On 14-Feb-2014 11:47, mathog wrote:
I'm thinking it will probably come down to building a debug version of
fontconfig, linking with that, and then tracking a whole lot of
pointers.
Did that with 2.11. The good news: all of the leaks associated with
text_reassemble.c in Inkscape went away. Whatever that problem was, it
was eliminated in the intervening versions. The bad news: Inkscape
still can't run FcFini() just before exit without crashing.
Ran inkscape in valgrind while it was linked with the 2.11 library
compiled with -g, and traced a huge number of memory leaks into
FcValueSave. These 3 lines leak when that function is called from
numerous locations:
v.u.s = FcStrdup (v.u.s);
v.u.m = FcMatrixCopy (v.u.m);
v.u.l = FcLangSetCopy (v.u.l);
but this one doesn't
v.u.c = FcCharSetCopy ((FcCharSet *) v.u.c);
What is this code trying to do? Because in general
sometype var;
/* set the values in var */
var = duplicate_function(var);
is going to leak if sometype has pointers to memory. The surprising
thing here is not that the 3 lines noted leak, but that FcCharSetCopy
doesn't. FcValueSave is called from 4 locations in fcpat.c, and all of
them leaked. The only other place FcValueSave seemed to be used was in
fccfg.c, but that included fcaliastail.h, which redefines FcValueSave so
that the function by that name is not called.
Regards,
David Mathog
mathog@xxxxxxxxxxx
Manager, Sequence Analysis Facility, Biology Division, Caltech
_______________________________________________
Fontconfig mailing list
Fontconfig@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/fontconfig
