Merge lp://qastaging/~mixxxdevelopers/mixxx/features_cuefile into lp://qastaging/~mixxxdevelopers/mixxx/trunk

Hey Phil,

Really sorry for taking so long to get to this! This all looks nice, I just have some general issues with lack of n-deck support and the duplication between EngineShoutcast and EngineRecord:

- The metadata recording is hard coded to work with only 2 decks. I'd really rather not add more 2-deck code if we don't have to. Can you use the [Master],num_decks control (alternatively, thread the PlayerManager into EngineRecord)? Every deck has an "orientation" control which indicates the side of the crossfader it is attached to. You can use this in conjunction with its volume/pregain faders to figure out if they are "hearable". PlayerInfo is N-Deck ready, so that shouldn't be a problem.

- Assuming you'll change this given the above comment anyway, but ControlObject::getControl shouldn't be used in code that gets called on a schedule. As written it'll be doing 5 locks of the global ControlObject static mutex per call (plus 5 hash lookups) to getActiveTracks() when it could be written just a little more verbosely to get 0 per call.

- There's now a fair amount of duplicated code between EngineShoutcast and EngineRecord. Can you pull the current-metadata-detection code out into a common class that they both use? You could even build it into the PlayerInfo singleton class since that already exists.

- Minor nit: EngineRecord::m_pMetaDataLife is an int so its prefix should be m_iMetaDataFile

- Maybe pull 5000 out into a constant? UPDATE_RATE_MS or something. also add a comment about what it is

thanks,
RJ

Hey Philip,

there's a singleton PlayerInfo class that EngineShoutcast uses to get track information per deck. Maybe you can extract metadata recording in there. This would be of great benefit for the library history feature, too.

Right now I am moving the 'Current Track' functionality into PlayerInfo.

I was planning on using a timestamp to make sure it updates at most once per X seconds. If another call is made within that timeframe it returns the last answer it made.

I moved all the functionality into PlayefInfo and refactored EngineRecord and EngineShoutCast to use it.

Awesome -- looks good for merging. I found these two issues:

1) TrackInfoObject::getId() only works for Mixxx library tracks. It won't work for tracks loaded from Rhythmbox, iTunes, Traktor, or Browse tracks, since all their IDs are -1. This might change in the future, but the best way is to probably just compare the TrackPointers directly. You're guaranteed that for tracks in the Mixxx library, there is only one TrackInfoObject* that exists for it, so QSharedPointer::operator= will work to tell if it's the same track. RB, iTunes, Traktor and Browse don't work this way, but at least they could be made to work that way when/if they ever keep track of the TIO's that they create. I would change all uses of getId() to compare the TrackPointers directly.

2) I tried out the cuefile recording and it didn't pick up 2 out of the 4 tracks I played. I think the problem is the math in PlayerInfo::getCurrentPlayingDeck. All the volume values are integers, while the engine volume values are 0.0 to 1.0 and the crossfader is -1.0 to 1.0, so these will get rounded off in integer conversion. Also, I think the xfvol calculation is slightly off since the orientation controls have non-intuitive values: left is 0, center is 1, right is 2.

This value for the xfval calculation worked for me (all 6 tracks I played got recorded in the correct order):

xfvol = 1 - 0.5 * fabs(m_listCOOrientation[chan]->get() - 1.0 - m_COxfader->get());

It maps tracks opposite from each other (e.g. oriented right crossfader left and vice-versa) to 0.0, and tracks that are exactly aligned w/ the crossfader (left/left, center/center, or right/right) to 1.0.