# HG changeset patch
# User Aidan Kehoe <kehoea@parhasard.net>
# Date 1264890339 0
# Node ID 7504864a986c1d7863e35f33c9b70b28f872305d
# Parent  0eccfd4850d60d5cf6405ee8158dd72c26820af5
Don't use Boyer-Moore if repeated octets & case-insensitive search.

2010-01-30  Aidan Kehoe  <kehoea@parhasard.net>

	* search.c (search_buffer):
	Don't use Boyer-Moore for case-insensitive search if the search
	pattern contains repeated Ibytes and the corresponding character
	has case information (or, equivalently, if one of its case
	equivalents would contain repeated Ibytes).

diff -r 0eccfd4850d6 -r 7504864a986c src/ChangeLog
--- a/src/ChangeLog	Sat Jan 30 19:23:43 2010 +0000
+++ b/src/ChangeLog	Sat Jan 30 22:25:39 2010 +0000
@@ -1,3 +1,11 @@
+2010-01-30  Aidan Kehoe  <kehoea@parhasard.net>
+
+	* search.c (search_buffer):
+	Don't use Boyer-Moore for case-insensitive search if the search
+	pattern contains repeated Ibytes and the corresponding character
+	has case information (or, equivalently, if one of its case
+	equivalents would contain repeated Ibytes).
+
 2010-01-28  Jerry James  <james@xemacs.org>
 
 	* Makefile.in.in: Remove internationalization rules, since the
diff -r 0eccfd4850d6 -r 7504864a986c src/search.c
--- a/src/search.c	Sat Jan 30 19:23:43 2010 +0000
+++ b/src/search.c	Sat Jan 30 22:25:39 2010 +0000
@@ -1425,6 +1425,37 @@
                           break;
                         }
                     }
+
+		  if (ichar_len (c) > 2)
+		    {
+		      /* Case-equivalence plus repeated octets throws off
+			 the construction of the stride table; avoid this.
+
+		         It should be possible to correct boyer_moore to
+		         behave correctly even in this case--it doesn't have
+		         problems with repeated octets when case conversion
+		         is not involved--but this is not a critical
+		         issue. */
+		      Ibyte encoded[MAX_ICHAR_LEN];
+		      Bytecount len = set_itext_ichar (encoded, c);
+		      int i, j;
+		      for (i = 0; i < len && boyer_moore_ok; ++i)
+			{
+			  for (j = 0; i < len && boyer_moore_ok; ++j)
+			    {
+			      if (encoded[i] == encoded[j])
+				{
+				  boyer_moore_ok = 0;
+				}
+			    }
+			}
+
+		      if (0 == boyer_moore_ok)
+			{
+			  break;
+			}
+		    }
+			  
                 } while (c != starting_c);
 
               if (!checked)