Hi all,

I have an odd issue that I wonder if anyone's seen before (before I dig into the guts of SVN::* myself...)

If I let SVN::Client automatically handle the connection pool itself, I get a segfault:

use strict;
use warnings;

use SVN::Client;

my $ctx = new SVN::Client;
$ctx->cat(\*STDOUT, "https://some/svn/url", "HEAD");
[download]

The backtrace of the core is:

(gdb) backtrace
#0  0xfe330f20 in ?? ()
#1  0xfeec41dc in ?? ()
#2  0xfeec42a8 in ?? ()
#3  0xfeec3f4c in ?? ()
#4  0xff098fdc in ?? ()
#5  0x00086e34 in xmlShell ()
#6  0x0007e9a8 in Perl_runops_standard ()
#7  0x00029084 in _PROCEDURE_LINKAGE_TABLE_ ()
#8  0x00028d7c in _GLOBAL_OFFSET_TABLE_ ()
#9  0x0002cd88 in S_call_list_body ()
#10 0x0002c95c in xmlCopyEntity ()
#11 0x00025d48 in perl_destruct ()
#12 0x00024cfc in main ()
[download]

So it occurs in the destruction of some xs code - presumably the SVN connection. On a hunch, I tried creating and freeing the pool manually:

use strict;
use warnings;

use SVN::Core;
use SVN::Client;

my $pool = SVN::Pool->new_default;
my $ctx = new SVN::Client;
$ctx->cat(\*STDOUT, "https://some/svn/url", "HEAD", $pool);
undef $pool;
[download]

Which works a treat. Anyone have any familiarity with the SVN::* internals before I go hunting?

Update: I loaded in all the symbol files I could think were relevant and discovered in the backtrace is xml related (and I'm linked against libxml2).