Made regexp robust against changes to a string's implementation.

RegExps now restart if their input string changes representation during preemption.
Cleaned up the handling of strings moving, so strings moved by GC and strings changing representation are handled equivalently.

[Lasse Reichstein, 2009-03-20 10:05:55]

Largish review.

[Lasse Reichstein, 2009-03-20 10:05:56]

Largish review.

[Erik Corry, 2009-03-20 12:45:42]

LGTM

http://codereview.chromium.org/42441/diff/1008/1009
File src/jsregexp.cc (right):

http://codereview.chromium.org/42441/diff/1008/1009#newcode605
Line 605: previous_index);
A comment would be in order here!

http://codereview.chromium.org/42441/diff/1008/1010
File src/jsregexp.h (right):

http://codereview.chromium.org/42441/diff/1008/1010#newcode2
Line 2: // Redistribution and use in source and binary forms, with or without
2009?

http://codereview.chromium.org/42441/diff/1008/1011
File src/regexp-macro-assembler-ia32.cc (right):

http://codereview.chromium.org/42441/diff/1008/1011#newcode967
Line 967: // AssertNoAllocation, since regexps might be preempted, and
preemption code
s/preemption code/another thread/

http://codereview.chromium.org/42441/diff/1008/1011#newcode976
Line 976: subject_ptr = String::cast(ConsString::cast(subject_ptr)->first());
I don't think you need to cast this to a string.

http://codereview.chromium.org/42441/diff/1008/1011#newcode981
Line 981: subject_ptr = String::cast(slice->buffer());
Or this

http://codereview.chromium.org/42441/diff/1008/1013
File test/cctest/test-api.cc (right):

http://codereview.chromium.org/42441/diff/1008/1013#newcode5745
Line 5745: const char* c_source =
It would be nice to test GC during execution of a long regexp with postive and
negative lookaheads too.

http://codereview.chromium.org/42441/diff/1008/1013#newcode5960
Line 5960: // Match 15-30 "a"'s against 14 and a "b".
14 "a"s

[Lasse Reichstein, 2009-03-20 13:26:11]

Addressed issues except:

http://codereview.chromium.org/42441/diff/1008/1013
File test/cctest/test-api.cc (right):

http://codereview.chromium.org/42441/diff/1008/1013#newcode5745
Line 5745: const char* c_source =
Are they more likely to fail?
In any case, that would be another test case, since this one doesn't do GC.