Skip Menu |

This queue is for tickets about the XML-LibXML CPAN distribution.

Report information
The Basics
Id: 117423
Status: new
Priority: 0/
Queue: XML-LibXML
People
Owner: Nobody in particular
Requestors: jkahrman [...] mathworks.com
Cc:
AdminCc:
Bug Information
Severity: (no value)
Broken in: (no value)
Fixed in: (no value)
Attachments
get_set_namespace
XML-LibXML-2.0126-patch

History Show all quoted text
  Thu Sep 01 21:13:36 2016 jkahrman [...] mathworks.com - Ticket created
Subject: 'uninitialized value' warning when calling setNamespace passing undef value returned from namespaceURI
I'm in the process of updating our perl build (perl 5.8.8 => 5.20; XML::LibXML 1.66 => XML-LibXML-2.0126). Using the newer versions, the attached code 'get_set_namespace' produces a new 'uninitialized value' warning. After some debugging and doing a little research, I think the problem is here: XML-LibXML/perl-libxml-mm.c 1204c1204 < if ( scalar != NULL && scalar != &PL_sv_undef ) { --- Show quoted text
> if ( scalar != NULL && SvOK(scalar) ) {
Basically, in the previous versions, the 'undef' being passed in happens to be the same 'undef' as &PL_sv_undef and is handled without a 'uninitialized value' warning. That appears to no longer be the case. From http://perldoc.perl.org/5.20.2/perlguts.html: "So to repeat always use SvOK() to check whether an sv is defined." Note: I think I've properly connected the dots here, but I've just started poking around in the Perl XS stuff, so I could be way off base.
Subject: get_set_namespace
Download get_set_namespace
application/octet-stream 561b

Message body not shown because it is not plain text.

  Fri Sep 16 12:20:18 2016 jkahrman [...] mathworks.com - Correspondence added
From: jkahrman [...] mathworks.com
After further testing, I ran into another 'uninitialized value' warning due to another comparison against &PL_sv_undef instead of using the SvOK macro. Rather than to continue to play wack-a-mole with these warnings, I went through and replaced all comparisons of variables to &PL_sv_undef with the SvOK macro. (See attached combined patch). All the default tests passed, but a sporadic issue I had been experiencing with the thread tests started occurring regularly on Linux and Mac (On Windows, the thread tests consistently stall before and after the patches). # XXX: Linux deadlocks/stalls after reference count warning/segfault?: PmmREFCNT_dec: REFCNT decremented below 0 for 13165c0! at t/90shared_clone_failed_rt_91800.t line 45. *** glibc detected *** perl: corrupted double-linked list: 0x0000000000c1e720 *** # XXX: Sporadic failures on maci PmmREFCNT_dec: REFCNT decremented below 0 for 7f9cd4d001b0! at t/90threads.t line 137. perl5.20.2(71235,0x7fff7c512300) malloc: *** error for object 0x7f9cd4d001b0: pointer being freed was not allocated *** set a breakpoint in malloc_error_break to debug I think this is the issue reported here: https://rt.cpan.org/Public/Bug/Display.html?id=91800 The change to use SvOK in "void DESTROY( node )" in "XML::LibXML::Node" in LibXML.xs appears to be the one that perturbs things (I suspect the comparison against &PL_sv_undef is masking the underlying issue), but I couldn't narrow down the issue any further. I've just abandoned running the thread tests for now as I don't believe we are using this feature.
Subject: XML-LibXML-2.0126-patch
Download XML-LibXML-2.0126-patch
application/octet-stream 4.6k

Message body not shown because it is not plain text.