Fix Chromium issue 47824.

Fix Chromium issue 47824. 
In rare cases a two-byte string was mistaken for an ascii-string.

[Mads Ager (chromium), 2010-06-29 14:20:09]

LGTM, thanks for fixing this.

[Vitaly Repeshko, 2010-06-29 14:36:31]

LGTM.

Thanks a lot for fixing this! I'm afraid I broke it in r4894:
 bool String::IsAsciiRepresentation() {
   uint32_t type = map()->instance_type();
-  if ((type & kStringRepresentationMask) == kConsStringTag &&
-      ConsString::cast(this)->second()->length() == 0) {
-    return ConsString::cast(this)->first()->IsAsciiRepresentation();
-  }
   return (type & kStringEncodingMask) == kAsciiStringTag;
 }


-- Vitaly

http://codereview.chromium.org/2858033/diff/1/5
File test/cctest/test-api.cc (right):

http://codereview.chromium.org/2858033/diff/1/5#newcode10980
test/cctest/test-api.cc:10980: i::String::cast(string->TryFlatten());
TryFlattenGetString returns the right string and requires no casting.

http://codereview.chromium.org/2858033/diff/1/5#newcode10985
test/cctest/test-api.cc:10985: uint16_t *uc16_buffer = new uint16_t[length + 1];
nit: Wrong placement of "*".

[Lasse Reichstein, 2010-06-30 07:14:41]

Reversing the change to ConstString::IsAsciiRepresentation would probably have
been an easier fix, but I think this is the better one.

http://codereview.chromium.org/2858033/diff/1/5
File test/cctest/test-api.cc (right):

http://codereview.chromium.org/2858033/diff/1/5#newcode10980
test/cctest/test-api.cc:10980: i::String::cast(string->TryFlatten());
Only if flattening succeeds, otherwise it just returns the original cons-string.
Likewise for TryFlatten, which returns the failure if flattenings fails.

I have added a new FlattenGetString that flattens unconditionally and returns
the non-cons string instead.

http://codereview.chromium.org/2858033/diff/1/5#newcode10985
test/cctest/test-api.cc:10985: uint16_t *uc16_buffer = new uint16_t[length + 1];
Good catch. Fixed.