[turbofan] Carefully update frame states during inlining.

[turbofan] Carefully update frame states during inlining.

During inlining we should pay attention to only rewire the outer frame
states of the inlinee, but leave any inner frame states of the inlinee
untouched.  Otherwise we might run into trouble once we start caching
graphs, or do getter/setter inlining.

R=jarin@chromium.org

Committed: https://crrev.com/fe690c3e82a6e6092e881fd2c5629ab53d29a1c8
Cr-Commit-Position: refs/heads/master@{#28466}

[Benedikt Meurer, 2015-05-19 08:18:10]

Hey Jaro,

This is the change we talked about briefly last week. I know it's not required
currently, but I feel better if we make sure that we never mess with inner frame
states during inlining. Please take a look.

Thanks,
Benedikt

[Jarin, 2015-05-19 08:24:27]

On 2015/05/19 08:18:10, Benedikt Meurer wrote:
> Hey Jaro,
> 
> This is the change we talked about briefly last week. I know it's not required
> currently, but I feel better if we make sure that we never mess with inner
frame
> states during inlining. Please take a look.
> 
> Thanks,
> Benedikt

It seems to me that if you just do not connect the outer state if there is
already one it is also wrong. You should really connect the frame state to the
end of the chain. Since we do not have any test coverage for that I would prefer
just CHECKing that there is no outer frame state.

[Jarin, 2015-05-19 09:09:32]

On 2015/05/19 08:24:27, jarin wrote:
> On 2015/05/19 08:18:10, Benedikt Meurer wrote:
> > Hey Jaro,
> > 
> > This is the change we talked about briefly last week. I know it's not
required
> > currently, but I feel better if we make sure that we never mess with inner
> frame
> > states during inlining. Please take a look.
> > 
> > Thanks,
> > Benedikt
> 
> It seems to me that if you just do not connect the outer state if there is
> already one it is also wrong. You should really connect the frame state to the
> end of the chain. Since we do not have any test coverage for that I would
prefer
> just CHECKing that there is no outer frame state.

As discussed offline, I now see that this patch is in fact connecting the frame
state to the end of the chain.

lgtm.

[Benedikt Meurer, 2015-05-19 09:09:54]

The CQ bit was checked by bmeurer@chromium.org

[commit-bot: I haz the power, 2015-05-19 09:10:00]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1134973006/1

[commit-bot: I haz the power, 2015-05-19 09:11:36]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-05-19 09:11:47]

Patchset 1 (id:??) landed as
https://crrev.com/fe690c3e82a6e6092e881fd2c5629ab53d29a1c8
Cr-Commit-Position: refs/heads/master@{#28466}