Fix issue 491: constantpool dump violates ARM debugger assertion for return p...

Fix issue 491: constantpool dump violates ARM debugger assertion for return point

The generation of the return sequence is now protected from having the constant pool emitted inside of it in both compilers.

BUG=http://code.google.com/p/v8/issues/detail?id=491
TEST=test/mjsunit/regress/regress-491.js

Committed: http://code.google.com/p/v8/source/detail?r=3215

[Søren Thygesen Gjesse, 2009-11-04 12:02:59]

I did not find a good place for sharing the code for the return sequence as one
code generator is using the virtual frame whereas the other is not.

[Kevin Millikin (Chromium), 2009-11-04 12:51:00]

Looks good to me.

http://codereview.chromium.org/362003/diff/2001/2009
File src/arm/assembler-arm.cc (right):

http://codereview.chromium.org/362003/diff/2001/2009#newcode1321
Line 1321: void Assembler::PostponeConstantPool(int instructions) {
I like "BlockConstPoolFor" to make the connection to "BlockConstPoolBefore"
obvious.

[Erik Corry, 2009-11-04 13:16:06]

LGTM.   At some point we should introduce a RAII class for controlling ranges
where we don't want extra stuff messsing up the instruction stream.