"While 'atomic' has a volatile qualifier, this is a historical
artifact and the pointer passed to it should not be volatile."
Furthermore "It is very important that all accesses to a
particular integer or pointer be performed using only this API"
(from https://developer.gnome.org/glib/2.68/glib-Atomic-Operations.html)
Hence initialization of atomic variables is changed to also use
this API, instead of directly initializing the value.
This also fixes a few cases where atomic variables were
accessed directly.
see also libs/pbd/pbd/g_atomic_compat.h
Playlist::ContentsChanged() is implicit.
Region::send_change()
-> Playlist::region_changed ()
-> Playlist::notify_contents_changed ()
-> Playlist::ContentsChanged () /* EMIT SIGNAL*/
This cuts the number of signals in half and also
allows to freeze/thaw the playlist to collect the
signals for each playlist into a single signal.
With a large session:
Editor::insert_time()
-> Playlist::shift()
-> 4000+ regions are modified
-> 4k calls to Region::send_change()
-> --"-- Playlist::region_changed ()
-> --"-- Playlist::region_bounds_changed ()
-> --"-- Playlist::notify_contents_changed ()
-> --"-- ContentsChanged () /* EMIT SIGNAL */
-> --"-- DiskReader::playlist_modified ()
-> 4k Session::request_overwrite_buffer events are queued
The butler thread processes them all in the background, but
this also enqueues 4k+ events to the GUI event pool since the
GUI subscribed to Playlist::ContentsChanged ().
However the GUI is inside Editor::insert_time() and cannot handle
events. So they keep accumulating, and eventually hits
"POOL OUT OF MEMORY - RECOMPILE WITH LARGER SIZE!"
-=-
This fixes the issue by collecting blocking ::notify_contents_changed
until all region_changed() events are processed, and a single call
to Playlist::flush_notifications() notifies the UI and disk-reader.
This works around JACK not allowing to directly access physical
input buffers by creating an explicit connection.
Ardour access input-buffers directly without connection
for input meters and AFL. This work in general since data
on those ports are always read unconditionally at the start of
each process cycle.
jack_port_get_buffer (jack_port_by_name (c, "system:capture_1"), n);
However with jack this fails if the given port is not connected.
Due to an internal jack optimization collecting input data is
skipped for ports that have no connections.
This fixes the following edge-case:
1. Place a custom transient marker.
2. Use Rhythm Ferret to analyze region.
3. Delete custom marker from (1) while rhythm-ferret
dialog is still open, and analyzed onset markers are visible.
PortAudio uses what it calls 'default suggested latencies' but in callback streaming mode, they can result in wildly inaccurate buffer sizing (e.g. the user requests a buffer size of 128 but PortAudio actually instructs ASIO to use a much bigger size).
What we do now is to improve PortAudio's suggested latency calculation by basing it on the actual buffer size requested by the user.
By default Apple uses a private TMP folder. While mktemp
returns "/tmp/xxx" the canonical path is "/private/tmp/xxx".
This lead to issues when tmp-prefix is removed when building
the session-archive.
See also e52bdc55ad
Since 6.0 take-name was not updated correctly:
The name changes in Session::non_realtime_stop(). At that
time tracks are still record-armed and ignored name changes.
Newly created tracks also never had a diskstream name set
correctly. This only happened at session-load via set_name(),
or config change.
This is is useful to determine if an undoable action was
performed before adding additional information (e.g. selection
changes) to the undo transaction.
Method Session:locations():range_starts_at(pos, slop, incl)
to search range by start point with some inaccuracy delta.
Similar to mark_at(pos, slop)
Method Session:locations():add_range(start, end)
to create new range and get it for later changes.
For some unknown reason, VC++2019 won't let us take the address of std::list::unique() - although other std::list members seem okay. I've spent weeks tracking this down but there's no fix available AFAICT.
I've flagged it up to the MSVC development team - just don't hold your breath !!
Do not simply allocate std::vector<> space but also initialize
elements. The data is later accessed as C-pointer array: &var[0].
With most compilers simply reserving space in the vector is
sufficient in order to later access the elements directly.
However actually placing objects in the vector before referencing
them is more correct.
If a plugin implements Vst::IComponent::set_state() but
does not implement Vst::IEditController::setComponentState()
nor dedicated get/setState interface for the IEditController,
querying the parameters using Vst::IEditController::getParamNormalized()
returns values that do reflect the restored state.
In this case the host needs to save/restore all control-parameters,
and ignore values reported by ::getParamNormalized().
This fixes a state save/restore issue with softube.vst3, however
vstpresets are still broken: the GUI (IComponent) is updated,
however the controller isn't.
