Remove HeapProxy and use the BlockHeapManager by default.

Remove HeapProxy and use the BlockHeapManager by default.

I've run some Dromaeo benchmarks and there's no significant performance difference between the 2 version (it's in the noise of the measure).

BUG=
R=chrisha@chromium.org

Committed: https://code.google.com/p/sawbuck/source/detail?r=2281

[Sébastien Marchand, 2014-08-25 20:18:35]

PTAL. I'll run some benchmarks to make sure that this isn't significantly slower
than the current implementation. I'll update the CL description with the numbers
once I have them.

[chrisha, 2014-08-26 19:25:17]

So much negative code, I love it!

lgtm, even without nits (!)

[Sébastien Marchand, 2014-08-27 13:59:27]

Thanks, finally pulling the plug on HeapProxy ! Committing.

[Sébastien Marchand, 2014-08-27 15:15:00]

struggling with a deadlock in the integration tests... It seems that it's
related to the stdio stream... Investigating...

[Sébastien Marchand, 2014-08-28 20:14:51]

I've finally found the problem ! It's due to the fact that we don't support the
flags in the WinHeap interceptors... and during the CRT initialization there's a
call to HeapAlloc with the HEAP_ZERO_MEMORY flag... This is when we allocate the
crt!__piob array, and this array has to be zero initialized if we don't want to
crash when we teardown the crt... 

I'll try to add support for this in the Windows Heap adapter

[chrisha, 2014-08-28 20:27:41]

Ah, excellent. Should be easy enough to add!


On Thu, Aug 28, 2014 at 4:14 PM, <sebmarchand@chromium.org> wrote:

> I've finally found the problem ! It's due to the fact that we don't
> support the flags in the WinHeap interceptors... and during the CRT
> initialization there's a call to HeapAlloc with the HEAP_ZERO_MEMORY
> flag... This is when we allocate the crt!__piob array, and this array
> has to be zero initialized if we don't want to crash when we teardown
> the crt...
>
> I'll try to add support for this in the Windows Heap adapter
>
> https://codereview.appspot.com/135780043/
>
> --
> You received this message because you are subscribed to the Google Groups
> "sawbuck-changes" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sawbuck-changes+unsubscribe@googlegroups.com.
> To post to this group, send email to sawbuck-changes@googlegroups.com.
> Visit this group at http://groups.google.com/group/sawbuck-changes.
> For more options, visit https://groups.google.com/d/optout.
>

[chrisha, 2014-08-28 20:28:08]

We should look at all other flags that end up being used, and add support
for those that make sense....


On Thu, Aug 28, 2014 at 4:27 PM, Chris Hamilton <chrisha@chromium.org>
wrote:

> Ah, excellent. Should be easy enough to add!
>
>
> On Thu, Aug 28, 2014 at 4:14 PM, <sebmarchand@chromium.org> wrote:
>
>> I've finally found the problem ! It's due to the fact that we don't
>> support the flags in the WinHeap interceptors... and during the CRT
>> initialization there's a call to HeapAlloc with the HEAP_ZERO_MEMORY
>> flag... This is when we allocate the crt!__piob array, and this array
>> has to be zero initialized if we don't want to crash when we teardown
>> the crt...
>>
>> I'll try to add support for this in the Windows Heap adapter
>>
>> https://codereview.appspot.com/135780043/
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "sawbuck-changes" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to sawbuck-changes+unsubscribe@googlegroups.com.
>> To post to this group, send email to sawbuck-changes@googlegroups.com.
>> Visit this group at http://groups.google.com/group/sawbuck-changes.
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>

[Sébastien Marchand, 2014-08-29 13:17:10]

Bug has been fixed ! Committing.

[Sébastien Marchand, 2014-08-29 13:21:42]

Committed patchset #3 (id:60001) manually as 2281.