On Tue, Jun 07 2022, Junio C Hamano wrote: > Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> writes: > >> On Fri, Jun 03 2022, Junio C Hamano wrote: >> >>> Indeed it makes it impossible to figure it out things like this >>> case. But ... >>> >>>> But this does look easy to "solve" with a quicker fix, just bringing >>>> back the "ci/print-test-failures.sh" step so you can at least expand it, >>>> and not have to go to the "summary" and download the *.zip of the log >>>> itself. As that shows we still have the raw log there, it just didn't >>>> make it to the new GitHub Markdown formatting mechanism. >>> >>> ... it seems a solution is possible? Care to send in a patch (or >>> perhaps Dscho already has a counter-proposal)? >> >> The only thing I have at the moment is: >> >> 1. git revert -m 1 bd37e9e41f5 >> 2. merge: https://lore.kernel.org/git/cover-v6-00.29-00000000000-20220525T094123Z-avarab@xxxxxxxxx/ >> 3. merge: https://lore.kernel.org/git/cover-v6-00.14-00000000000-20220525T100743Z-avarab@xxxxxxxxx/ >> >> I.e. to pick this in the sequence I'd proposed doing & have tested >> thoroughly. > > I know you two have difference in opinions, but throwing away > everything the other party did and forcing your stuff in is not a > very effective way to work together. I'm suggesting getting Johannes's patches in combined with the changes & bugfixes I'd proposed. So no, not throwing that work away, it would (applied up to 14/14) give you functionally the same end result that's on "next" now as far as the new GitHub Markdown output is concerned. The [3] above has links to the relevant CI output. I had tried to rebase the above [2] on top of "next" before this discussion started, I agree that would be ideal, but it's a much larger logical change that I don't have time to pursue now. I.e. there's a reason I proposed doing it in that order, a logical rebasing of [2] on top of bd37e9e41f5 would involve a lot of backing out of the existing direction taken there. I.e. the whole part where the split by "steps" provides much of the ci/* specific code in bd37e9e41f5 for free. >> It also addresses other noted some other regressions in "next", but as >> noted e.g. in [A] there's other issues in "next", e.g. that even the >> "raw" trace logs are altered as a side-effect of running with >> --github-workflow-markup, and of course the major UX slowdowns. > > Dscho? I know you do not care about the UX slowdown as much as > others, but I am sure you do not consider what is in 'next' is > perfect. It seems to need further work to go back to the feature > parity with what it replaced. Just to be clear [3] up to 14/14 would still exhibit this particular bug, but with 13/14 it wouldn't from the links in [3] the relevant outputs are: "next" (well, similar): https://github.com/avar/git/runs/6571972194?check_suite_focus=true [3] with 14/14: https://github.com/avar/git/runs/6588407676?check_suite_focus=true [3] with 13/14: https://github.com/avar/git/runs/6588579493?check_suite_focus=true I really would like to get is out of this long-running ci/ limbo, perhaps Johannes has some proposed patches, but I don't think fixing the outstanding bugs is going to be trivial or easy. Some of it's hard to tease apart, e.g. the altered *.out logs seem to require some tricky test-lib.sh and test-lib-functions.sh changes. I don't see why wouldn't get all of that code in now though, just hidden behind a flag, that would take the pressure off dealing with the current regressions, [2] with 13/14 would do that. Then once those outstanding issues are fixed we'd just need the one-line 14/14 change to flip the default CI output. But is it the smallest possible change on top of what's now in "next"? No, of course not. But I don't have those hypothetical patches, just the above. That's all I meant in reply to the "care to send in a patch?" upthread.
