Regexp failure with utf8-flagged string and byte-flagged pattern

On 9/22/07, Tels <nospam-abusebloodgate.com> wrote:
> Moin,
>
> On Saturday 22 September 2007 14:09:11 Tels wrote:
> > Moin,
> > > I meant to say earlier that the
win32/Makefile has some targets
> > > defined that make it easier to make minor
changes to the regex engine.
> > > If you have a look at the 'reonly' and
'test-reonly' you should be
> > > able to put them in your *nix Makefile and
use them. Then instead of
> > > doing a full test right off you can do a
> > >
> > > make test-reonly
> > >
> > > and make sure that all regex related tests
pass, and only do a full
> > > make test cycle once everything is working
ok. Ive been muttering
> > > about getting these targets added to the
"normal" makefile for a while
> > > but ive not got around to it yet and nobody
else has either.
> >
> > Ah, I might look into this.
>
> Sorry, developing a headache so skipping this for now.

Sorry to hear that. Hope you feel better.

> > > Also, the test file that needs to be updated
for this bug is
> > > t/op/pat.t, and note that the test count is
at the BOTTOM of file, the
> > > new test should go about a page above the
bottom (tests that have
> > > caused SEGV's in the past are kept last).
> >
> > And this, too. After it compiled fully and works,
of course :-P
>
> Attached is a patch that does what Yves suggested,
passes the test from the
> bug report, as well the test I added at t/op/pat.t -
however, I am *not*
> sure that the test I added really tests what it should
- as the file
> t/op/pat.t is in UTF-8 according to file

Hmm, thats a little surprising. Wouldnt have predicted that
at all.

> so I had to use xd6 and I just hope thats ok 

That looks fine to me. So does the patch.

> In any event, that should resolve this issue.

Nice one Tels. Thanks.

Yves
-- 
perl -Mre=debug -e "/just|another|perl|hacker/"

Tels <nospam-abusebloodgate.com> writes:

> Moin,
> 
> On Saturday 22 September 2007 14:09:11 Tels wrote:
> > Moin,
> > > I meant to say earlier that the
win32/Makefile has some targets
> > > defined that make it easier to make minor
changes to the regex engine.
> > > If you have a look at the 'reonly' and
'test-reonly' you should be
> > > able to put them in your *nix Makefile and
use them. Then instead of
> > > doing a full test right off you can do a
> > >
> > > make test-reonly
> > >
> > > and make sure that all regex related tests
pass, and only do a full
> > > make test cycle once everything is working
ok. Ive been muttering
> > > about getting these targets added to the
"normal" makefile for a while
> > > but ive not got around to it yet and nobody
else has either.
> >
> > Ah, I might look into this.
> 
> Sorry, developing a headache so skipping this for now.
> 
> > > Also, the test file that needs to be updated
for this bug is
> > > t/op/pat.t, and note that the test count is
at the BOTTOM of file, the
> > > new test should go about a page above the
bottom (tests that have
> > > caused SEGV's in the past are kept last).
> >
> > And this, too. After it compiled fully and works,
of course :-P
> 
> Attached is a patch that does what Yves suggested,
passes the test from the 
> bug report, as well the test I added at t/op/pat.t -
however, I am *not* 
> sure that the test I added really tests what it should
- as the file 
> t/op/pat.t is in UTF-8 according to file so I had to
use xd6 and I just 
> hope thats ok 
> 

The test just has byte sequences which form valid utf-8.
Most of the
script is really latin1.

And if you want to be sure, Dump() from Devel::Peek is your
friend.

Regards,
        Slaven

-- 
Slaven Rezic - slaven <at> rezic <dot> de

    tksm - Perl/Tk program for searching and replacing in
multiple files
    http://ptktools
.sourceforge.net/#tksm

On 22/09/2007, Tels <nospam-abusebloodgate.com> wrote:
> Attached is a patch that does what Yves suggested,
passes the test from the
> bug report, as well the test I added at t/op/pat.t -
however, I am *not*
> sure that the test I added really tests what it should
- as the file
> t/op/pat.t is in UTF-8 according to file so I had to
use xd6 and I just
> hope thats ok 
>
> In any event, that should resolve this issue.

Thanks, applied as #31961.