Now that Location signal(s) are only emitted when the
drag ended, the list can be kept in sync.
This also adds a further optimization to only update the
list once, even when multiple locations have been changed.
Section Markers only need to be updated when an actual
section-marker is modified OR flags change.
This also removes duplicate signal subscriptions and
caches sorted Location list when iterating over section markers.
This allows Ardour to re-open previously shown plugins,
and toggle externalUI.
This worked in the distant past, before Ardour 3.3 introduced
the Ardour Window Manager.
when the surround bus is removed, it (probably erroneously) causes
changes to the I/O routing for the auditioner, which will be
delivered to anyone who cares via an RCConfiguration ParameterChanged
signal. This will be emitted from a non-GUI thread, and so will not
be handled synchronously. The GUI thread will first handle the notification
that the surround bus was deleted, and that will include deleting the
MixerStrip for the surround bus. But another call_slot() request will
be pending to notify the now-deleted MixerStrip about the auditioner change.
Adding invalidator() calls ensures that the deletion of the MixerStrip removes
all of these pending call_slot() requests.
The box outlines a given region, the grab position may be earlier
than the region's position.
Previously this caused a crash (uncaught exception):
```gdb
__cxa_throw () at /lib/x86_64-linux-gnu/libstdc++.so.6
Temporal::timecnt_t::timecnt_t(Temporal::timecnt_t const&, Temporal::timepos_t const&) at ../libs/temporal/timeline.cc:79
RegionMoveDrag::setup_pointer_offset() at ../gtk2_ardour/editor_drag.cc:2226
```
deleting regionviews changed the region selection, but the regionview
ptr had not been removed from the RegionViewList. The selection change signal
caused an iteration over the list, and manipulated the deleted regionview.
The former was incorrectly implemented, and the latter has already been tested more
in real life.
We should likely remove ::remove_time also and use shift() there too, but that
requires testing negative shifts more broadly.
Editor::duplicate copies regions one at a time. This lead
to region layeres not being retained, particularly since
Playlist::duplicate() puts the duplicated layer at the top.
This fixes an issue where brining up a menu can directly activate an
action in the menu. Notably on macOS.
Nathan reports that this may also address #9515
Note: mac touchpads are not affected since right-click there is
effectively a "press and hold".
freetype on macOS does not scale fonts, font-scale and ui_scale
have to remain at 100% otherwise other UI elements change size
while the font does not.
The InstrumentSelector keeps a PluginPtr reference to all
instruments. By default Ardour does not destroy the editor
at exit, So 3 instances of Instrument Selector can remain
* TriggerClipPicker in the editor sidebar (owned by _editor)
* TriggerClipPicker on the TriggerPage (owned by trigger_page)
* SoundFileOmega (sfbrowser owned by Editor)
If a shared_ptr reference is retained, VST3 Module is not
unloaded at exit. This leads to issues with various plugins.
eg. Reason Rack does not terminate, Ardour hangs at exit.
When using "VS" the numeric control shows the "default speed"
when not rolling. Then when starting to roll default speed will
be the actual speed.
Previously the shuttle was not updated when `last_speed_displayed`
was the current speed. This dated back to before "VS" showed
default speed.
Previously [GtkApplicationDelegate applicationShouldTerminate:]
triggered both StartupFSM::queue_finish and ArdourUI::queue_finish.
Both of which registered an idle-callback to call ArdourUI::finish.
heap-use-after-free. Marker is deleted (and re-created), when
the tempo-map edit is aborted:
```
#0 0x7f77528ac017 in operator delete(void*) ../../../../src/libsanitizer/asan/asan_new_delete.cpp:160
#1 0x55f81062800d in TempoMarker::~TempoMarker() ../gtk2_ardour/marker.cc:794
#2 0x55f80ffb2fc0 in Editor::reset_tempo_marks() ../gtk2_ardour/editor_tempodisplay.cc:205
#3 0x55f80ffb2b19 in Editor::reset_metric_marks() ../gtk2_ardour/editor_tempodisplay.cc:185
#4 0x55f80ffb49fb in Editor::tempo_map_changed() ../gtk2_ardour/editor_tempodisplay.cc:301
#5 0x55f80ffbdf00 in Editor::abort_tempo_map_edit() ../gtk2_ardour/editor_tempodisplay.cc:850
#6 0x55f80fcf967a in TempoMarkerDrag::finished(_GdkEvent*, bool) ../gtk2_ardour/editor_drag.cc:333
```
Since no movement occurred, the tempo-map was not changed.
however we need to drop the lock and writable thread-pointer...
