Fix an issue in the ARM port where a left shift was predicted to have...

Fix an issue in the ARM port where a left shift was predicted to have
a Smi result when it had an int32 result.

BUG=none
TEST=test/mjsunit/shifts.js

[Erik Corry, 2010-08-19 07:45:27]

Great, thanks.  Please add a test to the mjsunit directory.  If you put a magic
comment:
// Flags: --nofull-compiler
then the test will be run by the optimizing compiler.

http://codereview.chromium.org/3195004/diff/1/2
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/3195004/diff/1/2#newcode1240
src/arm/codegen-arm.cc:1240: } else {  // Token::SHL
Instead of Token::SHL being in a comment it should be in an ASSERT.

[Rodolph Perfetta, 2010-08-19 17:21:17]

applied comments and added new unit test.

[Rodolph Perfetta, 2010-08-19 17:21:32]

http://codereview.chromium.org/3195004/diff/1/2
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/3195004/diff/1/2#newcode1240
src/arm/codegen-arm.cc:1240: } else {  // Token::SHL
On 2010/08/19 07:45:27, Erik Corry wrote:
> Instead of Token::SHL being in a comment it should be in an ASSERT.

Done.

[Erik Corry, 2010-08-23 07:43:40]

LGTM

http://codereview.chromium.org/3195004/diff/5001/6002
File test/mjsunit/shifts.js (right):

http://codereview.chromium.org/3195004/diff/5001/6002#newcode30
test/mjsunit/shifts.js:30: // test a few corners cases with shifts
I fixed the punctuation here.

[Erik Corry, 2010-08-23 07:54:21]

Committed as r5315.