add success and error callbacks to setLocalDescription and setRemoteDesc...

[fippo]

as discussed in issue #73.

[juberti]

I see what you meant before about reusing functions - I suggest adding some sort of webrtc.noop() and webrtc.error() (which calls console.error, at least for now) functions to adapter.js that we can use to simplify this code.

[fippo]

I just moved the errors two adapter.js -- setRemoteDescriptionFailure and setLocalDescriptionFailure.
I also noticed that error callbacks for createOffer and createAnswer are missing sometimes. I fixed this in one instance, added createOfferFailure and createAnswerFailure but haven't grepped yet.

(style) question about the success callbacks, e.g.
https://github.com/GoogleChrome/webrtc/blob/master/samples/web/content/datachannel/js/main.js#L118 call setLocalDescription-setRemoteDescription-createAnswer sequentially, but shouldn't setRemoteDescription be called in the success callback of setLocalDescription and likewise createAnswer in the callback of setRemoteDescription? videopipe.js does the latter at least.

Arguably, errors are very unlikely when two peerconnection objects on the same page talk to each outher.

[fippo]

moved the success callbacks to adapter.js as well.

[juberti]

I don't like the idea of dumping functions called setRDFailure/setLDFailure into the global namespace, as they might collide with other real names in application code. It also isn't clear what these functions do from their names - hence I prefer the simpler and more explicit webrtc.noop and webrtc.error suggested above.

[fippo]

mh... but dumping a global webrtc object into adapter.js isn't ideal either.
How about putting this into a common.js which is shared by all demos but not pulled in by anyone that just wants adapter.js as a shim that hides browser differences?

[juberti]

I hate to make a new file for just these little nibs of code. Are you worried about a webrtc object stomping on apps' webrtc objects? We could just pick a different name.

[fippo]

Right, it doesn't make sense for these little nibs of code. We could bold and see if anyone complains about this webrtc object :-)

[fippo]

ping. Apparently FF32 has a regression (https://bugzil.la/1063971) which causes it not to work for at least the data channel example (https://groups.google.com/forum/#!topic/discuss-webrtc/u5NHYR-4rpc) when the mandatory callbacks are missing.

[juberti]

Add a comment about that this does

[juberti]

would prefer to keep these args on the same line, as long line is < 80 chars (here and below)

[fippo]

Hitting 82 chars in some places... reduce it to a single line where possible and two lines elsewhere or two lines (putting webrtc.noop and webrtc.error on the second) everywhere?

(no preference, anything that makes jshint happy is fine with me :-)

[fippo]

too many merge conflicts... and it looks this has been taken care of already in the meantime.