On Wed, Nov 27, 2019 at 08:07:30AM -0500, Ed Maste wrote:
> On Wed, 27 Nov 2019 at 11:24, Jeff King <peff@xxxxxxxx> wrote:
> >
> > > Extend test-lib.sh to add a FREEBSD prereq (akin to MINGW) and add !FREEBSD
> > > to these tests.
> >
> > Before we start growing more system-specific prereqs here, can we add a
> > layer of indirection? Convert this test to use a REGEX_ALLOW_ILLSEQ
> > prereq (or maybe there's a better name), and then set it for both mingw
> > and freebsd?
>
> Thanks, and that makes sense, but should we make the sense the
> opposite way - set REGEX_ALLOW_ILLSEQ only for glibc (including
> Cygwin, I guess)?
Possibly. There's a long tail of platforms we support, and we don't
really know how they behave (are they OK, or is it that nobody has
bothered to check yet?). I'd be inclined to blacklist platforms where we
know it doesn't work, in order to let the test run on as many as
possible.
I guess we'd need to invert the prereq then to systems that actually
report ILLSEQ (maybe REGEX_ILLSEQ, and set it for FreeBSD?).
> This also applies only to two cases ("latin1 + locale" and "latin1 +
> locale + invalid needle"). There are other !MINGW tests in
> t4210-log-i18n.sh which do not use invalid UTF-8 and work fine on
> FreeBSD.
Hmm, yeah. It's not clear to me if !MINGW is there because of the ILLSEQ
issue, or just because it doesn't handle these tests at all. So maybe it
would make sense to just set it for FreeBSD for now, and mingw folks can
clean it up further if they want to dig into it.
-Peff
