Fix UdpPacketSocket to handle callbacks after the socket is destroyed.

Fix UdpPacketSocket to handle callbacks after the socket is destroyed.

PPB_UDPSocket_Private doesn't make guarantees about callbacks not being called
after Close() - we may still get some callbacks asynchronously with error codes
different from PP_ERROR_ABORTED. UdpPacketSocket wasn't handling this case
properly.

BUG=143839
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=153972

[Wez, 2012-08-29 18:03:36]

I do like the fact that using WeakPtrFactory deals with this in a type safe way,
but the PPAPI way to do this is to use a CompletionCallbackFactory - have you
considered using that?

Is the PPB_UdpSocket_Private behaviour by-design?  PPAPI doesn't usually
dispatch callbacks once resources have been torn-down, I thought?

[Sergey Ulanov, 2012-08-29 19:37:18]

On 2012/08/29 18:03:36, Wez wrote:
> I do like the fact that using WeakPtrFactory deals with this in a type safe
way,
> but the PPAPI way to do this is to use a CompletionCallbackFactory - have you
> considered using that?

I think these two approaches are essentially the same in terms of efficiency,
but I prefer one with WeakPtrFactory because it's more consistent with the rest
of the code.

> Is the PPB_UdpSocket_Private behaviour by-design?  PPAPI doesn't usually
> dispatch callbacks once resources have been torn-down, I thought?

I think it's by-design. The Close() method documentation just says that pending
callbacks are aborted. PPAPI guarantees that each callback is called exactly
once. Even after resource is destroyed it still has to call the callback.
Normally the callbacks are aborted and called with PP_ERROR_ABORTED, but it's
not always possible. E.g. in this particular bug the resource is already
destroyed - so SendTo() has to return PP_ERROR_BADRESOURCE and it must do it
asynchronously because the callback is marked as required, so it posts the
callback on the current thread
(http://code.google.com/searchframe#OAMlx_jo-ck/src/ppapi/thunk/enter.cc&l=149).
Because resource doesn't exist anymore it's not possible to associate the posted
async callback invocation and the corresponding Close() call.

[Wez, 2012-08-29 20:06:47]

lgtm

[commit-bot: I haz the power, 2012-08-29 20:16:55]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/sergeyu@chromium.org/10900028/1