| 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) |
