BBC: 4288c5b broke test in CPAN module XML::Easy #16833
Comments
From @jkeenanAs originally reported in ##### "ExtUtils::Mkbootstrap::Mkbootstrap('blib/arch/auto/XML/Easy/Easy.bs')", I provided a patch (which has not yet been applied). Slaven ##### Change REG_INFTY to 2**16-1, instead of 2**15-1 This commit doubles the upper limit that unbounded regular expression We probably should make this a 32 bit value, but doing this doubling was See http://nntp.perl.org/group/perl.perl5.porters/251413 and followups :040000 040000 5180c35445d2089ef994fcb0634d5e0064197f50 Thank you very much. |
From @jkeenanThis is perl 5, version 29, subversion 7 (v5.29.7 (v5.29.6-69-gc011174cae)) built for x86_64-linux Copyright 1987-2018, Larry Wall Perl may be copied only under the terms of either the Artistic License or the Complete documentation for Perl, including FAQ lists, should be found on |
From @jkeenanOn Sat, 02 Feb 2019 15:18:52 GMT, jkeenan@pobox.com wrote:
Although the test failure reported in this ticket occurred during the 5.29 development cycle, I'm inclined to say that this should *not* be classified as a blocker for perl-5.30 -- at least not until we have other failure reports due to the same commit. I've supplied a patch to the upstream test file. If it's not the best patch, I'm sure the CPAN maintainer will let us know. At this point, however, I think the benefit of the change in blead outweighs the (minor) CPAN breakage. My two cents. Thank you very much. |
|
The RT System itself - Status changed from 'new' to 'open' |
From @khwilliamsonOn 3/8/19 4:34 PM, James E Keenan via RT wrote:
I added all BBC tickets to be 5.30 blockers to ensure adequate |
From @khwilliamsonRemoved from blockers list |
From @iabynOn Wed, Apr 24, 2019 at 10:21:05AM -0700, Karl Williamson via RT wrote:
I concur(*). As as aside, I've just attached an improved patch to the rt.cpan.org (*) "I concur, said William. I agree completely. If you think it's for the -- |
That patch has not yet been applied, nor has a new CPAN release been issued. @zefram can you take a look? Thank you very much. |
--- t/syntax_main.t- 2019-04-24 11:47:08.365496801 +0100
+++ t/syntax_main.t 2019-04-24 12:01:50.459442769 +0100
@@ -35,16 +35,34 @@
# This code checks whether the regexp iteration limit bug (#60034) is
# present. The regexp match expression checks for getting the wrong
-# result with a long input, and suffices to diagnose the bug. However,
-# running that test on a pre-5.10 perl causes the stack to grow large,
+# result with a long input, and suffices to diagnose the bug.
+# for a pattern like /X*/, where X is sub-pattern that can match variable
+# length string, e.g. (ab?), it is currently known that:
+#
+# on < 5.10.0, the old recursive engine will crash on too long a match;
+# on < 5.29.4, /X*/ is misinterpreted as /X{0,32767}/
+# on 5.29.4+, /X*/ is misinterpreted as /X{0,65535}/
+#
+# Running that test on a pre-5.10 perl causes the stack to grow large,
# and if there's a limited stack size then this may overflow it and
# cause perl to crash. All pre-5.10 perls have the iteration limit
# bug, so there's no need to run the proper test on those verions.
# 5.10 fixed the stack issue, so it's safe to run the proper test there.
-my $have_iterlimit_bug = "$]" < 5.010 || do {
- local $SIG{__WARN__} = sub { };
- ("a"x40000) !~ /\A(?:X?[a-z])*\z/;
-};
+
+my $iterlimit; # if defined, sets an upper limit for iterations
+
+if ($] < 5.010) {
+ $iterlimit = 0;
+}
+else {
+ local $SIG{__WARN__} = sub { };
+ for my $i (32767, 65535) {
+ if (("a"x($i+1)) !~ /\A(?:X?[a-z])*\z/) {
+ $iterlimit = $i;
+ last;
+ }
+ }
+}
my $data_in = IO::File->new("t/read.data", "r") or die;
my $line = $data_in->getline;
@@ -75,7 +93,7 @@
}
SKIP: {
skip "perl bug affects long inputs", 2
- if $have_iterlimit_bug && length($input) >= 32766;
+ if defined $iterlimit && length($input) >= $iterlimit;
is upgraded($input) =~ $recogniser{$prod}, !$syntax_error;
is downgraded($input) =~ $recogniser{$prod}, !$syntax_error;
} |
|
If the only thing remaining is a new CPAN release and this isn't a blocker any more, what's the outstanding on this? |
toddr
added
the
Closable?
|
Given the blocker was removed, 5.30 is released, and there is a report to the module in question, I am closing this ticket. |
toddr
removed
the
Closable?
Migrated from rt.perl.org#133816 (status was 'open')
Searchable as RT133816$
The text was updated successfully, but these errors were encountered: