Revision: 467
http://vcs.pcre.org/viewvc?view=rev&revision=467
Author: ph10
Date: 2009-10-19 12:43:18 +0100 (Mon, 19 Oct 2009)
Log Message:
-----------
Fix study problem with JavaScript compatibility flag and back references.
Modified Paths:
--------------
code/trunk/pcre_study.c
code/trunk/testdata/testinput2
code/trunk/testdata/testoutput2
Modified: code/trunk/pcre_study.c
===================================================================
--- code/trunk/pcre_study.c 2009-10-19 11:42:08 UTC (rev 466)
+++ code/trunk/pcre_study.c 2009-10-19 11:43:18 UTC (rev 467)
@@ -314,18 +314,26 @@
logic is that a recursion can only make sense if there is another
alternation that stops the recursing. That will provide the minimum length
(when no recursion happens). A backreference within the group that it is
- referencing behaves in the same way. */
+ referencing behaves in the same way.
+
+ If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
+ matches an empty string (by default it causes a matching failure), so in
+ that case we must set the minimum length to zero. */
case OP_REF:
- ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));
- if (cs == NULL) return -2;
- do ce += GET(ce, 1); while (*ce == OP_ALT);
- if (cc > cs && cc < ce)
- {
- d = 0;
- had_recurse = TRUE;
+ if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
+ {
+ ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));
+ if (cs == NULL) return -2;
+ do ce += GET(ce, 1); while (*ce == OP_ALT);
+ if (cc > cs && cc < ce)
+ {
+ d = 0;
+ had_recurse = TRUE;
+ }
+ else d = find_minlength(cs, startcode, options);
}
- else d = find_minlength(cs, startcode, options);
+ else d = 0;
cc += 3;
/* Handle repeated back references */
Modified: code/trunk/testdata/testinput2
===================================================================
--- code/trunk/testdata/testinput2 2009-10-19 11:42:08 UTC (rev 466)
+++ code/trunk/testdata/testinput2 2009-10-19 11:43:18 UTC (rev 467)
@@ -3160,4 +3160,13 @@
ABX
BAXBAD
+/(\3)(\1)(a)/<JS>
+ cat
+
+/(\3)(\1)(a)/SI<JS>
+ cat
+
+/(\3)(\1)(a)/SI
+ cat
+
/-- End of testinput2 --/
Modified: code/trunk/testdata/testoutput2
===================================================================
--- code/trunk/testdata/testoutput2 2009-10-19 11:42:08 UTC (rev 466)
+++ code/trunk/testdata/testoutput2 2009-10-19 11:43:18 UTC (rev 467)
@@ -10431,4 +10431,36 @@
BAXBAD
No match
+/(\3)(\1)(a)/<JS>
+ cat
+ 0: a
+ 1:
+ 2:
+ 3: a
+
+/(\3)(\1)(a)/SI<JS>
+Capturing subpattern count = 3
+Max back reference = 3
+Options:
+No first char
+Need char = 'a'
+Subject length lower bound = 1
+No set of starting bytes
+ cat
+ 0: a
+ 1:
+ 2:
+ 3: a
+
+/(\3)(\1)(a)/SI
+Capturing subpattern count = 3
+Max back reference = 3
+No options
+No first char
+Need char = 'a'
+Subject length lower bound = 3
+No set of starting bytes
+ cat
+No match
+
/-- End of testinput2 --/
