Omnibox: Fix HQP Crash on Bad FormatURL and can_inline Interaction

chrome/browser/autocomplete/history_quick_provider.cc

Index: chrome/browser/autocomplete/history_quick_provider.cc
diff --git a/chrome/browser/autocomplete/history_quick_provider.cc b/chrome/browser/autocomplete/history_quick_provider.cc
index 0f0405f918c96770f95cfec4c0708713fcf5016a..06e83002a577bc8f5965cb714ebe0a5b46e326d4 100644
--- a/chrome/browser/autocomplete/history_quick_provider.cc
+++ b/chrome/browser/autocomplete/history_quick_provider.cc
@@ -277,7 +277,17 @@ AutocompleteMatch HistoryQuickProvider::QuickMatchToACMatch(
   match.contents_class =
       SpansFromTermMatch(new_matches, match.contents.length(), true);
 
-  if (history_match.can_inline()) {
+  // Set |inline_autocompletion| and |allowed_to_be_default_match| if possible.
+  // The second part of this test can happen if the only match(es) of the user's
+  // term occur in places FormatUrl() decides to omit in the formatted url.
+  // In these cases, it's impossible to set |inline_autocompletion| correctly
+  // and hence the match cannot be the default match.  I (mpearson@) believe
+  // this is likely caused by the mismatch that offsets are originally
+  // computed with respect to the cleaned-up URL yet then applied and
+  // updated by FormatUrl() as if they applied to the original string.
+  // See crbug.com/252630.
+  // TODO(mpearson): try removing the second clause after fixing 252630.

[Peter Kasting, 2014/04/14 22:13:21]

Nit: removing the second clause -> replacing the second clause with a DCHECK

(And then remove the DCHECK inside the conditional.)

[Mark P, 2014/04/14 23:34:20]

Done.
Done.

+  if (history_match.can_inline() && !new_matches.empty()) {
     DCHECK(!new_matches.empty());
     size_t inline_autocomplete_offset = new_matches[0].offset +
         new_matches[0].length;