Bug #122958 for XML-LibXML: libxml2-2.9.5 breaks t/48_reader_undef_warning_on_empty_str

Tue Sep 05 02:40:51 2017 dmacks [...] netspace.org - Ticket created

CC:
Subject:	libxml2-2.9.5 breaks t/48_reader_undef_warning_on_empty_str_rt106830.t
Date:	Tue, 05 Sep 2017 02:19:23 -0400
To:	bug-XML-LibXML [...] rt.cpan.org
From:	Daniel Macks <dmacks [...] netspace.org>

The new version of libxml2 produces a test failure on my OS X 10.11 (Apple's perl-5.18.2). XML-LibXML-2.0129 reports: #   Failed test 'xml_reader is undef' #   at t/48_reader_undef_warning_on_empty_str_rt106830.t line 33. # Looks like you failed 1 test of 2. t/48_reader_undef_warning_on_empty_str_rt106830.t .. Dubious, test returned 1 (wstat 256, 0x100) Failed 1/2 subtests I think the cause is a libxml2 change in the creation of the parser for a zero-length string. Previously, it short-circuited and treated this case as if a NULL were passed, returning NULL. Now, a parser is created as for any other actual string. The relevant upstream change is probably: https://github.com/GNOME/libxml2/commit/94f6ce838c6a903ef5d0251f061a3af1816b4c50 dan   --- Daniel Macks dmacks@netspace.org

Thu Sep 07 10:20:55 2017 paul [...] city-fan.org - Correspondence added

From:

paul [...] city-fan.org

On Tue Sep 05 02:40:51 2017, dmacks@netspace.org wrote: Show quoted text

> The new version of libxml2 produces a test failure on my OS X 10.11 > (Apple's perl-5.18.2). XML-LibXML-2.0129 reports: > > #   Failed test 'xml_reader is undef' > #   at t/48_reader_undef_warning_on_empty_str_rt106830.t > line 33. > # Looks like you failed 1 test of 2. > t/48_reader_undef_warning_on_empty_str_rt106830.t .. > Dubious, test returned 1 (wstat 256, 0x100) > Failed 1/2 subtests > > I think the cause is a libxml2 change in the creation of the parser > for a zero-length string. Previously, it short-circuited and treated > this case as if a NULL were passed, returning NULL. Now, a parser is > created as for any other actual string. The relevant upstream change > is probably: > > https://github.com/GNOME/libxml2/commit/94f6ce838c6a903ef5d0251f061a3af1816b4c50

I'm also seeing a failure in t/32xpc_variables.t: # Failed test ' TODO : Add test name' # at t/32xpc_variables.t line 114. # Looks like you failed 1 test of 35. t/32xpc_variables.t ................................ Dubious, test returned 1 (wstat 256, 0x100) Failed 1/35 subtests Cause is probably similar. This is on Fedora Linux (same on Rawhide and ancient Fedora 13).

Thu Sep 07 10:20:56 2017 The RT System itself - Status changed from 'new' to 'open'

Thu Sep 07 12:22:01 2017 ppisar [...] redhat.com - Correspondence added

From:

ppisar [...] redhat.com

Dne Út 05.zář.2017 02:40:51, dmacks@netspace.org napsal(a): Show quoted text

> I think the cause is a libxml2 change in the creation of the parser > for a zero-length string. Previously, it short-circuited and treated > this case as if a NULL were passed, returning NULL. Now, a parser is > created as for any other actual string. The relevant upstream change > is probably: > > https://github.com/GNOME/libxml2/commit/94f6ce838c6a903ef5d0251f061a3af1816b4c50 >

I confirm both failures are triggered by this libxml2 commit.

