On Sun, Feb 10, 2008 at 04:34:42AM -0500, Mike Frysinger wrote: > On Saturday 09 February 2008, Tony Abou-Assaleh wrote: > > We are getting closer to an actual release :O) > > > > I made a new test release grep-2.5.4-20080209b. You can download it from: > > while `make check` passes for me, i'm guessing the failing tests from 2.5.3 > were dropped rather than fixed ? for example, the -o -i brokeness ... > > fails with 2.5.4 but not 2.5.1: > $ echo W | grep -o -i W > <nothing> > -mike
Interesting. After I figured out you probably aren't using UTF-8 and changed LC_ALL to C after the pipe, I can confirm this, except on a version of 2.5.3 with the multibyte patches from debian (that is, their patches 55,60,61,63,64,65,66,67). In en_GB.UTF-8 all versions work fine for me. For what it's worth (in the absence of copyright assignments, probably not a lot), with those patches it isn't appreciably slower in multibyte locales and gives the best results I've seen from the testsuite (42 failures in foad1, 2 in fmbtest, 3 in yesno - instead of 58, 2, 3 with the 2.5.3 release [ all with cs_CZ.UTF-8 ]). Adding the latest versions of foad1 and yesno to that fixes all the failures in yesno, obviously, but adds an extra failure in foad1 for the new -h -H test. Looking at the reult, it was because debian force it to output '(standard input)' and patch the existing foad1.sh accordingly. Ken -- das eine Mal als Tragödie, das andere Mal als Farce
