crash with unicode characters in regex comment #8656
Comments
From h.f.s.@web.deThis is a bug report for perl from h.f.s.@web.de, Hi, perl crashes when I use an umlaut in a comment to a regex in a This is the smallest example is was able to create: #!/usr/bin/perl The version of Parse::RecDescent is 1.94 from CPAN. The error message is: The stacktrace is: This was with my system perl version. Continue below... Flags: Site configuration information for perl v5.8.8: Configured by Debian Project at Sun Aug 6 15:47:28 UTC 2006. Summary of my perl5 (revision 5 version 8 subversion 8) configuration: Locally applied patches: @INC for perl v5.8.8: Environment for perl v5.8.8: I locally compiled perl v5.9.4 and was able to reproduce the problem The stacktrace: Program received signal SIGSEGV, Segmentation fault. Regards, Site configuration information for perl 5.9.4: Configured by hermi at Wed Nov 1 18:16:47 CET 2006. Summary of my perl5 (revision 5 version 9 subversion 4) configuration: Locally applied patches: @INC for perl 5.9.4: Environment for perl 5.9.4: |
From guest@guest.guest.xxxxxxxxI investigated a bit further. With both perl 5.8.8 and 5.9.4 I guess the problems are related. The backtrace is: Program received signal SIGABRT, Aborted. |
|
The RT System itself - Status changed from 'new' to 'open' |
From guest@guest.guest.xxxxxxxxWhoops, I must have things seriously mixed up to think that -u had anything to perl -e 'eval "m/ää/;"' Program received signal SIGABRT, Aborted. Please excuse the noise. |
From @andk
> use encoding 'utf8'; There are known bugs in C< use encoding 'utf8'; > which usually go That said, I cannot reproduce your bug with any of 10 different -- Summary of my perl5 (revision 5 version 9 subversion 5) configuration: Characteristics of this binary (from libperl): |
From @demerphqOn 11/2/06, Andreas J. Koenig <andreas.koenig.gmwojprw@franz.ak.mind.de> wrote:
This rings a bell for me, didnt we have an issue related to this Note that backtrack stack, the error is NOT being emitted by the regex cheers, |
From BQW10602@nifty.comOn Thu, 02 Nov 2006 03:45:34 +0100, (Andreas J. Koenig) wrote
The following causes crash on perl 5.8.1 and later #!perl ... and I have often (but not always) encountered such a message: Regards, |
From BQW10602@nifty.comOn Thu, 2 Nov 2006 09:23:53 +0100, demerphq wrote
I think a chunk (see below) at the bottom of S_regatom() 6381: if (ret && PL_encoding && PL_regkind[OP(ret)] == EXACT) { ä encoded in 'utf8' is downgraded by sv_utf8_downgrade() at line 6387, I don't see what this chunk is for. Regards, |
From BQW10602@nifty.comOn Thu, 02 Nov 2006 22:55:23 +0900, SADAHIRO Tomoyuki wrote
I think the below example must clarify the circumstances. #!perl use encoding 'utf8'; use Devel::Peek; While sv_utf8_upgrade() is encoding-aware, sv_utf8_downgrade() isn't Regards, |
From shouldbedomo@mac.comOn 2006–11–02, at 13:57, SADAHIRO Tomoyuki wrote:
That's interesting: the original bug report was using an extended On 2006–11–01, at 21:40, Hermann Schwarting (via RT) wrote:
Firstly, with 32-bit debugging bleadperl@29183 on Mac OS X, Hermann's 32-bit_perl-current$ ./perl -Ilib -I/usr/local/lib/perl5/site_perl/ my $parser = testParser(); sub testParser { test : /\. # Gebäude my $parser = testParser(); sub testParser { test : /\. # Gebäude This may be due to architectural differences between Mac OX on PPC Secondly, making your regex extended stops perl misbehaving for me, 32-bit_perl-current$ ./perl -w -Ilib (gdb) b szone_error Breakpoint 1, 0x90114024 in szone_error () I hope that the gdb output means something to somebody. Please let me |
