Oilpan: ENABLE(GC_PROFILING): Don't emit free list entries for LargeObject heap

Oilpan: ENABLE(GC_PROFILING): Don't emit free list entries for LargeObject heap

Before this CL, the tracing.json output by the profiler couldn't be parsed
correctly by dino.js visualizer, as it contained invalid entries from LargeObject heap.
This CL fixes the issue by disabling dump of free list entries for LargeObject heap.

BUG=454273

[kouhei (in TOK), 2015-03-10 02:15:25]

kouhei@chromium.org changed reviewers:
+ keishi@chromium.org, oilpan-reviews@chromium.org

[haraken, 2015-03-10 02:18:31]

This fix is ok, but can we fix LargeObjectHeap::snapshotFreeList so that it
doesn't output invalid a free list snapshot?

[kouhei (in TOK), 2015-03-10 09:09:06]

haraken: According to keishi and tkent, LargeObject doesn't have a free list so
this patch should be correct? The other stats for the LargeObject heap is
collected from ThreadState::snapshot().

[haraken, 2015-03-10 11:05:48]

On 2015/03/10 09:09:06, kouhei wrote:
> haraken: According to keishi and tkent, LargeObject doesn't have a free list
so
> this patch should be correct? The other stats for the LargeObject heap is
> collected from ThreadState::snapshot().

You're right. However, snapshotFreeList is a virtual method of BaseHeap, so I
think it should be allowed to call snapshotFreeList for LargeObjectHeap
(LargeObjectHeap::snapshotFreeList is already implemented so that it does
nothing). I guess the profiler should be fixed so that it doesn't crash when it
encounters an empty entry in the json.

[keishi, 2015-03-11 02:25:24]

On 2015/03/10 11:05:48, haraken wrote:
> On 2015/03/10 09:09:06, kouhei wrote:
> > haraken: According to keishi and tkent, LargeObject doesn't have a free list
> so
> > this patch should be correct? The other stats for the LargeObject heap is
> > collected from ThreadState::snapshot().
> 
> You're right. However, snapshotFreeList is a virtual method of BaseHeap, so I
> think it should be allowed to call snapshotFreeList for LargeObjectHeap
> (LargeObjectHeap::snapshotFreeList is already implemented so that it does
> nothing). I guess the profiler should be fixed so that it doesn't crash when
it
> encounters an empty entry in the json.

Done. Let me know if it doesn't work.

[kouhei (in TOK), 2015-03-11 05:51:42]

On 2015/03/11 02:25:24, keishi wrote:
> On 2015/03/10 11:05:48, haraken wrote:
> > On 2015/03/10 09:09:06, kouhei wrote:
> > > haraken: According to keishi and tkent, LargeObject doesn't have a free
list
> > so
> > > this patch should be correct? The other stats for the LargeObject heap is
> > > collected from ThreadState::snapshot().
> > 
> > You're right. However, snapshotFreeList is a virtual method of BaseHeap, so
I
> > think it should be allowed to call snapshotFreeList for LargeObjectHeap
> > (LargeObjectHeap::snapshotFreeList is already implemented so that it does
> > nothing). I guess the profiler should be fixed so that it doesn't crash when
> it
> > encounters an empty entry in the json.
> 
> Done. Let me know if it doesn't work.

Thanks! closing this as the patch is no longer needed.