*** empty log message ***
authorRichard Stallman <rms@gnu.org>
Mon, 7 Jun 1993 08:06:10 +0000 (08:06 +0000)
committerRichard Stallman <rms@gnu.org>
Mon, 7 Jun 1993 08:06:10 +0000 (08:06 +0000)
regex.c

diff --git a/regex.c b/regex.c
index bf3e968..4bb30f7 100644 (file)
--- a/regex.c
+++ b/regex.c
@@ -4130,11 +4130,27 @@ re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
                detect that here, the alternative has put on a dummy
                failure point which is what we will end up popping.  */
 
-           /* Skip over open/close-group commands.  */
-           while (p2 + 2 < pend
-                  && ((re_opcode_t) *p2 == stop_memory
-                      || (re_opcode_t) *p2 == start_memory))
-             p2 += 3;                  /* Skip over args, too.  */
+           /* Skip over open/close-group commands.
+              If what follows this loop is a ...+ construct,
+              look at what begins its body, since we will have to
+              match at least one of that.  */
+           while (1)
+             {
+               if (p2 + 2 < pend
+                   && ((re_opcode_t) *p2 == stop_memory
+                       || (re_opcode_t) *p2 == start_memory))
+                 p2 += 3;
+               else if (p2 + 6 < pend
+                        && (re_opcode_t) *p2 == dummy_failure_jump)
+                 p2 += 6;
+               else
+                 break;
+             }
+
+           p1 = p + mcnt;
+           /* p1[0] ... p1[2] are the `on_failure_jump' corresponding
+              to the `maybe_finalize_jump' of this case.  Examine what 
+              follows.  */
 
             /* If we're at the end of the pattern, we can change.  */
             if (p2 == pend)
@@ -4152,11 +4168,7 @@ re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
              {
                register unsigned char c
                   = *p2 == (unsigned char) endline ? '\n' : p2[2];
-               p1 = p + mcnt;
 
-                /* p1[0] ... p1[2] are the `on_failure_jump' corresponding
-                   to the `maybe_finalize_jump' of this case.  Examine what 
-                   follows.  */
                 if ((re_opcode_t) p1[3] == exactn && p1[5] != c)
                   {
                    p[-3] = (unsigned char) pop_failure_jump;
@@ -4182,6 +4194,54 @@ re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
                       }
                  }
              }
+            else if ((re_opcode_t) *p2 == charset)
+             {
+               register unsigned char c
+                  = *p2 == (unsigned char) endline ? '\n' : p2[2];
+
+                if ((re_opcode_t) p1[3] == exactn
+                   && ! (p2[1] * BYTEWIDTH > p1[4]
+                         && (p2[1 + p1[4] / BYTEWIDTH]
+                             & (1 << (p1[4] % BYTEWIDTH)))))
+                  {
+                   p[-3] = (unsigned char) pop_failure_jump;
+                    DEBUG_PRINT3 ("  %c != %c => pop_failure_jump.\n",
+                                  c, p1[5]);
+                  }
+                  
+               else if ((re_opcode_t) p1[3] == charset_not)
+                 {
+                   int idx;
+                   /* We win if the charset_not inside the loop
+                      lists every character listed in the charset after.  */
+                   for (idx = 0; idx < p2[1]; idx++)
+                     if (! (p2[2 + idx] == 0
+                            || (idx < p1[4]
+                                && ((p2[2 + idx] & ~ p1[5 + idx]) == 0))))
+                       break;
+
+                   if (idx == p2[1])
+                      {
+                       p[-3] = (unsigned char) pop_failure_jump;
+                        DEBUG_PRINT1 ("  No match => pop_failure_jump.\n");
+                      }
+                 }
+               else if ((re_opcode_t) p1[3] == charset)
+                 {
+                   int idx;
+                   /* We win if the charset inside the loop
+                      has no overlap with the one after the loop.  */
+                   for (idx = 0; idx < p2[1] && idx < p1[4]; idx++)
+                     if ((p2[2 + idx] & p1[5 + idx]) != 0)
+                       break;
+
+                   if (idx == p2[1] || idx == p1[4])
+                      {
+                       p[-3] = (unsigned char) pop_failure_jump;
+                        DEBUG_PRINT1 ("  No match => pop_failure_jump.\n");
+                      }
+                 }
+             }
          }
          p -= 2;               /* Point at relative address again.  */
          if ((re_opcode_t) p[-1] != pop_failure_jump)