It seems that without explicit initialization of the members they remain
uninitialized. The lack of initialization caused crashing, because
PavuApplication::on_activate() creates the MainWindow only if mainWindow
is NULL, and if it's some random value instead, on_activate() calls
selectTab() on a non-existent window.
Fixes: https://gitlab.freedesktop.org/pulseaudio/pavucontrol/issues/49
This introduces a new file for clarity. Options
handling changes so that --tab changes the tab
if the window is already opened. Other options
are only used at start time.
Due to another bug the MIME type guessing of PNG files was broken on my
Ubuntu system. This made pavucontrol crash on startup (while loading the
necessary icons) without a helpful error message (requiring a lot of
debugging effort). Although this is not originally pavucontrol's fault,
I think that pavucontrol could be more robust about such a problem,
especially because:
A) In a complex database like the one to guess MIME types (that is populated
from hundres of different packages) this can happen from time to time and
B) pavucontrol already has some fallback options in place - they are just not
active in this specific case.
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=88813
pa_stream_peek can return NULL if either the buffer is empty or if it
has a hole. In either case we need to avoid derefencing the data
pointer. Additionally, if there is a hole, we need to call pa_stream_drop,
if the buffer is empty we should not call it.
BugLink: https://bugs.debian.org/735898
RoleWidget::create contains:
x->get_widget_derived("streamWidget", w);
But streamWidget is defined as following in the glade file:
<object class="GtkEventBox" id="streamWidget">
Where as RoleWidget is derived from [Minimal]StreamWidget, which is derived
from Gtk::VBox, so this is clearly wrong.
Adding:
printf("rolewidget type: %s\n", g_type_name(G_TYPE_FROM_INSTANCE(w->gobj())));
for debugging shows that this really leads to RoleWidget being instantiated
as an EventBox (yet things still work due to sheer luck).
This commit fixes this, by putting the streamWidget id at the right level of
the hierarchy in the glade file (and likewise for cardWidget and deviceWidget).
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Widgets (unlike Windows and Dialogs) returned by Gtk::Builder::get_widget*
start owned by the GtkBuilder object, the idea being that they will get
added to a container before the scope of the GtkBuilder object ends, and it
thus automatically gets destroyed.
But in the various ::create methods in pavucontrol, a pointer to the widget
gets returned, so that it can be added to a cointainer by the caller.
However as soon as the ::create method exits the GtkBuilder object owning
the widget, and thus also the widget gets destroyed, and we end up returning
free-ed memory.
This commit fixes this by making all ::create methods take a reference on
the widget before returning it, and having all the callers unreference the
widget after adding it to a container.
https://bugs.freedesktop.org/show_bug.cgi?id=83144https://bugzilla.redhat.com/show_bug.cgi?id=1133339
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Add a horizontal scrollbar inside each tab automatically when the window
is not wide enough. Add arrows for scrolling the tab bar when the tabs
don't fit inside the window.
ellipsize labels to make the window resizable even with long label text
and add tooltips to provide a way to read the full text
Originally from Sebastian Wick <sebastian@sebastianwick.net>
If we know if a certain port is available/unavailable, we can print
that out, as a help to the user (and as debugging for ourselves).
A profile is also available/unavailable if all ports which have that
profile are available/unavailable.
Credit goes to David Henningson for the original idea and some of the code.
The prepareMenu() call can change the active profile selection, which
in turn will cause a "set card profile" command to be sent to the
server if the "updating" flag is not set, so the "updating" flag needs
to be set when calling prepareMenu() from updateCard().
This caused a problem with disconnecting bluetooth headsets: as part
of the disconnection procedure, module-bluetooth-device sets the card
profile to "off". At that point module-card-restore doesn't do
anything, because the change is marked as "don't save". But the
profile change event is then sent to pavucontrol, which updates its
view, and pavucontrol sends the new profile ("off") back to
pulseaudio, and this time the profile change iss marked as "please
save", so module-card-restore restores the "off" profile when the
device is connected again, even though the user never requested the
"off" profile to be chosen.
As a developper, I find it annoying to restart pavucontrol everytime I restart
pulseaudio, moreover the error dialog sometimes needs an additional click
before restarting.
Add it as a command line option so that default behavior is not changed.
This makes sure we inhibit autosuspend only for network sources (which
was the main purpose of adding autosuspend, since constantly monitoring
those is network heavy).
This only seemed to affect GTK3 and it seems to introduce
a GTK warning relating to 'gtk_widget_size_allocate' similar
to those fixed in the previous commit, but nothing seems to
be the worse for it.
