Revert 241291 (Fix a leak in ResourceDispatcher on cancel).

Revert 241291 (Fix a leak in ResourceDispatcher on cancel).

The CL fixed a memory leak when requests are cancelled, but
caused crashes under certain circumstances.  The leaks will
still need to be fixed after this lands.

BUG=chrome-os-partner:27911, 328092
> Fix a long standing leak on cancellation of renderer-
> initiated URLRequests.  ResourceDispatcher was not calling
> peer->OnCompletedRequest on cancellation of async requests,
> despite it being expected by WebURLLoaderImpl::Context
> (The only ResourceLoaderBridge::Peer).
>
> This shouldn't impact externally visible behavior at all,
> other than fixing the leak, because once
> WebURLLoaderImpl::Context::Cancel() is called, it will never
> call back into WebURLLoaderImpl.
>
> BUG=328092
>
> Review URL: https://codereview.chromium.org/113893007

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=273149

[mmenke, 2014-05-27 20:33:19]

This isn't an exact revert, because it's no longer necessary to call
"SiteIsolationPolicy::OnRequestComplete(request_id);" as of
https://codereview.chromium.org/187063006.

[mmenke, 2014-05-27 20:34:52]

On 2014/05/27 20:33:19, mmenke wrote:
> This isn't an exact revert, because it's no longer necessary to call
> "SiteIsolationPolicy::OnRequestComplete(request_id);" as of
> https://codereview.chromium.org/187063006.

It may be simpler to, when fixing the leak, just make Cancel clean up properly,
instead of waiting to receive the corresponding cancel from the browser
process...  Though I'm not really sure.  I'm not entirely sure exactly what else
needs to be cleaned up.

[jam, 2014-05-28 00:39:24]

On 2014/05/27 20:34:52, mmenke wrote:
> On 2014/05/27 20:33:19, mmenke wrote:
> > This isn't an exact revert, because it's no longer necessary to call
> > "SiteIsolationPolicy::OnRequestComplete(request_id);" as of
> > https://codereview.chromium.org/187063006.
> 
> It may be simpler to, when fixing the leak, just make Cancel clean up
properly,
> instead of waiting to receive the corresponding cancel from the browser
> process...  Though I'm not really sure.  I'm not entirely sure exactly what
else
> needs to be cleaned up.

lgtm  like this, or it's also fine to do the more involved solution, just as
long as the chromeos bug is fixed

[mmenke, 2014-05-28 01:40:23]

The CQ bit was checked by mmenke@chromium.org

[commit-bot: I haz the power, 2014-05-28 01:42:23]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/mmenke@chromium.org/301753002/1

[commit-bot: I haz the power, 2014-05-28 03:31:15]

FYI, CQ is re-trying this CL (attempt #1).
The failing builders are:
  mac_gpu on tryserver.chromium.gpu
(http://build.chromium.org/p/tryserver.chromium.gpu/builders/mac_gpu/builds/9758)

[commit-bot: I haz the power, 2014-05-28 03:58:02]

Change committed as 273149