Re: [Lmuse-developer] playpos problem with wave track

[Werner S. <ws...@se...>]

On Tuesday 13 January 2004 17:02, Robert Jonsson wrote:
> Hi,
>
> > you are right, it will last some time util it is ready for a release.
> > Many things have changed and the next *.med file version is incompatible.
> > Im looking into make it at least somewhat backwards compatible.
>
> Changing the fileformat seems inevitable. Perhaps it would be possible to
> just make a conversion utility if one knows what tags should be present and
> which to filter out.
> A suppose there are no formal documents (DTDs or similar) which specify the
> formats?

there is no formal description. I've changed the serial number so it should be
possible to at least read the songdata of pre 0.7 MusE files (we can
implement an internal converter).


>
> > Some features are only partly implemented. For example the midi/audio
> > mixer integration. Timing feels stable and the new software synthesizer
> > integration is also more stable and has better performance.
> > New features to play with is the rendering of midi files to wave files
> > with software syntis in JACK "freewheel mode" - faster than realtime.
>
> I'll be trying to build it one of these nights, will be interesting. I'll
> make sure to keep a journal of the bugs I find.
>
> > I'm currently testing midi play, import and export.
> > Last bug i found: when playing long sysex messages at song start jack
> > zombifies MusE. I'm not sure what to do about this.
>
> Basically what you are saying is that the sysex messages are so big that
> MusE is not able to complete within it's timeout?

yes, and its worse. I fixed the logic, so no more JACK zombification. But
the sysex needs some seconds to transfer, so the whole song timing is 
destroyed.

>
> I don't know alot about how these things work... I guess the sysex messages
> are somekind of setup of the synth that happens when the song is
> initialized. Could it not be done before ROLL then?
> Or is it that MusE does not know what type of data it is so it can't know
> what to do in advance...
> Sorry, thinking out aloud.

I believe that all sysex messages at tick 0 should be sorted out and treated 
as
"initialisation data". This leaves the question where to edit them and when
to play them. Currently "initialization data" like GmOn GSOn or XgOn are send
to the midi device when seeking to tick zero and explicitly when pressing
Midi->DeviceInit.
Another approach whould be to create a special init measure on midi import,
collect all init data into this measure and move the rest of the song one
(ore more) measure(s) to the right. Maybe MusE can calculate how many init 
measures 
are needed to satisfy all timing requirements.


Werner