config_map.erase() is only called in ExportHandler::finish_timespan().
When an export fails (throw) or is aborted, the export-handler's
config remains as is and the next export will run it again.
The export-handler is global, per session and
ExportHandler::add_export_config() only ever inserts or ignores insert.
This is in preparation to fix:
1) export to invalid path -> fail, error is thrown
2) correct path -> new config is inserted in the map
3) try to export again, first runs the not-completed export from (1)
-> constant errors.
Some MIDNAM files use Patch-Banks that do *not* have a common midi bank
(CC0, CC32) but the Patchbank is instead just a collection of patches.
There may be more than 127 patches in a patch-bank and each of those
may be in a different midi-bank (e.g. Roland SC-88 Pro).
Midnams like these are now properly supported via dropdowns (in the
patch add/modify dialog and timeline), but ignored by the grid
(patch-widget). More work is needed.
This fixes an issue with pitch-shift using the most recently used
time-stretch algorithm, which may be "resample.." (effective NO-OP).
Also leave a ToDo note regarding rubberband crispness levels for later.
Leave the user in control while the slider is being dragged.
Previously there was a feedback loop:
User-drags slider -> value changes -> value is rounded
-> slider is updated with rounded value (while the user still drags)
Explicit cast to float (rangestep is unsigned int) when calculating
interface steps.
Also prefer to round to nearest value-point (rather than round-down).
This forces liblilv to collect external files in the template-folder
and symlink to the original file instead of re-using the external-file
folder of the session that was used to create the template.
This is in preparation to replace the Ringbuffer and avoid over-using
"buf" as variable (it's already used for vectors, as parameter name and
Audiobuffers.
currently this is only used to resolve midi events for plugins
(this is conceptually not correct, note offs should be resolved by
the disk-reader only), but it calls into all processors now (incl
disk-reader if present), which is handy (e.g. flush delaylins)
