[Musickit-developer] PLEASE TEST MUSICKIT (Changes in memory allocation retain/release)

[Stephen B. <st...@br...>]

Hi all,

As you will have noticed I've just submitted a substantial number of changes 
to MKPart and MKNote, along with some supplementary changes to associated 
objects.

The changes are to do with memory management. I have gone over the MKNote and 
MKPart classes very carefully to make the functions and methods conform to 
the "official" OpenStep memory management guidelines (for a summary, see my 
e-mail on Tuesday, "Retains and MKNotes (important)").

There are some internal methods in MKNote to do with hashtables that are not 
yet examined, but almost all of the rest of those 2 classes has been.

I have done some basic tests on the classes, but have not tested them 
thoroughly.

Could I make a request: could as many of you as possible update from CVS, 
then compile and run any MK apps or programs you have, just to get a good 
idea of how solid the classes are, post-changes?

I don't want to get the new release underway without a bit more confirmation 
that I have not broken things too badly :-)

Here's what I have tested so far:
create note
create part
create score
set note parameters: MK_freq, custom string param, timetag, duration
add note to part, part to score
add new partinfo note to part, with MK_synthPatch parameter
save as scorefile
save as midifile (now no longer crashes on GNUstep!)
release note, part and score

create score
load score from saved scorefile (only tested with output from 1st test)
call up 1st part
call up note from part
log [note description]  (all parameters logged correctly)
release score.

ScorePlayer.app (MacOSX):
load "BachFugue" scorefile
"play" (though of course there's no audio since it was a DSP56k based 
scorefile).
[but at least it does not crash]

Cheers,
Stephen Brandon
st...@br...