This allows UI scripts (saved in preferences) to access
os.* functions (non-sandboxed), while preventing other
scripts to do so.
Lua scripts that can run os commands can execute arbitrary
code on the system. While this is a nice feature, it can be
equally dangerous.
Without this, Editor::button_release_handler() will handle a button3 press
without checking for an active drag, resulting in two drags, and two reversible
commands, and an abort ...
This involved a significant change in event handling for automation region views,
but it brings it into line with how it works for other things. On button press
we initiate a drag, then if no motion occurs, the Drag returns false during
finalization, and only then do we continue through Editor::button_release_handler()
to eventually end up in ::add_automation_event().
Although it is a substantial change, the fact that it now works the same
way for audio regions, automation regions and automation tracks seems
like a definite plus.
Drag::_raw_grab_time is set to the correct time-domain. However
during ::motion() and other operations the adjusted time was
always AudioTime.
Most notably this caused editor->selection->time to have
different time-domains for start and end.
This reverts commit c578695a64.
When hovering over a control point there is still the "Fader"
cursor shown. It is also handy to be able to directly modify
a control-point right after adding it.
A freehand draw operation can be still be initiated above,
below, or left/right of a given control point.
There is a DisplaySuspender in MidiRegionView::init(), and another invoked
during the addition of a RegionView to a StreamView. Before this change, both
suspenders were created to invoke a full "model-changed" redisplay. Since every
RegionView must be added to a StreamView to be visible, we only need the second
suspender to invoke the model-change. The first one can be just a view
change (and probably isn't necessary at all, but I did not want to change that
much at this point in time).
VelocityGhostRegion used the visibility of the "parent" note canvas item of a
lollipop canvas item to determine the lolli's visibility. But during the
construction of the MidiRegionView, the note's container is not yet visible, so
this fails.
In addition this logic would hide lollis for notes that are outside the current
visible note range of the track (because the parent note item was not visible).
This change adds a method to MidiRegionView to decide if a note is within the
region's time range, and if so, we show the lollipop item. This means that
lollis for notes outside the note-range will still be visible, which seems more
correct. In addition, the nascent condition of the parent note's container no
longer affects lolli visibility.
The bindings use will all work just fine without this focus grab; the focus
grab in turn confuses the use of some/all global bindings (such as the spacebar
to toggle tranport state)
* Fix crash when inserting note to empty region
* Consistently use fine-grained modifier
* Don't allow to edit BBT as text field
This is currently not supported.
* Fix GTK_IS_CELL_EDITABLE assertions (Tab to next)
cb78043adc "Indicate selection extents in time ruler" introduced use of
superclock before it has been set. But as it only is used at time 0, the
actual superclock doesn't matter. Still, we work around it in order to
avoid triggering any warnings of uninitialized use.
In some startup flows, the entire loading process happens inside
StartupFSM::start(). In others, that call gets things moving but
we return from it and loading is not complete until later.
Deleting the StartupFSM while still inside the ::start() call
led to a use-after-free error. This new code will leak the StartupFSM
in some startup flows.
The old state in this commit would lead to an existing session
being loaded with the claim that it was a new session. This went
unnoticed until 4bed642d71, where newness impacted the time
domain of the session (and actually led to it being set to a random memory
value).
Variables by these names are only used from the local wscript and when
running "waf configure", which already for other reasons only can run at
the top-level.
These variables are thus not mandatory and not used.
'top' was a constant that was set to '.', even when inside
subdirectories. It is thus not really top.
I don't know if the intent was to use the actual top (which is available
as bld.top_dir), but for now we make it explicit what we have and do.