite_focus=true https://github.com/git/git/runs/5699946885?check_suite_focus=true#step:4:1826 The specific breakage (which has little to do with the comparison between CI error pages) is that diff_setup() is called and populate diff_options.parseopts member but diff_setup_done() does not seem to be called and ends up leaking it. I wonder why options->parseopts is not freed immediately after diff_opt_parse() calls parse_options(), but need to be kept until diff_setup_done().
