Fixing inconsistency with behavior of `ClearGettingPorts`.

webrtc/p2p/client/basicportallocator.cc

Index: webrtc/p2p/client/basicportallocator.cc
diff --git a/webrtc/p2p/client/basicportallocator.cc b/webrtc/p2p/client/basicportallocator.cc
index fabbc429e5292a941ec726f0571cbb113f01be47..d38fd50dba834cf3c0e2ab39bd225b5503eb2ff0 100644
--- a/webrtc/p2p/client/basicportallocator.cc
+++ b/webrtc/p2p/client/basicportallocator.cc
@@ -253,7 +253,6 @@ void BasicPortAllocatorSession::StartGettingPorts() {
 
 void BasicPortAllocatorSession::StopGettingPorts() {
   ASSERT(rtc::Thread::Current() == network_thread_);
-  network_thread_->Post(RTC_FROM_HERE, this, MSG_CONFIG_STOP);
   ClearGettingPorts();
   // Note: this must be called after ClearGettingPorts because both may set the
   // session state and we should set the state to STOPPED.
@@ -266,6 +265,7 @@ void BasicPortAllocatorSession::ClearGettingPorts() {
   for (uint32_t i = 0; i < sequences_.size(); ++i) {
     sequences_[i]->Stop();
   }
+  network_thread_->Post(RTC_FROM_HERE, this, MSG_CONFIG_STOP);

[pthatcher1, 2016/07/22 18:11:22]

ClearGettingPorts sends MSG_CONFIG_STOP?  That's confusing.   Should we call it
MSG_CONFIG_CLEAR?  Or should we call this
StopGettingPortsUntilContinualGatheringTriggered?

[Taylor Brandstetter, 2016/07/22 19:58:42]

In my opinion the issue is more that "ClearGettingPorts" is not ideally named.
"Stop/Clear" are actually more like "PermanentlyStop/Stop". But it would be
difficult to change those names now since they're pure virtual and have
overrides in downstream dependencies.

By the way: Originally "ClearGettingPorts" was
"StopGettingPorts(keep_session_running=true)" and it *did* post MSG_CONFIG_STOP:
https://codereview.webrtc.org/1359363003/diff/120001/webrtc/p2p/client/basicp...

But I think you requested 2 separate methods and MSG_CONFIG_STOP was
unintentially lost as part of that refactoring.

   state_ = SessionState::CLEARED;
 }