[arm] Fix pushing of stale register in CallConstructStub.

[arm] Fix pushing of stale register in CallConstructStub.

This fixes a recent regression where the register usually holding the
original receiver was pushed onto the stack even when the stub was not
compiled for super calls (i.e. IsSuperConstructorCall). This led to
untagged values being present in the stack frame.

R=jacob.bramley@arm.com
BUG=chromium:512573
LOG=N

Committed: https://crrev.com/85d3b16386849be735dd5c189eeec9c9f70204ac
Cr-Commit-Position: refs/heads/master@{#29785}

[jbramley, 2015-07-22 07:51:36]

https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64.cc
File src/arm64/code-stubs-arm64.cc (right):

https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64....
src/arm64/code-stubs-arm64.cc:2758: __ Push(argc, function, feedback_vector,
index, orig_construct);
Why not pass orig_construct=NoReg for the !is_super case? That way you don't
need the complexity in this function (or in GenerateRecordCallTarget).

I don't think the same will work for ARM, but it's a separate Push there anyway.

[Michael Starzinger, 2015-07-22 08:07:41]

https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64.cc
File src/arm64/code-stubs-arm64.cc (right):

https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64....
src/arm64/code-stubs-arm64.cc:2758: __ Push(argc, function, feedback_vector,
index, orig_construct);
On 2015/07/22 07:51:36, jbramley wrote:
> Why not pass orig_construct=NoReg for the !is_super case? That way you don't
> need the complexity in this function (or in GenerateRecordCallTarget).
> 
> I don't think the same will work for ARM, but it's a separate Push there
anyway.

The way the 8-register push is implemented, it assumes that 5 registers are
always valid and only treats the last 3 registers as "var-args" that could be
invalid.

[jbramley, 2015-07-22 08:09:01]

On 2015/07/22 08:07:41, Michael Starzinger wrote:
>
https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64.cc
> File src/arm64/code-stubs-arm64.cc (right):
> 
>
https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64....
> src/arm64/code-stubs-arm64.cc:2758: __ Push(argc, function, feedback_vector,
> index, orig_construct);
> On 2015/07/22 07:51:36, jbramley wrote:
> > Why not pass orig_construct=NoReg for the !is_super case? That way you don't
> > need the complexity in this function (or in GenerateRecordCallTarget).
> > 
> > I don't think the same will work for ARM, but it's a separate Push there
> anyway.
> 
> The way the 8-register push is implemented, it assumes that 5 registers are
> always valid and only treats the last 3 registers as "var-args" that could be
> invalid.

Ah, fair enough. I think I'd like to fix that, but perhaps not for this patch.

lgtm

[Michael Starzinger, 2015-07-22 08:22:01]

On 2015/07/22 08:09:01, jbramley wrote:
> On 2015/07/22 08:07:41, Michael Starzinger wrote:
> >
>
https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64.cc
> > File src/arm64/code-stubs-arm64.cc (right):
> > 
> >
>
https://codereview.chromium.org/1241223003/diff/1/src/arm64/code-stubs-arm64....
> > src/arm64/code-stubs-arm64.cc:2758: __ Push(argc, function, feedback_vector,
> > index, orig_construct);
> > On 2015/07/22 07:51:36, jbramley wrote:
> > > Why not pass orig_construct=NoReg for the !is_super case? That way you
don't
> > > need the complexity in this function (or in GenerateRecordCallTarget).
> > > 
> > > I don't think the same will work for ARM, but it's a separate Push there
> > anyway.
> > 
> > The way the 8-register push is implemented, it assumes that 5 registers are
> > always valid and only treats the last 3 registers as "var-args" that could
be
> > invalid.
> 
> Ah, fair enough. I think I'd like to fix that, but perhaps not for this patch.

Acknowledged. SGTM.

[Michael Starzinger, 2015-07-22 08:27:56]

The CQ bit was checked by mstarzinger@chromium.org

[commit-bot: I haz the power, 2015-07-22 08:28:05]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1241223003/1

[commit-bot: I haz the power, 2015-07-22 08:30:47]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-07-22 08:30:48]

Try jobs failed on following builders:
  v8_presubmit on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_presubmit/builds/4354)

[Michael Starzinger, 2015-07-22 08:37:54]

The CQ bit was checked by mstarzinger@chromium.org

[commit-bot: I haz the power, 2015-07-22 08:38:07]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1241223003/1

[commit-bot: I haz the power, 2015-07-22 08:49:40]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-07-22 08:50:10]

Patchset 1 (id:??) landed as
https://crrev.com/85d3b16386849be735dd5c189eeec9c9f70204ac
Cr-Commit-Position: refs/heads/master@{#29785}