On Tuesday 20 November 2007, Tony Abou-Assaleh wrote: > > src/search.c:check_multibyte_string() calls wcrtomb() on a const string > > but wcrtomb() in most cases will be modifying the first argument (which > > is the const string) > > That code wasn't touched in at least 2 years. Is it easy to fix? Would you > submit a patch? :O)
no idea, ive never hacked on grep before ... it's scary. a cursory look shows getting rid of the const marking isnt really feasible, which means either use a different function, or dupe the string ... the latter is trivial to do, but the former requires widechar knowledge which i lack, sorry > > > 2) Remove known failing cases from the release > > > > can you iterate which those are ? i see 2 failing tests: foad1.sh and > > yesno.sh ... seems like the foad1.sh tests show a pretty "bad" problem in > > grep that should get fixed rather than ignored ... or is my understanding > > of what is being tested incorrect ? seems like failing "-o -i" is > > pretty "bad" ... especially as it's a regression from grep-2.5.1 ... > > Yes, foad1.sh and yesno.sh. But not all the tests in these script fail. It > is easier to just remove these files from the distribution tarball, but I > think if there is an easy enough way to disable only some tests that would > be better. > > The failing cases have been failing for a long time. while true, the last grep version people have been using is 2.5.1 which was released a long time ago and does not have this bug: echo word | grep -o -i WORD > > rather than rewriting yet another testsuite by hand, have you seen > > autotest ? it's an attempt by the autotools guys for streamlining test > > integration for autotool-based projects > > Sounds promising! It is currently listed as experimental, but we should > keep autotest on the radar. ive found it to be pretty usable in automake-1.10/autoconf-2.61 (other than one bug which has been fixed when running tests in $srcdir!=$builddir) if you'd like some help with grep integration, i dont mind -mike
signature.asc
Description: This is a digitally signed message part.
