Re: [PATCH v4 0/8] PCRE v2, PCRE v1 JIT, log -P & fixes

["Ævar Arnfjörð Bjarmason" <avarab@gmail.com>]

On Fri, Jun 2, 2017 at 6:10 PM, Johannes Schindelin
<Johannes.Schindelin@gmx.de> wrote:
> Hi,
>
> On Fri, 2 Jun 2017, Junio C Hamano wrote:
>
>> Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:
>>
>> > See <20170525200528.22037-1-avarab@gmail.com> for v3
>> > (https://public-inbox.org/git/20170525200528.22037-1-avarab@gmail.com/).
>> >
>> > This is on top of "[PATCH v4 00/31] Easy to review grep & pre-PCRE
>> > changes".
>> >
>> > Changes:
>> >
>> > Ævar Arnfjörð Bjarmason (8):
>> >   grep: don't redundantly compile throwaway patterns under threading
>> >   grep: skip pthreads overhead when using one thread
>> >   log: add -P as a synonym for --perl-regexp
>> >   grep: add support for the PCRE v1 JIT API
>> >   grep: un-break building with PCRE < 8.32
>> >   grep: un-break building with PCRE < 8.20
>> >
>> > No changes.
>> >
>> >   grep: un-break building with PCRE >= 8.32 without --enable-jit
>> >
>> > NEW: It turns out that a PCRE version that supports JIT, but is built
>> > without JIT support will fail at link time since there's no
>> > pcre_jit_exec symbol.
>> >
>> > It also turns out (contrary to what I claimed on list before, my
>> > mistake) that there's no way to detect this through some macro. All
>> > the pcre include files are the same with/without --enable-jit, only
>> > the object file differs.
>> >
>> > So there's now a NO_LIBPCRE1_JIT flag to the Makefile, which is off by
>> > default, but turned on on MinGW. I have not tested that
>> > config.mak.uname change, but everything else I could test on Linux.
>> >
>> > The reason for why it's NO_LIBPCRE1_JIT not USE_LIBPCRE1_JIT is that
>> > in practice pretty much everyone who builds pcre builds it with JIT
>> > (I've looked through various Linux/BSD distro build files), it's MinGW
>> > that's the exception here. Given the performance gain it makes sense
>> > to make it the default.
>> >
>> >   grep: add support for PCRE v2
>> >
>> > Almost no changes, just:
>> >
>> >  * A trivial change to stop redundantly assigning to pcre2_jit_on,
>> >    mistakenly left over from an earlier version.
>> >
>> >  * Updated commit message / perf numbers for the extra patches in the
>> >    series both here and in v3.
>>
>> Nicely summarised and matches what I received; thanks, will replace.
>
> For the record: I spent the entire development time I had today on trying
> to get PCRE2 to build and to figure out which PCRE2 tests fail and why.
>
> I hoped to get to the bottom why the JIT is disabled in PCRE1, but ran out
> of time.
>
> I seem to have gotten PCRE2 to build and figured out why the tests failed
> (spoiler: all of the failures were bogus and no indication of an
> incorrectly-built PCRE2).
>
> I barely had time to build `pu` (forcing PCRE2) and to run the test
> scripts whose file names contain the substring "grep". Seems to work so
> far, but this is by no means comprehensive testing; it is more like hushed
> and rushed testing on a Friday night when I should have stopped working 10
> minutes ago.
>
> Will continue with testing Git for Windows using PCRE2 next week and keep
> you posted,

Thanks a lot for testing it. Great to hear that it (definitely almost) works!

If the grep tests it's very likely that all of them will pass, the
only tests I run when developing this series (outside of the full run
for list submission) are t[0-9]*grep*.sh t[0-9]*log*.sh tests, since
those are the only ones impacted by it.