blob: 9cf7e3e104f007c2dc42fb19d7e577f2eb38655b (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
|
--- icu/source/i18n/regexcmp.cpp (revision 31398)
+++ icu/source/i18n/regexcmp.cpp (revision 31782)
@@ -3307,8 +3307,29 @@
case URX_STRING_I:
- // TODO: Is the case-folded string the longest?
- // If so we can optimize this the same as URX_STRING.
- loc++;
- currentLen = INT32_MAX;
+ // TODO: This code assumes that any user string that matches will be no longer
+ // than our compiled string, with case insensitive matching.
+ // Our compiled string has been case-folded already.
+ //
+ // Any matching user string will have no more code points than our
+ // compiled (folded) string. Folding may add code points, but
+ // not remove them.
+ //
+ // There is a potential problem if a supplemental code point
+ // case-folds to a BMP code point. In this case our compiled string
+ // could be shorter (in code units) than a matching user string.
+ //
+ // At this time (Unicode 6.1) there are no such characters, and this case
+ // is not being handled. A test, intltest regex/Bug9283, will fail if
+ // any problematic characters are added to Unicode.
+ //
+ // If this happens, we can make a set of the BMP chars that the
+ // troublesome supplementals fold to, scan our string, and bump the
+ // currentLen one extra for each that is found.
+ //
+ {
+ loc++;
+ int32_t stringLenOp = (int32_t)fRXPat->fCompiledPat->elementAti(loc);
+ currentLen = safeIncrement(currentLen, URX_VAL(stringLenOp));
+ }
break;
|
