Found via `codespell -q 3 -S *.po,./.git,./share/patchfiles,./libs,./msvc_extra_headers,./share/web_surfaces,*.patch -L acount,ba,buss,busses,discreet,doubleclick,hsi,ontop,retrn,ro,scrollin,seh,siz,sord,sur,te,trough,ue`
VST preset restore has to happen in the GUI thread, specifically
the plugin's own UI event-loop (which on Linux and Windows is
not usually Ardour's Thread 1). However if there is no plugin
GUI, it should be safe to use Ardour's UI thread.
A widget's style (font) is only set after the widget has been
packed and inherits relevant information from the top-level window.
`set_size_request_to_display_given_text()` produces incorrect
results when it is called for widgets without a top-level
(here ARDOUR_UI::setup_windows() _main_window.add(), changes the
font from "Sans 11" to "Sans 9" for ArdourDropdowns).
`set_size_request_to_display_given_text()` must subscribe to
`signal_style_changed()`, however in case of ArdourButton UI-scaling
of elements also needs to be taken into account. It is preferable
to allow for multiple sizing-texts directly.
RegionFactory::create(Region&,..) copies all region properties.
There are however various operations that use a new Source
as intermediate to create a new Region from another Region.
Those operations should also inherit various region properties
(such as mute, opaque, locked, ..)
This likely broke in 3.x when Ardour migrated to context dependent
shortcuts (rather than GTK managed menu shortcuts).
While there is Gtkmm2ext::Application::hide, there is currently no
bindable action and hence no way to add a dedicated
shortcut. Furthermore, since this is a special NSApplication menu,
updating the menu-item when the action is bound is not trivial.
gtk_application_hide() is only implemented for macOS and
there is likely a better use for "H" shortcut anyway.
Querying available buffersizes and sample-rates requires access
to the device. Almost all ALSA devices are limited to
a single user-space application so we unconditionally try
to request access to the device.
Even though SIGINT is intended to be sent from terminals only,
it is more successful at terminating various child processes than
closing stdin and sending SIGTERM.
This effectively triggers a Session::request_overwrite_buffer
via Playlist::notify_contents_changed when the time-domain of a
Region changes.
When the time domain of a MIDI region changes, Region boundaries
and events positions change ever so slightly. The playlist has
to be rendered again to take this into account.
When MIDI regions with different time-domains are layered, notes
may not be resolved in correct order. Previously changing time-domains
to address this did not result in the playlist being reread.
This adds a special case of "flush/resolve" to restore the
state of an upper layer opaque MIDI region while at the same time
resolving notes of a lower layer region.
This honors transparent/opaque property of MIDI regions.
Previously all MIDI regions played as if they were unconditionally
transparent.
Further work is needed for backwards compatibility (and default
setting).
The iterator into the model can become incorrect in the sense that it is no longer the correct
next event to play. This can happen at least with a tempo map change, and possible under other
conditions. Catch this when it happens, and act as if we reached the end of the trigger
