Handle disconnection events properly for unauthenicated clients.

Handle disconnection events properly for unauthenticated clients.

Previously when receiving disconnection event the host was stopping
screen recorder even when there are other active sessions. Obviously
this is wrong.

BUG=128643
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=138185

[Sergey Ulanov, 2012-05-18 04:22:58]

I'll be OOO tomorrow, so please CQ if it looks good.

[Wez, 2012-05-18 20:20:57]

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
File remoting/host/chromoting_host.cc (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
remoting/host/chromoting_host.cc:255:
recorder_->RemoveConnection(client->connection());
nit: It should be safe to RemoveConnection() for a connection that was never
added in the first place, so do you need this to be inside the
is_authenticated() block?

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
remoting/host/chromoting_host.cc:259:
OnClientDisconnected(client->client_jid()));
This changes the semantics of OnClientDisconnected such that you won't see it
for clients that never authenticate; is that what you intend?

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_session.h
File remoting/host/client_session.h (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_sessio...
remoting/host/client_session.h:112: bool is_authenticated() { return
is_authenticated_;  }
You can check |auth_input_filter_.get_stub()| here and avoid having a separate
|is_authenticated_| flag.

[Sergey Ulanov, 2012-05-21 18:20:13]

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
File remoting/host/chromoting_host.cc (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
remoting/host/chromoting_host.cc:255:
recorder_->RemoveConnection(client->connection());
On 2012/05/18 20:20:57, Wez wrote:
> nit: It should be safe to RemoveConnection() for a connection that was never
> added in the first place, so do you need this to be inside the
> is_authenticated() block?

It doesn't make sense to remove the connection if we know it was never added.

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
remoting/host/chromoting_host.cc:259:
OnClientDisconnected(client->client_jid()));
On 2012/05/18 20:20:57, Wez wrote:
> This changes the semantics of OnClientDisconnected such that you won't see it
> for clients that never authenticate; is that what you intend?

Yes. I think for such clients the observers will see only the OnAccessDenied()
event.

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_session.h
File remoting/host/client_session.h (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_sessio...
remoting/host/client_session.h:112: bool is_authenticated() { return
is_authenticated_;  }
On 2012/05/18 20:20:57, Wez wrote:
> You can check |auth_input_filter_.get_stub()| here and avoid having a separate
> |is_authenticated_| flag.

OnConnectionClosed() resets auth_input_filter_ and we need this method to return
correct value when OnSessionClosed() is called.

[Wez, 2012-05-21 20:33:50]

LGTM w/ some comments below.

nit: The description has a typo.

I think this CL highlights some improvements we should make to the
ChromotingHost/ClientSession layout, too.

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
File remoting/host/chromoting_host.cc (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
remoting/host/chromoting_host.cc:255:
recorder_->RemoveConnection(client->connection());
On 2012/05/21 18:20:13, sergeyu wrote:
> On 2012/05/18 20:20:57, Wez wrote:
> > nit: It should be safe to RemoveConnection() for a connection that was never
> > added in the first place, so do you need this to be inside the
> > is_authenticated() block?
> 
> It doesn't make sense to remove the connection if we know it was never added.

True, but you're testing whether |recorder_| is set, so the current arrangement
is fragile in the face of changes to recorder & client lifetime, e.g. if we add
support for multiple clients then this code will break.  Always calling
RemoveConnection() makes the code more robust to those sorts of change.

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_session.h
File remoting/host/client_session.h (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_sessio...
remoting/host/client_session.h:112: bool is_authenticated() { return
is_authenticated_;  }
On 2012/05/21 18:20:13, sergeyu wrote:
> On 2012/05/18 20:20:57, Wez wrote:
> > You can check |auth_input_filter_.get_stub()| here and avoid having a
separate
> > |is_authenticated_| flag.
> 
> OnConnectionClosed() resets auth_input_filter_ and we need this method to
return
> correct value when OnSessionClosed() is called.

Hmmm, yes; the teardown logic in ChromotingHost really belongs in ClientSession
- ClientSession should just tell the host when it needs the ScreenRecorder, and
when it's done with it.

[Sergey Ulanov, 2012-05-21 22:27:54]

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
File remoting/host/chromoting_host.cc (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/chromoting_ho...
remoting/host/chromoting_host.cc:255:
recorder_->RemoveConnection(client->connection());
On 2012/05/21 20:33:50, Wez wrote:
> On 2012/05/21 18:20:13, sergeyu wrote:
> > On 2012/05/18 20:20:57, Wez wrote:
> > > nit: It should be safe to RemoveConnection() for a connection that was
never
> > > added in the first place, so do you need this to be inside the
> > > is_authenticated() block?
> > 
> > It doesn't make sense to remove the connection if we know it was never
added.
> 
> True, but you're testing whether |recorder_| is set, so the current
arrangement
> is fragile in the face of changes to recorder & client lifetime, e.g. if we
add
> support for multiple clients then this code will break.  Always calling
> RemoveConnection() makes the code more robust to those sorts of change.

Done.

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_session.h
File remoting/host/client_session.h (right):

http://codereview.chromium.org/10409021/diff/5001/remoting/host/client_sessio...
remoting/host/client_session.h:112: bool is_authenticated() { return
is_authenticated_;  }
On 2012/05/21 20:33:50, Wez wrote:
> On 2012/05/21 18:20:13, sergeyu wrote:
> > On 2012/05/18 20:20:57, Wez wrote:
> > > You can check |auth_input_filter_.get_stub()| here and avoid having a
> separate
> > > |is_authenticated_| flag.
> > 
> > OnConnectionClosed() resets auth_input_filter_ and we need this method to
> return
> > correct value when OnSessionClosed() is called.
> 
> Hmmm, yes; the teardown logic in ChromotingHost really belongs in
ClientSession
> - ClientSession should just tell the host when it needs the ScreenRecorder,
and
> when it's done with it.

Added TODO.

[commit-bot: I haz the power, 2012-05-21 22:29:10]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/10409021/9003