Viewing the PDF spec in evince, it got into an infinite loop. The backtrace looked like this:

(gdb) bt
#0  0x0000003822c1b484 in _clip_paths_are_rectilinear (clip=0x7fffb5b6cd40, boxes=0x7fffb5b6cb50, count=0x7fffb5b6cb5c)
    at cairo-clip.c:517
#1  _cairo_clip_path_to_boxes (clip=0x7fffb5b6cd40, boxes=0x7fffb5b6cb50, count=0x7fffb5b6cb5c) at cairo-clip.c:853
#2  _cairo_clip_get_boxes (clip=0x7fffb5b6cd40, boxes=0x7fffb5b6cb50, count=0x7fffb5b6cb5c) at cairo-clip.c:1320
#3  0x0000003822c1cf45 in _cairo_clip_to_boxes (clip=0x7fffb5b6cba0, extents=0x7fffb5b6cae0, boxes=0x7fffb5b6cb50, num_boxes=
    0x7fffb5b6cb5c) at cairo-clip.c:1388
#4  0x0000003822c4b084 in _cairo_surface_fallback_fill (surface=0x184d810, op=CAIRO_OPERATOR_OVER, source=0x7fffb5b6cc20, path=
    0x3822eab158, fill_rule=CAIRO_FILL_RULE_WINDING, tolerance=0.10000000000000001, antialias=CAIRO_ANTIALIAS_DEFAULT, clip=
    0x7fffb5b6cd40) at cairo-surface-fallback.c:1154
#5  0x0000003822c483aa in _cairo_surface_fill (surface=0x184d810, op=CAIRO_OPERATOR_OVER, source=0x7fffb5b6cc20, path=0x3822eab158, 
    fill_rule=CAIRO_FILL_RULE_WINDING, tolerance=0.10000000000000001, antialias=CAIRO_ANTIALIAS_DEFAULT, clip=0x7fffb5b6cd40)
    at cairo-surface.c:2163
#6  0x0000003822c219aa in _cairo_gstate_fill (gstate=0x3822eaae70, path=0x3822eab158) at cairo-gstate.c:1283
#7  0x0000003822c18c05 in cairo_fill_preserve (cr=0x3822eaae40) at cairo.c:2277
#8  0x0000003822c18c41 in cairo_fill (cr=0x3822eaae40) at cairo.c:2253
#9  0x00007f428892d9e0 in ?? () from /usr/lib64/gtk-2.0/2.10.0/engines/libclearlooks.so
#10 0x00007f428891c408 in gdk_rectangle_intersect () from /usr/lib64/gtk-2.0/2.10.0/engines/libclearlooks.so

It doesn't seem to be easily reproducable. 

I don't know the exact version of cairo I was using, but the file looked like this:

-rwxr-xr-x. 1 root root 3556033 Apr 24 09:04 /usr/lib64/libcairo.so.2.10907.0

which suggests master as of 9am EDT April 24th.