Fix GC problem with ToBooleanStub

Use the generic version of the stub (without IC patching) to avoid triggering a GC when invoking the ToBooleanStub in optimized code.

[fschneider, 2011-07-27 11:59:36]

This is an attempt to fix the safepoint problem with the ToBoolean stub. It
would be nice to have a regression test, but I fear, it may not easy to create
one.

[Vitaly Repeshko, 2011-07-27 13:09:05]

Almost looks good. Two questions. Don't we need CallRuntimeSaveDoubles in the
stub itself? Is there are a deep reason why ARM is taking the argument in a
register and other platforms on the stack?


-- Vitaly

[fschneider, 2011-07-27 14:30:08]

New approach. As discussed we just don't invoke a stub that does patching.

[Sven Panne, 2011-07-27 14:46:33]

LGTM with a tiny stylistic change

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h
File src/code-stubs.h (right):

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h#newcode938
src/code-stubs.h:938: static Types all_types() { return Types(kAllTypes); }
Removing kAllTypes and directly using

 (1 << NUMBER_OF_TYPES) - 1

in all_types() looks a bit cleaner, I think.

[Vitaly Repeshko, 2011-07-27 14:49:37]

LGTM

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h
File src/code-stubs.h (right):

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h#newcode924
src/code-stubs.h:924: bool IsEmpty() const { return set_.IsEmpty(); }
Maybe add IsAll()?

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h#newcode938
src/code-stubs.h:938: static Types all_types() { return Types(kAllTypes); }
On 2011/07/27 14:46:33, Sven wrote:
> Removing kAllTypes and directly using
> 
>  (1 << NUMBER_OF_TYPES) - 1
> 
> in all_types() looks a bit cleaner, I think.

I agree.

[fschneider, 2011-07-27 14:56:19]

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h
File src/code-stubs.h (right):

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h#newcode924
src/code-stubs.h:924: bool IsEmpty() const { return set_.IsEmpty(); }
On 2011/07/27 14:49:38, Vitaly Repeshko wrote:
> Maybe add IsAll()?

Done.

http://codereview.chromium.org/7465066/diff/5002/src/code-stubs.h#newcode938
src/code-stubs.h:938: static Types all_types() { return Types(kAllTypes); }
On 2011/07/27 14:46:33, Sven wrote:
> Removing kAllTypes and directly using
> 
>  (1 << NUMBER_OF_TYPES) - 1
> 
> in all_types() looks a bit cleaner, I think.

Done.