Change 16091682 by jiho@jiho-earthsea-medley-review-fix_segfault-git5 on 2010/06/16 01:13:54

	Fix buffer allocation problem of hunspell, expecially for Korean Affix which
	doesn't use UTF-8 as its base encoding.
	
	PRESUBMIT=passed
	BUG=2661104
	R=jayr
	DELTA=40  (9 added, 25 deleted, 6 changed)
	OCL=16077686

Affected files ...

... //depot//hunspell_1_2_8/README.google#2 edit
... //depot//hunspell_1_2_8/src/hunspell/atypes.hxx#2 edit
... //depot//hunspell_1_2_8/src/hunspell/hunspell.cxx#3 edit

==== //depot//src/hunspell/atypes.hxx#1 - /google/src/files/16091682/depot//hunspell_1_2_8/src/hunspell/atypes.hxx ====
--- /google/src/files/13856229/depot//src/hunspell/atypes.hxx	2009-12-08 16:54:13.000000000 -0500
+++ /google/src/files/16091682/depot//src/hunspell/atypes.hxx	2010-06-16 04:13:54.000000000 -0400
@@ -19,7 +19,13 @@
 #define SETSIZE         256
 #define CONTSIZE        65536
 #define MAXWORDLEN      100
-#define MAXWORDUTF8LEN  256
+// Note(jiho@google.com): Korean Hunspell dictionary doesn't use UTF-8 directly.
+// It decomposes one single Korean character to three characters. So, one single
+// UTF-8 Korean character(3 bytes) will eventually take 9 bytes. At least
+// MAXWORDLEN * 3 bytes are required for handling Korean dictionary.
+// And the Korean dictionary has a word with 513 byte length. I've changed this
+// value to 550 to cover the word. If not, hunspell dies with segfault.
+#define MAXWORDUTF8LEN 550
 
 // affentry options
 #define aeXPRODUCT      (1 << 0)