Bump dictionary versions to 3-0

chrome/tools/convert_dict/dic_reader.cc

Index: chrome/tools/convert_dict/dic_reader.cc
diff --git a/chrome/tools/convert_dict/dic_reader.cc b/chrome/tools/convert_dict/dic_reader.cc
index 967f07e5d4480e957cbeec10bde3f609b3380d78..bef23823bce534b2c3910eaac81c44b52f4009cb 100644
--- a/chrome/tools/convert_dict/dic_reader.cc
+++ b/chrome/tools/convert_dict/dic_reader.cc
@@ -119,9 +119,13 @@ bool PopulateWordSet(WordSet* word_set, FILE* file, AffReader* aff_reader,
 
     if (found == word_set->end()) {
       word_set->insert(std::make_pair(utf8word, affix_vector));
+    } else if (affix_index == 0) {
+      // If we have no affix, add the no-affix case to the affix-vector.
+      found->second.insert(affix_index);

[rpetterson, 2012/12/15 03:53:47]

I'm not sure I see how this fixes the issue where later affixes need to
overwrite affixes earlier in the same file?

[please use gerrit instead, 2012/12/15 04:19:14]

This fixes the following case:

Word/S                   # <--- affix_index = 1
Word                     # <--- affix_index = 0

The dictionary writer here expected the final set of affixes indexes to be [1,
0]. Without this line, unit test SpellCheckTest.SpellCheckText unit test fails
for a few languages.

The case that you fixed was the following:

Word/S                   # <--- affix_index = 1
Word/!                   # <--- affix_index = 2

We need the final set of affix indexes to be [2].

[please use gerrit instead, 2012/12/15 22:00:19]

Rlp: Let me investigate more into how convert_dict works and bdic, dic, aff file
formats.

     } else {
-      // The affixes of the delta file should override those in the
-      // dictionary file.
+      // The later added affixes should override the earlier added affixes. This
+      // is useful for overrding the affixes in dic file from dic_delta file,
+      // for example.
       found->second.swap(affix_vector);
     }
   }