[Po4a-Devel] Re: Complex regular subexpression recursion limit (32766) exceeded

Hello, if all you did to change the existing to this form was removing spaces and comments, then yes, that's almost the same regex. You also need to remove the x modifier, at the very end, that says "extended regex", ie, regex where spaces and comments are allowed (and ignored). I think that it's a good idea to have such complex regex to be extended. Comments are more than welcome here. Don't you think? If you did another edit during the transformation, I did not spot it... Mt ----- Le 18 Juil 21, à 18:28, Jean-Christophe Helary lists(a)traduction-libre.org a écrit : > Hello Martin, > > Thank you for the reference... > Well, I'd love to provide a patch but for that I'd need to understand that > regex... :( > > Basically, the regex is this, right ? > > $buffer =~ > /^((?:.*?\n)?(?:[^%]|(?<!\\)(?:\\\\)*\\%)*?)(${RE_ESCAPE}(?:begin|end)\{.*)$/sx > > JC > >> On Jul 19, 2021, at 0:39, Martin Quinson <martin.quinson(a)ens-rennes.fr&gt; wrote: >> >> Hello Jean-Christophe, >> >> this is https://github.com/mquinson/po4a/issues/37 that was "fixed" in the past >> by adding a new line to org.texi. >> I still fail to understand how to fix it properly, and a patch would be more >> than welcome here. >> >> Thanks, Mt >> >> ----- Le 18 Juil 21, à 17:11, Jean-Christophe Helary lists(a)traduction-libre.org >> a écrit : >> >>> When trying to convert org.texi to PO I get the following error: >>> >>> Complex regular subexpression recursion limit (32766) exceeded at >>> /opt/local/lib/perl5/5.26/Locale/Po4a/TeX.pm line 697. >>> >>> That happens between line 11180 and 11200 of the file and it does not seem to be >>> triggered by any weird code. >>> >>> Checking the error on the web, I found that it is possibly related to using * in >>> a regex: >>> >>> https://metacpan.org/pod/XML::Easy::Syntax >>> >>>> BUGS >>>> >>>> Many of these regular expressions are liable to tickle a serious bug in perl's >>>> regexp engine. The bug is that the * and + repeat operators don't always match >>>> an unlimited number of repeats: in some cases they are limited to 32767 >>>> iterations. Whether this bogus limit applies depends on the complexity of the >>>> expression being repeated, whether the string being examined is internally >>>> encoded in UTF-8, and the version of perl. In some cases, but not all, a false >>>> match failure is preceded by a warning "Complex regular subexpression recursion >>>> limit (32766) exceeded". >>>> >>>> This bug is present, in various forms, in all perl versions up to at least 5.8.9 >>>> and 5.10.0. Pre-5.10 perls may also overflow their stack space, in similar >>>> circumstances, if a resource limit is imposed. >>>> >>>> There is no known feasible workaround for this perl bug. The regular expressions >>>> supplied by this module will therefore, unavoidably, fail to accept some >>>> lengthy valid inputs. Where this occurs, though, it is likely that other >>>> regular expressions being applied to the same or related input will also suffer >>>> the same problem. It is pervasive. Do not rely on this module (or perl) to >>>> process long inputs on affected perl versions. >>> >>> Line 697 of TeX.pm is: >>> >>> # detect \begin and \end (if they are not commented) >>> → if ($buffer =~ /^((?:.*?\n)? # $1 is >>> (?:[^%] # either not a % >>> | # or >>> (?<!\\)(?:\\\\)*\\%)*? # a % preceded by an odd nb of \ >>> ) # $2 is a \begin{ with the end of the line >>> (${RE_ESCAPE}(?:begin|end)\{.*)$/sx >>> >>> >>> If there could be a way to simplify this regex, maybe the issue would go away... >>> >>> org.texi is the only file in the emacs distribution that chokes on this regex. >>> >>> -- >>> Jean-Christophe Helary @brandelune >>> https://mac4translators.blogspot.com >>> https://sr.ht/~brandelune/omegat-as-a-book/ >>> _______________________________________________ >>> Devel mailing list -- devel(a)lists.po4a.org >>> To unsubscribe send an email to devel-leave(a)lists.po4a.org >> _______________________________________________ >> Devel mailing list -- devel(a)lists.po4a.org >> To unsubscribe send an email to devel-leave(a)lists.po4a.org > > -- > Jean-Christophe Helary @brandelune > https://mac4translators.blogspot.com > https://sr.ht/~brandelune/omegat-as-a-book/