Remove BrowserThread-specific implementation of MessageLoopProxy.

Remove BrowserThread-specific implementation of MessageLoopProxy.

The plan is to merge MessageLoopProxy with MessageLoopProxyImpl.
We don't need BrowserThreadMessageLoopProxy because all MessageLoop
instances have proxy anyway.

[jam, 2012-05-02 14:49:13]

lgtm

[Sergey Ulanov, 2012-05-02 18:16:23]

John, thanks for review,
With this change chrome now DCHECKs with the following callstack when started:
[25753:25753:1138772815284:FATAL:browser_thread_impl.cc(358)] Check failed:
thread. 
Backtrace:
	base::debug::StackTrace::StackTrace() [0x7f5f16a2df5e]
	logging::LogMessage::~LogMessage() [0x7f5f16a65bd2]
	content::BrowserThread::GetMessageLoopProxyForThread() [0x7f5f18be842c]
	PrefService::CreatePrefService() [0x7f5f1af40ee0]
	BrowserProcessImpl::CreateLocalState() [0x7f5f1bb3ffa8]
	BrowserProcessImpl::local_state() [0x7f5f1bb3fe56]
	(anonymous namespace)::InitializeLocalState() [0x7f5f1b955718]
	ChromeBrowserMainParts::PreCreateThreadsImpl() [0x7f5f1b954a8c]
	ChromeBrowserMainParts::PreCreateThreads() [0x7f5f1b95439f]
	content::BrowserMainLoop::CreateThreads() [0x7f5f18bde853]
	(anonymous namespace)::BrowserMainRunnerImpl::Initialize() [0x7f5f18be1c9e]
	BrowserMain() [0x7f5f18bdd6fa]
	(anonymous namespace)::RunNamedProcessTypeMain() [0x7f5f18bb560a]
	(anonymous namespace)::ContentMainRunnerImpl::Run() [0x7f5f18bb52cd]
	content::ContentMain() [0x7f5f18bb4844]

I.e. it looks like we are trying to initialize JsonPrefStore with the file
thread before that thread is started. What do you think would be the right fix
for this issue? Why do we need to initialize local state before browser threads
are started?

[Sergey Ulanov, 2012-05-02 21:27:04]

Looks like the problem below can be fixed by switching JsonPrefStore to worker
pool - http://codereview.chromium.org/10344007/ .
There was another problem on shutdown - DeleteSoon() wasn't working properly
when called after threads are destroyed. Fixed it in patch set 3.

On 2012/05/02 18:16:23, sergeyu wrote:
> John, thanks for review,
> With this change chrome now DCHECKs with the following callstack when started:
> [25753:25753:1138772815284:FATAL:browser_thread_impl.cc(358)] Check failed:
> thread. 
> Backtrace:
> 	base::debug::StackTrace::StackTrace() [0x7f5f16a2df5e]
> 	logging::LogMessage::~LogMessage() [0x7f5f16a65bd2]
> 	content::BrowserThread::GetMessageLoopProxyForThread() [0x7f5f18be842c]
> 	PrefService::CreatePrefService() [0x7f5f1af40ee0]
> 	BrowserProcessImpl::CreateLocalState() [0x7f5f1bb3ffa8]
> 	BrowserProcessImpl::local_state() [0x7f5f1bb3fe56]
> 	(anonymous namespace)::InitializeLocalState() [0x7f5f1b955718]
> 	ChromeBrowserMainParts::PreCreateThreadsImpl() [0x7f5f1b954a8c]
> 	ChromeBrowserMainParts::PreCreateThreads() [0x7f5f1b95439f]
> 	content::BrowserMainLoop::CreateThreads() [0x7f5f18bde853]
> 	(anonymous namespace)::BrowserMainRunnerImpl::Initialize() [0x7f5f18be1c9e]
> 	BrowserMain() [0x7f5f18bdd6fa]
> 	(anonymous namespace)::RunNamedProcessTypeMain() [0x7f5f18bb560a]
> 	(anonymous namespace)::ContentMainRunnerImpl::Run() [0x7f5f18bb52cd]
> 	content::ContentMain() [0x7f5f18bb4844]
> 
> I.e. it looks like we are trying to initialize JsonPrefStore with the file
> thread before that thread is started. What do you think would be the right fix
> for this issue? Why do we need to initialize local state before browser
threads
> are started?

[jam, 2012-05-06 06:51:10]

On 2012/05/02 21:27:04, sergeyu wrote:
> Looks like the problem below can be fixed by switching JsonPrefStore to worker
> pool - http://codereview.chromium.org/10344007/ .
> There was another problem on shutdown - DeleteSoon() wasn't working properly
> when called after threads are destroyed. Fixed it in patch set 3.

It would be preferable to use a mechanism to automatically release the lock,
instead of having to remember to do so at each early return. You can use
AutoLock in a scoped_ptr.

> 
> On 2012/05/02 18:16:23, sergeyu wrote:
> > John, thanks for review,
> > With this change chrome now DCHECKs with the following callstack when
started:
> > [25753:25753:1138772815284:FATAL:browser_thread_impl.cc(358)] Check failed:
> > thread. 
> > Backtrace:
> > 	base::debug::StackTrace::StackTrace() [0x7f5f16a2df5e]
> > 	logging::LogMessage::~LogMessage() [0x7f5f16a65bd2]
> > 	content::BrowserThread::GetMessageLoopProxyForThread() [0x7f5f18be842c]
> > 	PrefService::CreatePrefService() [0x7f5f1af40ee0]
> > 	BrowserProcessImpl::CreateLocalState() [0x7f5f1bb3ffa8]
> > 	BrowserProcessImpl::local_state() [0x7f5f1bb3fe56]
> > 	(anonymous namespace)::InitializeLocalState() [0x7f5f1b955718]
> > 	ChromeBrowserMainParts::PreCreateThreadsImpl() [0x7f5f1b954a8c]
> > 	ChromeBrowserMainParts::PreCreateThreads() [0x7f5f1b95439f]
> > 	content::BrowserMainLoop::CreateThreads() [0x7f5f18bde853]
> > 	(anonymous namespace)::BrowserMainRunnerImpl::Initialize() [0x7f5f18be1c9e]
> > 	BrowserMain() [0x7f5f18bdd6fa]
> > 	(anonymous namespace)::RunNamedProcessTypeMain() [0x7f5f18bb560a]
> > 	(anonymous namespace)::ContentMainRunnerImpl::Run() [0x7f5f18bb52cd]
> > 	content::ContentMain() [0x7f5f18bb4844]
> > 
> > I.e. it looks like we are trying to initialize JsonPrefStore with the file
> > thread before that thread is started. What do you think would be the right
fix
> > for this issue? Why do we need to initialize local state before browser
> threads
> > are started?