Most significant changes:
- Factor out FileSource from AudioFileSource, use for SMFSource too
- Explicitly pass embedded rather than mysterious name mangling or whatever
- Destroy a ton of duplicated or very-nearly-duplicated code
- Clean up and document all that weird source stuff in session.cc
git-svn-id: svn://localhost/ardour2/branches/3.0@4609 d708f5d6-7413-0410-9779-e7cbd77b26cf
Re-addition of this sort of thing is now officially punishable by death ;)
git-svn-id: svn://localhost/ardour2/branches/3.0@4603 d708f5d6-7413-0410-9779-e7cbd77b26cf
Add IdentityConverter for when no conversion is actually needed.
git-svn-id: svn://localhost/ardour2/branches/3.0@4595 d708f5d6-7413-0410-9779-e7cbd77b26cf
This has 3 main benefits:
- All conversion code is in one place (less duplication, potential bugs)
- The conversion method can be passed to things that are ignorant
of the actual time units involved, information required, etc.
(In the future it would be nice to have user selectable tempo/frame time)
- It should be relatively simple now to support tempo changes part-way
through a MIDI region (at least architecturally speaking)
git-svn-id: svn://localhost/ardour2/branches/3.0@4594 d708f5d6-7413-0410-9779-e7cbd77b26cf
Ardour should now be able to more or less tolerate crazy incoming MIDI (except for SYSEX).
git-svn-id: svn://localhost/ardour2/branches/3.0@4592 d708f5d6-7413-0410-9779-e7cbd77b26cf
Fixes crash bug when seeking back and forth from start to end of session.
Not sure about other things, but it makes a lot more sense now anyway...
git-svn-id: svn://localhost/ardour2/branches/3.0@4590 d708f5d6-7413-0410-9779-e7cbd77b26cf
Correctly layer automation regions to match the stacking of their 'real' counterparts.
git-svn-id: svn://localhost/ardour2/branches/3.0@4585 d708f5d6-7413-0410-9779-e7cbd77b26cf
The time stamp of an event is now always tempo, from file to model and
back again. Frame time is only relevant at playback or recording time,
in the audio thread (MidiModel and MidiBuffer).
I think perhaps we don't need to change the actual time from double (which is
convenient for math), it is the time base conversion that caused problems.
Using a correct equality comparison (i.e. not == which is not correct for
floating point) should probably make the undo issues go away, in 99.99% of
cases anyway.
There's almost certainly some regressions in here somewhere, but they do not
seem to be time related. The bugs I'm hitting in testing are old ones that
seem unrelated now, so it's checkpoint time.
This sets us up for fancy things like tempo map import and tempo/meter changes
halfway through MIDI regions, but for now it's still assumed that the tempo
at the start of the region is valid for the duration of the entire region.
git-svn-id: svn://localhost/ardour2/branches/3.0@4582 d708f5d6-7413-0410-9779-e7cbd77b26cf
One thing left, this weird "end-marker-is-free" in <Config>. Is this really a config option? Anyone?
git-svn-id: svn://localhost/ardour2/branches/3.0@4577 d708f5d6-7413-0410-9779-e7cbd77b26cf
Use iterator interface of Sequence to read events in a MIDISource rather than Sequence::read, avoiding timestamp confusion.
Disable no longer useful Sequence::read.
git-svn-id: svn://localhost/ardour2/branches/3.0@4570 d708f5d6-7413-0410-9779-e7cbd77b26cf
- Remove tab width assumption in struct members that looked atrocious at ts=4
- Line up function names for easier/faster readability
git-svn-id: svn://localhost/ardour2/branches/3.0@4563 d708f5d6-7413-0410-9779-e7cbd77b26cf
