Fixing some issues with ICE restart signaling.

Fixing some issues with ICE restart signaling.

In summary, there were some cases where we were changing the ICE
ufrag/password when we shouldn't have been, and some cases where we
weren't changing them when we should have been.

[Taylor Brandstetter, 2015-11-18 22:31:58]

Description was changed from

==========
Fixing some issues with ICE restart signaling.

Basically, there are some situations where we were changing the ICE
ufrag/password when we shouldn't have been, and some cases where we
weren't changing them when we should have been.
==========

to

==========
Fixing some issues with ICE restart signaling.

In summary, there were some cases where we were changing the ICE
ufrag/password when we shouldn't have been, and some cases where we
weren't changing them when we should have been.
==========

[Taylor Brandstetter, 2015-11-18 22:40:17]

deadbeef@webrtc.org changed reviewers:
+ pthatcher@webrtc.org

[Taylor Brandstetter, 2015-11-18 22:40:17]

I was about to work on the "SetConfiguration causes the next CreateOffer to
trigger an ICE restart" logic, but in looking at the code I noticed some bugs
with our existing logic:

1. If the offerer changes the ufrag/password on just ONE m-line, the answer will
change the ufrag/password on ALL m-lines.
2. If the offerer changes the ufrag/password, and CreateAnswer is called twice,
the second call will return the old ufrag/password.
3. Very corner-casey, but if there are different ufrags/passwords for each
m-line, and BUNDLE is enabled, CreateAnswer will force the ufrags/passwords to
match. Meaning, like with issue #1, the answerer is changing the ufrag/password
without the offerer doing the same.

This CL fixes those bugs and adds some unit tests. And will make it easier to
properly implement SetConfiguration.

https://codereview.webrtc.org/1453813005/diff/1/talk/session/media/mediasessi...
File talk/session/media/mediasession.cc (left):

https://codereview.webrtc.org/1453813005/diff/1/talk/session/media/mediasessi...
talk/session/media/mediasession.cc:1359: // as BUNDLE requires.
Actually, BUNDLE doesn't require this, as far as I can tell. And this code
introduces bugs sometimes. So I just removed it. Didn't seem to break any unit
tests.

[Taylor Brandstetter, 2015-12-10 21:21:36]

Actually, hold off on reviewing this (if you ever get to it in your queue). I
have another CL that will conflict with it, so I need to do some refactoring.

[pthatcher1, 2015-12-15 08:00:13]

Do you want me to review this one now, or wait until we get the "3x transport
options" thing resolved in the other CL?

[Taylor Brandstetter, 2015-12-15 19:16:48]

On 2015/12/15 08:00:13, pthatcher1 wrote:
> Do you want me to review this one now, or wait until we get the "3x transport
> options" thing resolved in the other CL?

It would be better to wait. I'll have to mostly redo this CL due to the 3x
transport_options, and there are other problems with it that I realized later.