Properly report OOM when deoptimizer allocation fails

Properly report OOM when deoptimizer allocation fails

R=mstarzinger@chromium.org
BUG=

Committed: https://crrev.com/db04a5ad6f34da95ad54cda97ca70d6805cc5978
Cr-Commit-Position: refs/heads/master@{#27900}

[Michael Starzinger, 2015-04-16 15:32:23]

https://codereview.chromium.org/1052823003/diff/1/src/deoptimizer.cc
File src/deoptimizer.cc (right):

https://codereview.chromium.org/1052823003/diff/1/src/deoptimizer.cc#newcode2784
src/deoptimizer.cc:2784: void
Deoptimizer::EnsureCodeForDeoptimizationEntry(Isolate* isolate,
As mentioned offline: I am not entirely sure that it is safe to trigger a GC
within this method. It seems the call-site in
Deoptimizer::PatchCodeForDeoptimization is un-handlified and has some raw
pointers into the heap (i.e. deopt_data pointing to a DeoptimizationInputData)
that would become stale. Can we guarantee that this call-site will never trigger
an expansion of the deopt entry table?

https://codereview.chromium.org/1052823003/diff/1/src/deoptimizer.cc#newcode2813
src/deoptimizer.cc:2813: V8::FatalProcessOutOfMemory(
I like that we are properly reporting the allocation failure via
V8::FatalProcessOutOfMemory now, we should definitely land that part.

[Erik Corry Chromium.org, 2015-04-16 16:39:11]

https://codereview.chromium.org/1052823003/diff/1/src/deoptimizer.cc
File src/deoptimizer.cc (right):

https://codereview.chromium.org/1052823003/diff/1/src/deoptimizer.cc#newcode2784
src/deoptimizer.cc:2784: void
Deoptimizer::EnsureCodeForDeoptimizationEntry(Isolate* isolate,
On 2015/04/16 15:32:23, Michael Starzinger wrote:
> As mentioned offline: I am not entirely sure that it is safe to trigger a GC
> within this method. It seems the call-site in
> Deoptimizer::PatchCodeForDeoptimization is un-handlified and has some raw
> pointers into the heap (i.e. deopt_data pointing to a DeoptimizationInputData)
> that would become stale. Can we guarantee that this call-site will never
trigger
> an expansion of the deopt entry table?

Nice catch.  Removed the GC, now just report the OOM.

[Michael Starzinger, 2015-04-16 16:44:21]

LGTM. Only the CL description is off now.

[Erik Corry Chromium.org, 2015-04-16 17:04:54]

The CQ bit was checked by erikcorry@chromium.org

[commit-bot: I haz the power, 2015-04-16 17:05:07]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1052823003/20001

[commit-bot: I haz the power, 2015-04-16 18:05:38]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-16 18:05:50]

Patchset 2 (id:??) landed as
https://crrev.com/db04a5ad6f34da95ad54cda97ca70d6805cc5978
Cr-Commit-Position: refs/heads/master@{#27900}