* Implemented case-insensitive back-reference matching in irregexp-ia32.

Makes irregexp-ia32 feature complete wrt. regexps.

[Lasse Reichstein, 2008-12-04 14:12:00]

Code-review, please.

[Erik Corry, 2008-12-04 14:36:58]

LGTM with comments.

http://codereview.chromium.org/12944/diff/1/2
File src/assembler-ia32.cc (right):

http://codereview.chromium.org/12944/diff/1/2#newcode1562
Line 1562: void Assembler::loop(Label* loop_to) {
From comp.lang.asm.x86
The loop instruction can only branch -128 to 127 bytes from eip. On Intel
processors, the loop instruction has always been horribly slow, so people
generally avoid it. (AMD made it fast on the K6, but they intentionally made
it slow in Athlon to avoid software bugs from people relying on its speed.)

I wonder if this is a good idea?

http://codereview.chromium.org/12944/diff/1/4
File src/regexp-macro-assembler-ia32.cc (right):

http://codereview.chromium.org/12944/diff/1/4#newcode290
Line 290: __ push(edi);
Please test this on a Mac before submitting.

http://codereview.chromium.org/12944/diff/1/4#newcode669
Line 669: int RegExpMacroAssemblerIA32::CaseInsensitiveCompareUC16(uc16**
buffer,
The interpreter has this routine too.  They should probably be unified.

http://codereview.chromium.org/12944/diff/1/7
File test/mjsunit/regexp.js (right):

http://codereview.chromium.org/12944/diff/1/7#newcode268
Line 268: assertEquals("xaBcAbCABCx,aBc",
String(/x(...)\1\1x/i.exec("xaBcAbCABCx")));
You should use \u to add unicode tests here without making the file non-ASCII. 
For both ASCII and non-ASCII there should be at least one test that hits the end
of the string, one with a zero length capture and one with an unbound capture.

[Lasse Reichstein, 2008-12-05 15:19:21]

http://codereview.chromium.org/12944/diff/1/2
File src/assembler-ia32.cc (right):

http://codereview.chromium.org/12944/diff/1/2#newcode1562
Line 1562: void Assembler::loop(Label* loop_to) {
Re: loop. I turned out not to use it anyway, so let's remove it.

http://codereview.chromium.org/12944/diff/1/7
File test/mjsunit/regexp.js (right):

http://codereview.chromium.org/12944/diff/1/7#newcode268
Line 268: assertEquals("xaBcAbCABCx,aBc",
String(/x(...)\1\1x/i.exec("xaBcAbCABCx")));
Indeed, more tests.
Adding non-ASCII characters won't do a lot now, but it'll be  good to have them
for when we start compiling specifically for ASCII strings.