Experiment with redirecting all BrowserThreads (but UI/IO) to TaskScheduler

content/public/test/test_browser_thread.cc

Index: content/public/test/test_browser_thread.cc
diff --git a/content/public/test/test_browser_thread.cc b/content/public/test/test_browser_thread.cc
index ae83f78f78aad1455f8a9f98aa6c4f003ef4499b..b5148bf46e4bd43f7cba08012cd8a968ffd8b888 100644
--- a/content/public/test/test_browser_thread.cc
+++ b/content/public/test/test_browser_thread.cc
@@ -40,15 +40,32 @@ class TestBrowserThreadImpl : public BrowserThreadImpl {
 };
 
 TestBrowserThread::TestBrowserThread(BrowserThread::ID identifier)
-    : impl_(new TestBrowserThreadImpl(identifier)) {
-}
+    : impl_(new TestBrowserThreadImpl(identifier)), identifier_(identifier) {}
 
 TestBrowserThread::TestBrowserThread(BrowserThread::ID identifier,
                                      base::MessageLoop* message_loop)
-    : impl_(new TestBrowserThreadImpl(identifier, message_loop)) {}
+    : impl_(new TestBrowserThreadImpl(identifier, message_loop)),
+      identifier_(identifier) {}
 
 TestBrowserThread::~TestBrowserThread() {
-  Stop();
+  // The upcoming BrowserThreadImpl::ResetGlobalsForTesting() call requires that
+  // |impl_| have triggered the shutdown phase for its BrowserThread::ID. This
+  // either happens when the thread is stopped (if real) or destroyed (when fake
+  // -- i.e. using an externally provided MessageLoop).
+  impl_.reset();
+
+  // Resets BrowserThreadImpl globals binding |identifier_| to |impl_|. This is

[robliao, 2016/12/08 01:57:17]

How does the below set the binding from |identifier_| to |impl_|?

ResetGlobalForTesting looks like it just clears everything.

[gab, 2016/12/08 18:49:41]

I meant to say that it clears that very binding, reworded.

+  // fine since the underlying MessageLoop has already been flushed and deleted
+  // in Stop(). In the case of an externally provided MessageLoop however, this
+  // means that TaskRunners obtained through
+  // |BrowserThreadImpl::GetTaskRunnerForThread(identifier_)| will no longer
+  // recognize their BrowserThreadImpl for RunsTasksOnCurrentThread(). This
+  // happens most often when such verifications are made from
+  // MessageLoop::DestructionObservers. Callers that care to work around that
+  // should instead use this shutdown sequence: TestBrowserThread::Stop() ->

[robliao, 2016/12/08 01:57:17]

Nit: Might be easier to read this as a sequenced list.

[gab, 2016/12/08 18:49:41]

Done.

+  // ~MessageLoop() -> ~TestBrowserThread() (~TestBrowserThreadBundle() does
+  // this).
+  BrowserThreadImpl::ResetGlobalsForTesting(identifier_);
 }
 
 bool TestBrowserThread::Start() {