Front page | perl.perl5.porters |
Postings from March 2007
The attached patch fixes (or at last papers over) the htmlview.t failure that was recently introduced by change #30584 [1] (or rather, a bug that was unearthed by #30584). The failure was seen at least in Tru64 and HP-UX and only under UTF-8 locales [2], [3]. The failure is (at least in Tru64, I haven't seen the details in HP-UX) that two of the html item anchors don't turn out as expected: # ! <li><strong><a name="mat" class="item">Mat<!></a></strong> # ! <li><strong><a name="mat___" class="item">Mat<!></a></strong> # ! <li><strong><a name="mat2" class="item">Mat</a></strong> # ! <li><strong><a name="mat" class="item">Mat</a></strong> The patch doesn't fix the bug, it just changes the regex so that the bug is not hit. The bug itself requires demerphq :-) I tried looking at whether the [[:punct:]] would be different under locales "C" and "fi_FI.UTF-8", but not that easy -- the [[:punct:]] are identical. (Also note how in the second case '2' gets lost.) NOTE that perl -C is *not* used, and Pod::Html does no utf8 stuff, and the $text in fragment_id_readable() does *not* have UTF8 bit on, but still the locale being a UTF-8 locale changes how things match. Attached are my best attempts at finding what is different, namely the re debug traces from inside the fragement_id_readable() $text substitute statements, one with locale "C" and one with a UTF-8 locale when matching with $text as 'Mat<!>': [1] http://public.activestate.com/cgi-bin/perlbrowse/p/30584 [2] http://www.nntp.perl.org/group/perl.daily-build.reports/2007/03/msg45847.html [3] http://www.nntp.perl.org/group/perl.daily-build.reports/2007/03/msg45852.htmlThread Next