Semi-Realtime MIDI - separation of voices [WIP] #2794
Conversation
shoogle
force-pushed
the
semi-realtime-midi-voices
branch
2 times, most recently
from
25ca630
to
1fbbfb3
Compare
| seg = setNoteRest(seg, track, NoteVal(), Fraction::fromTicks(rchord->tick() - seg->tick()), Direction::AUTO); // This *appears* to work in all voices, but see FIXME below. | ||
| while (seg->tick() < rchord->tick()) // might encounter segments from other voices before the one we want | ||
| seg = seg->next(Segment::Type::ChordRest); // FIXME: in higher voices this sometimes does not find the ChordRest we just added 2 lines above with setNoteRest() | ||
| } |
There was a problem hiding this comment.
This is where the problem occurs.
I use setNoteRest to add a rest so that I can guarantee there will be a ChordRest at the tick immediately after it. I then try to move to the end of the rest with seg->next(Segment::Type::ChordRest). This works fine in Voice 1 and usually in higher voices too, but sometimes in higher voices it moves too far and skips past the end of the rest.
shoogle
force-pushed
the
semi-realtime-midi-voices
branch
from
1fbbfb3
to
cacf5f0
Compare
|
What's the current status? I still have the PR marked as WIP. |
|
Status is unchanged at present. TLDR: ChordRests and ties are being added to the score but are subsequently ignored. Does this mean anything to you? My code separates the voices and adds the notes to ChordRests in each voice, but for some reason the new ChordRests do not appear to be "visible" to the other voices. Perhaps I need to do something like call |
|
For any playback changes probably score->setPlaylistDirty() has to be called to force the midi event list to be recreated on next play.. |
|
This PR now needs a rebase. |
|
@shoogle, would you like to rebase this PR or should we better archive it for now? |
dmitrio95
added
the
archived
|
Archiving this for now, if it is planned to be updated feel free to reopen this PR. |
Definitely a work-in-progress at the moment. The voice recalculation works perfectly, but there are bugs when it comes to entering the newly calculated voice into the score. The pre-existing
setNoteRest()function appears to have problems with higher voices. It seems like ChordRests are being created but then ignored, so when the chord looks for the next ChordRest the new ones are missed and things get corrupted.Voice separation works some of the time, but crashes consistently with certain inputs, such as when playing short low notes at the same time as long high notes. For example, playing an E whole-note while playing C quarters causes a crash, but playing E quarters against a C whole note is fine.