From BQW10602@nifty.comOn Thu, 2 Nov 2006 18:24:54 +0100, Dominic Dunlop wrote
No, I don't think it is different. Actually Hermann's code I think another bug in Parse::RecDescent is involved in this. #!perl This causes weird messages: Variable "$errortext" is not available at Parse/RecDescent.pm line 2917. The reason is that the invalid pattern in the comment is Parse::RecDescent 1082-1091 '' =~ m$ldel$pattern$rdel" and $@) If Parse::RecDescent::_warn works well, the warning should be Token pattern "m/\. # [c-a] Wrongly the modifiers are neglected in the evalled code. Inline Patchdiff -urN Parse-RecDescent-1.94/lib/Parse/RecDescent.pm Parse-RecDescent-new/lib/Parse/RecDescent.pm
--- Parse-RecDescent-1.94/lib/Parse/RecDescent.pm Wed Apr 09 17:29:37 2003
+++ Parse-RecDescent-new/lib/Parse/RecDescent.pm Fri Nov 03 12:27:33 2006
@@ -1081,9 +1081,9 @@
if (!eval "no strict;
local \$SIG{__WARN__} = sub {0};
- '' =~ m$ldel$pattern$rdel" and $@)
+ '' =~ m$ldel$pattern$rdel$mod" and $@)
{
- Parse::RecDescent::_warn(3, "Token pattern \"m$ldel$pattern$rdel\"
+ Parse::RecDescent::_warn(3, "Token pattern \"m$ldel$pattern$rdel$mod\"
may not be a valid regular expression",
$_[5]);
$@ =~ s/ at \(eval.*/./;
Regards,
SADAHIRO Tomoyuki |
From @andk(TheDamian added to CC)
> On Thu, 2 Nov 2006 18:24:54 +0100, Dominic Dunlop wrote > diff -urN Parse-RecDescent-1.94/lib/Parse/RecDescent.pm Parse-RecDescent-new/lib/Parse/RecDescent.pm I've also put the patch on CPAN as file: $CPAN/authors/id/A/AN/ANDK/patches/Parse-RecDescent-1.94-SADAHIRO-01.patch.gz This is not to steal you the show, just a test bed for the upcoming -- |
From BQW10602@nifty.comOn Thu, 02 Nov 2006 22:55:23 +0900, SADAHIRO Tomoyuki wrote
How to crash perl in this case: 1. When SIZE_ONLY is true, regcomp.c, 6379-6404 When the above chunk is removed, following tests in ext/Encode/t/encoding.t print "not " unless "\x{3AF}" =~ /\xDF/; print "not " unless "\xDF" =~ /\xDF/; Perhaps it is necessary to interpret \xHH in regex as encoded Note: real literals like /Gebäude/ are converted to unicode P.S. though the fix is not yet, a test suite for this report Regards, |
From BQW10602@nifty.comOn Fri, 03 Nov 2006 06:45:04 +0100, Andreas J. Koenig wrote
Thank you. I'm wondering how this module is maintained currently. Regards, |
From BQW10602@nifty.comOn Fri, 03 Nov 2006 16:23:13 +0900, SADAHIRO Tomoyuki wrote
Here is a patch (encoding.patch.gz). - if PL_encoding, regatom() recodes only octal and hexadecimal escapes. Though these changes concern the functionality of encoding.pm Regards, |
From BQW10602@nifty.comOn Sat, 04 Nov 2006 21:02:10 +0900, SADAHIRO Tomoyuki wrote
Then the patch is renewed, encoding2.gz - Test suites in the above patch emit messeges including non-ASCII Therefore the difference is only in messeges from test suites. Regards, |
From @demerphqOn 11/4/06, SADAHIRO Tomoyuki <bqw10602@nifty.com> wrote:
Sadahiro San++ Yves -- |
From @TuxOn Sat, 04 Nov 2006 21:53:50 +0900, SADAHIRO Tomoyuki <bqw10602@nifty.com>
All tests successful (1 subtest UNEXPECTEDLY SUCCEEDED), 39 tests and 340 subtests skipped.
-- |
|
@rgs - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#40641 (status was 'resolved')
Searchable as RT40641$
The text was updated successfully, but these errors were encountered: