Reland: Preprocess structured stack trace on GC to get rid of code reference.

Reland: Preprocess structured stack trace on GC to get rid of code reference.

BUG=v8:2340
LOG=N

Committed: https://crrev.com/46b3582f48a3026b3087f4548b7cb3b2ff5b1102
Cr-Commit-Position: refs/heads/master@{#28102}

[Yang, 2015-04-28 07:55:47]

yangguo@chromium.org changed reviewers:
+ hpayer@chromium.org

[Yang, 2015-04-28 07:55:47]

The GC stress failures are due to GC triggering while we add a stack trace to
the weak fixed array. Specifically, this:

- We add a new stack trace to the weak fixed array via WeakFixedArray::Add.
- There is no remaining slot in the current weak fixed array, so we allocate a
larger one and copy the old weak fixed array into it.
- WeakFixedArray::Set allocates a weak cell. This triggers a GC.
- Right after GC, we iterate the (old) weak fixed array in
Heap::PreprocessStackTraces, which turns code references into Smis.
- Back in WeakFixedArray::Add, we return the new weak fixed array, which has not
been cleared by Heap::PreprocessStackTraces, as it iterated over the old weak
fixed array.
- The next time we run Heap::PreprocessStackTraces, which now iterates over the
new weak fixed array, we try to preprocess stack traces that have already been
preprocessed. Casting a Smi into a code object fails.

[Hannes Payer (out of office), 2015-04-28 10:37:05]

lgtm

[Yang, 2015-04-28 10:49:01]

The CQ bit was checked by yangguo@chromium.org

[commit-bot: I haz the power, 2015-04-28 10:49:04]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1109093002/20001

[commit-bot: I haz the power, 2015-04-28 11:13:29]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-28 11:13:39]

Patchset 2 (id:??) landed as
https://crrev.com/46b3582f48a3026b3087f4548b7cb3b2ff5b1102
Cr-Commit-Position: refs/heads/master@{#28102}