Fix assertion caused by wrong unicharset (issue #2301)

Credit to OSS-Fuzz: This fixes an issue which was reported by OSS-Fuzz, see details at https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=13592. OSS-Fuzz triggered this assertion: contains_unichar_id(unichar_id):Error:Assert failed:in file ../../src/ccutil/unicharset.h, line 502 Signed-off-by: Stefan Weil <[email protected]>
tesseract-ocr · Mar 10, 2019 · 91d0a71 · 91d0a71
1 parent 0e72733
commit 91d0a71
Show file tree

Hide file tree

Showing 4 changed files with 10 additions and 11 deletions.
diff --git a/src/dict/context.cpp b/src/dict/context.cpp
@@ -43,16 +43,17 @@ const int case_state_table[6][4] = {
      5, -1, 2, -1},
 };
 
-int Dict::case_ok(const WERD_CHOICE &word, const UNICHARSET &unicharset) const {
+int Dict::case_ok(const WERD_CHOICE &word) const {
   int state = 0;
   int x;
+  const UNICHARSET* unicharset = word.unicharset();
   for (x = 0; x < word.length(); ++x) {
     UNICHAR_ID ch_id = word.unichar_id(x);
-    if (unicharset.get_isupper(ch_id))
+    if (unicharset->get_isupper(ch_id))
       state = case_state_table[state][1];
-    else if (unicharset.get_islower(ch_id))
+    else if (unicharset->get_islower(ch_id))
       state = case_state_table[state][2];
-    else if (unicharset.get_isdigit(ch_id))
+    else if (unicharset->get_isdigit(ch_id))
       state = case_state_table[state][3];
     else
       state = case_state_table[state][0];

diff --git a/src/dict/dict.cpp b/src/dict/dict.cpp
@@ -678,7 +678,7 @@ void Dict::adjust_word(WERD_CHOICE *word,
                        bool debug) {
   bool is_han = (getUnicharset().han_sid() != getUnicharset().null_sid() &&
                  word->GetTopScriptID() == getUnicharset().han_sid());
-  bool case_is_ok = (is_han || case_ok(*word, getUnicharset()));
+  bool case_is_ok = (is_han || case_ok(*word));
   bool punc_is_ok = (is_han || !nonword || valid_punctuation(*word));
 
   float adjust_factor = additional_adjust;

diff --git a/src/dict/dict.h b/src/dict/dict.h
@@ -293,7 +293,7 @@ class Dict {
   void SettupStopperPass2();
   /* context.cpp *************************************************************/
   /// Check a string to see if it matches a set of lexical rules.
-  int case_ok(const WERD_CHOICE &word, const UNICHARSET &unicharset) const;
+  int case_ok(const WERD_CHOICE& word) const;
   /// Returns true if the word looks like an absolute garbage
   /// (e.g. image mistakenly recognized as text).
   bool absolute_garbage(const WERD_CHOICE &word, const UNICHARSET &unicharset);

diff --git a/src/dict/stopper.cpp b/src/dict/stopper.cpp
@@ -2,7 +2,6 @@
  ** Filename:    stopper.c
  ** Purpose:     Stopping criteria for word classifier.
  ** Author:      Dan Johnson
- ** History:     Mon Apr 29 14:56:49 1991, DSJ, Created.
  **
  ** (c) Copyright Hewlett-Packard Company, 1988.
  ** Licensed under the Apache License, Version 2.0 (the "License");
@@ -49,7 +48,7 @@ bool Dict::AcceptableChoice(const WERD_CHOICE& best_choice,
 
   bool no_dang_ambigs = !best_choice.dangerous_ambig_found();
   bool is_valid_word = valid_word_permuter(best_choice.permuter(), false);
-  bool is_case_ok = case_ok(best_choice, getUnicharset());
+  bool is_case_ok = case_ok(best_choice);
 
   if (stopper_debug_level >= 1) {
     const char *xht = "UNKNOWN";
@@ -107,15 +106,14 @@ bool Dict::AcceptableResult(WERD_RES *word) const {
     tprintf("\nRejecter: %s (word=%c, case=%c, unambig=%c, multiple=%c)\n",
             word->best_choice->debug_string().string(),
             (valid_word(*word->best_choice) ? 'y' : 'n'),
-            (case_ok(*word->best_choice, getUnicharset()) ? 'y' : 'n'),
+            (case_ok(*word->best_choice) ? 'y' : 'n'),
             word->best_choice->dangerous_ambig_found() ? 'n' : 'y',
             word->best_choices.singleton() ? 'n' : 'y');
   }
 
   if (word->best_choice->length() == 0 || !word->best_choices.singleton())
     return false;
-  if (valid_word(*word->best_choice) &&
-      case_ok(*word->best_choice, getUnicharset())) {
+  if (valid_word(*word->best_choice) && case_ok(*word->best_choice)) {
     WordSize = LengthOfShortestAlphaRun(*word->best_choice);
     WordSize -= stopper_smallword_size;
     if (WordSize < 0)