Bug in PIL's quantize()
Germán M. Bravo <[email protected]> Wed, 20 Jun 2012 17:55:38 -0500
| Newsgroups | gmane.comp.python.image |
|---|---|
| Message-ID | <CA+98ED=j2AeCNsbYZ3k87hOi+6vymbWgFBevMv4X+741_2fs5g@mail.gmail.com> |
I found a nasty bug in PIL's quantize(), in Quant.c,
map_image_pixels_from_median_box(), the `int pixelVal;` definition
must be `unsigned long pixelVal;` instead... This was making PIL crash
under certain circumstances in my system (64 bit Mac OS X Lion 10.7.4
using llvm-gcc4.2).
Patch follows:
-------- CUT HERE --------
--- libImaging/Quant.c.orig 2012-06-20 17:53:49.000000000 -0500
+++ libImaging/Quant.c 2012-06-20 17:53:51.000000000 -0500
@@ -914,7 +914,7 @@
unsigned long bestdist,bestmatch,dist;
unsigned long initialdist;
HashTable h2;
- int pixelVal;
+ unsigned long pixelVal;
h2=hashtable_new(unshifted_pixel_hash,unshifted_pixel_cmp);
for (i=0;i<nPixels;i++) {
-------- CUT HERE --------
_______________________________________________
Image-SIG maillist - [email protected]
http://mail.python.org/mailman/listinfo/image-sig