On Sat, Apr 09, 2022 at 04:36:26PM +0100, Adam Dinwoodie wrote: > On Fri, Apr 08, 2022 at 12:56:30PM +0200, Ævar Arnfjörð Bjarmason wrote: > > > > <snip> > > > > I converted all the straightforward cases, where these tests were either > > a bit misleading, or we'd actually skip testing on some systems > > needlessly e.g. if they supported \t in a name but not \n. > > > > This leaves only 8 remaining cases of FUNNYNAMES, all of those similarly > > seem like subtle potential issues. I.e. we're creating files with > > characters like "?" or "*" in the name. > > > > But the prerequisite never checks for that, we're just implicitly > > assuming that a FS that can do [\t\n"] an also do [*?+] or whatever. > > > > In the case of the "rm" test we'd unconditionally create a file with a > > space in its name, but then conditional on FUNNYNAMES remove it. > > > > <snip> > > That's a significantly neater patch than I was expecting! I can see > Junio's added my quick fix for the v2.36.0 rc; what's the process from > here in this circumstance? Wait for v2.36.0 to be released properly, > then submit the patches? > > Fixing t9902 using the same scheme as above adds the diff below, after > applying the patch above to the v2.36.0-rc1 tag; I've confirmed this > works as expected on Cygwin, with the test gated by FS_NAME_TAB passing > and the test gated by FS_NAME_BACKSLASH being skipped. > > <snip> Ævar, are you able to submit your patch here to provide the better long-term fix? Once you've submitted your signed-off version, I can submit the t9902-specific changes on top, or I'm happy for you to just integrate them into your commit if that's easier.
