ports/171353: Segmentation fault when use'ing the perl modules XML::LibXML and GD together (related to pthreads?)

[Slaven Rezic]

>Number:         171353
>Category:       ports
>Synopsis:       Segmentation fault when use'ing the perl modules XML::LibXML and GD together (related to pthreads?)
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    freebsd-ports-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Wed Sep 05 19:10:01 UTC 2012
>Closed-Date:
>Last-Modified:
>Originator:     Slaven Rezic
>Release:        9.0-RELEASE
>Organization:
>Environment:
FreeBSD cvrsnica.herceg.de 9.0-RELEASE FreeBSD 9.0-RELEASE #0: Tue Jan  3 07:46:30 UTC 2012     root at farrell.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC  amd64

CPU: Intel(R) Core(TM) i5-2500T CPU @ 2.30GHz (2307.82-MHz K8-class CPU)

>Description:
The following perl one-liner crashes on my system with a segmentation fault:

perl -e 'use XML::LibXML; use GD;'

All components used here (perl, p5-XML-LibXML, p5-GD) are installed as packages.

Interestingly, if I swap both use statements, then there's no crash:

perl -e 'use GD; use XML::LibXML' 

I could also reproduce the problem with self-compiled perls (5.10.1, 5.14.2, 5.16.0) and the latest module versions from CPAN. Some of this perls did not produce a segfault, but an endless loop in the global destruction phase. Only with a perl with compiled-in threading there was no problem.

The backtrace looks like following:

	#0  0x000000080508935b in pthread_mutex_lock () from /lib/libthr.so.3
	#1  0x0000000803b436c5 in xmlRMutexLock () from /usr/local/lib/libxml2.so.5
	#2  0x0000000803b3eb0c in xmlCatalogCleanup () from /usr/local/lib/libxml2.so.5
	#3  0x0000000803adbbda in xmlCleanupParser () from /usr/local/lib/libxml2.so.5
	#4  0x000000080383eac4 in XS_XML__LibXML_END (cv=0xf343c8) at LibXML.xs:1566
	#5  0x00000000004eab56 in Perl_pp_entersub () at pp_hot.c:2778
	#6  0x000000000043f36d in Perl_call_sv (sv=0xf343c8, flags=13) at perl.c:2705
	#7  0x000000000044444a in Perl_call_list (oldscope=1, paramList=0xf343e0) at perl.c:4782
	#8  0x000000000043a3cf in perl_destruct (my_perl=0x802735220) at perl.c:565
	#9  0x000000000041ddc2 in main (argc=2, argv=0x7fffffffccb0, env=0x7fffffffccc8) at perlmain.c:131

If the problem caused an endless loop (see above), then this loop happens at the same place, somewhere after pthread_mutex_lock.

Regards,
    Slaven

>How-To-Repeat:
See above.
>Fix:
Sorry, none known, other than the workaround of swapping both "use" statements, which is often not feasible in complex systems.


>Release-Note:
>Audit-Trail:
>Unformatted: