pil: fix crash in 64-bit mode

Closes #15563.

Signed-off-by: Adam Vandenberg <flangy@gmail.com>

1 files changed, 21 insertions, 2 deletions

diff --git a/Library/Formula/pil.rb b/Library/Formula/pil.rb
index 323e83ff5..e9865b1e7 100644
--- a/Library/Formula/pil.rb
+++ b/Library/Formula/pil.rb
@@ -1,15 +1,21 @@
 require 'formula'
 
 class Pil < Formula
-  url 'http://effbot.org/downloads/Imaging-1.1.7.tar.gz'
   homepage 'http://www.pythonware.com/products/pil/'
+  url 'http://effbot.org/downloads/Imaging-1.1.7.tar.gz'
   sha1 '76c37504251171fda8da8e63ecb8bc42a69a5c81'
 
   option 'with-little-cms', 'Compile with little-cms support.'
 
   depends_on :freetype
   depends_on 'jpeg' => :recommended
-  depends_on 'little-cms' unless build.include? 'with-little-cms' # => :optional
+  depends_on 'little-cms'=> :optional unless build.include? 'with-little-cms'
+
+  # The patch is to fix a core dump in Bug in PIL's quantize() with 64 bit architectures.
+  # http://mail.python.org/pipermail/image-sig/2012-June/007047.html
+  def patches
+    DATA
+  end
 
   def install
     # Find the arch for the Python we are building against.
@@ -51,3 +57,16 @@ class Pil < Formula
     EOS
   end
 end
+
+__END__
+--- a/libImaging/Quant.c
++++ b/libImaging/Quant.c
+@@ -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++) {