Thu Sep 07 12:46:18 2017 ppisar [...] redhat.com - Correspondence added

From:

ppisar [...] redhat.com

Dne Čt 07.zář.2017 12:22:01, ppisar napsal(a): Show quoted text

> Dne Út 05.zář.2017 02:40:51, dmacks@netspace.org napsal(a):

> > I think the cause is a libxml2 change in the creation of the parser > > for a zero-length string. Previously, it short-circuited and treated > > this case as if a NULL were passed, returning NULL. Now, a parser is > > created as for any other actual string. The relevant upstream change > > is probably: > > > > https://github.com/GNOME/libxml2/commit/94f6ce838c6a903ef5d0251f061a3af1816b4c50 > >

> I confirm both failures are triggered by this libxml2 commit.

This is not true, I made some mistake. The t/48_reader_undef_warning_on_empty_str_rt106830.t is of course caused by the mentioned 94f6ce838c6a903ef5d0251f061a3af1816b4c50 commit. But the t/32xpc_variables.t failure (no exception after unregistering XPath variables) is caused by commit <https://github.com/GNOME/libxml2/commit/c851970c6ec2166df804b1d551df0998de182fc4>: commit c851970c6ec2166df804b1d551df0998de182fc4 (HEAD, refs/bisect/bad) Author: Nick Wellnhofer <wellnhofer@aevum.de> Date: Sat May 27 15:26:11 2017 +0200 Rework final handling of XPath results Move cleanup of XPath stack to xmlXPathFreeParserContext. This avoids memory leaks if valuePop fails in some error cases. Found with libFuzzer and ASan. Rework handling of the final XPath result object in xmlXPathCompiledEvalInternal and xmlXPathEval to avoid useless error messages.

Wed Sep 20 09:25:51 2017 ppisar [...] redhat.com - Correspondence added

From:

ppisar [...] redhat.com

Dne Čt 07.zář.2017 12:46:18, ppisar napsal(a): Show quoted text

> But the t/32xpc_variables.t failure (no exception after unregistering > XPath variables) is caused by commit > <https://github.com/GNOME/libxml2/commit/c851970c6ec2166df804b1d551df0998de182fc4>: >

I reported it to libxml2 authors <https://bugzilla.gnome.org/show_bug.cgi?id=787941> because I thing this is a regression. The other test failure is just an improvement that XML-LibXML should adapt to, e.g. like this: --- a/t/48_reader_undef_warning_on_empty_str_rt106830.t +++ b/t/48_reader_undef_warning_on_empty_str_rt106830.t @@ -30,9 +30,13 @@ use XML::LibXML::Reader; my $xml_reader = XML::LibXML::Reader->new(string => $empty_xml_doc); # TEST - ok (scalar(!defined($xml_reader)), - 'xml_reader is undef', - ); + SKIP: { + skip 'libxml2 accepts empty strings since 2.9.5 version', 1 + if (XML::LibXML::LIBXML_VERSION() >= 20905); + ok (scalar(!defined($xml_reader)), + 'xml_reader is undef', + ); + } # TEST eq_or_diff(

Wed Sep 20 10:27:55 2017 NWELLNHOF [...] cpan.org - Correspondence added

On Wed Sep 20 09:25:51 2017, ppisar wrote: Show quoted text

> Dne Čt 07.zář.2017 12:46:18, ppisar napsal(a):

> > But the t/32xpc_variables.t failure (no exception after unregistering > > XPath variables) is caused by commit > > <https://github.com/GNOME/libxml2/commit/c851970c6ec2166df804b1d551df0998de182fc4>: > >

> > I reported it to libxml2 authors > <https://bugzilla.gnome.org/show_bug.cgi?id=787941> because I thing > this is a regression.

Fixed in libxml2 with https://git.gnome.org/browse/libxml2/commit/?id=3157cf4 Show quoted text

> The other test failure is just an improvement that XML-LibXML should > adapt to, e.g. like this:

Right.

Mon Oct 09 10:56:44 2017 mjgardner [...] cpan.org - Correspondence added

Message body is not shown because it is too large.

Mon Oct 09 10:56:46 2017 mjgardner [...] cpan.org - Correspondence added

Message body is not shown because it is too large.

Bug #122958 for XML-LibXML: libxml2-2.9.5 breaks t/48_reader_undef_warning_on_empty_str_rt106830.t