Hi Dscho, On Wed, Oct 30, 2019 at 3:17 PM Johannes Schindelin <Johannes.Schindelin@xxxxxx> wrote: > > Hi Elijah, > > On Wed, 30 Oct 2019, Elijah Newren wrote: > > > On Mon, Oct 28, 2019 at 6:20 PM Junio C Hamano <gitster@xxxxxxxxx> wrote: > > > > > > Johannes Schindelin <Johannes.Schindelin@xxxxxx> writes: > > > > > > >> Still, I rely pretty heavily on t6036, t6042, t6043, and t6046 for > > > >> sanity in the face of refactoring and rewriting -- and as mentioned > > > >> before they have caught refactoring bugs in those areas that appear at > > > >> first blush as "overzealous", ... > > > > > > > > One idea would be to try to guard those extra careful tests behind the > > > > `EXPENSIVE` prereq. > > > > > > Yeah, I like that---I think it is perfectly in line with the spirit > > > of EXPENSIVE, too. > > > > Or perhaps EXPENSIVE_ON_WINDOWS, since it's actually pretty cheap on > > linux and not that bad on Mac > > Why the complexity? If you separate out the expensive tests (even if > they are only expensive in terms of run time on Windows), it will make > the regression tests so much more readable to the occasional reader > (making them less expensive in terms of reading time...). The "extra careful" things you were complaining about with the new test I was adding to t6043 was true of every single test in that file...and likely much of t6036, t6042, and perhaps even t6046 (though those have fewer tests than t6043). I have no clue where I'd even begin to draw the line between them. If it's possible, it sounds extremely complex. Just using the EXPENSIVE_ON_WINDOWS prereq that already exists would be easy and simple. Or did you mean you wanted me to duplicate every single test and attempt to trim down the duplicates somehow? That'd be a rather large undertaking that sounds rather unappealing on a few fronts, but maybe that's what you had in mind? > > However, if we're going down that route, perhaps t9001-send-email.sh > > could be wrapped in an EXPENSIVE prerequisite? That single test file > > takes an inordinate percentage of overall runtime. One one box with a > > few extra cpus, that test both starts first and finishes last...and > > it's not far from that on even normal boxes. > > I would be okay with that. > > No, let me try that again. I would be _totally_ okay with that. Ooh, sweet, sounds like I should propose it.
