| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
|
|
| |
ASSERT_* less and EXPECT_* more where we need to free resources.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
| |
These cleanup-marked lines where pretty bogus, as they wouldn't even be
executed if something ASSERT_'ed before it. Also, to have a clean slate,
the cleanup - i.e. remove all layers and surfaces - is done in TearDown.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Do not use pthread_cancel to shutdown control_thread. This could cause
stale locks in libwayland-client when the cancel was received when
the display lock (according to the current 1.3 implementation) is held.
To alleviate this, one could disable thread_cancellation in
control_thread during wl_* calls, this however is one more hack to the
cancel-to-shutdown hack. Instead, use a clean interface to send an
event to the control_thread when it's time to shutdown.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
|
|
| |
sync_and_acquire_instance coded as a macro that checks
the initialized field and returns ILM_FAILED if no initilized.
As sync_and_acquire_instance() is only to be used from ilm_*
implementation functions, this should be safe.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
| |
Use an old-fashioned while loop for the prepare_read+
dispatch_pending loop.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
| |
Replace valid with a needed "initialized" flag, also don't use
this flag to determine thread termination, pthread_cancel() is
used.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
| |
After display ownership is managed in ilmCommon, it is important
to destroy components in the reverse order they were created.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
| |
Get rid of the "main_ctx" name as it does not apply anymore.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
| |
Renamed get_instance() to sync_and_acquire_instance() to cleanly state
that there is more than just return an instance pointer. Acquires the
state lock and does a display_roundtrip_queue()
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
| |
* child global listener renamed.
* main global listener removed.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
| |
Also remove the "mostly-child" main globals listener.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
| |
First step towards single "context", use main_ctx for everything.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
|
| |
* Implements a dispatch loop pretty much like the wayland
documentation suggests.
* Acquire the context lock before each dispatching call.
* Make sure to cancel_read even when the thread is cancelled.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Using a private event queue for all ilm_* internal
wayland communication. This separates the ivi_*
objects from whatever is used externally.
* Init not happens _only_ on ilm_init
This also means no more is initialized polling and
possibly exit(0)ing from get_instace()
* Fixed destroy to happen in main thread where it will
actually be executed.
* Needed to adapt ilm_commitChanges() to new internal
wl_event_queue.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
|
| |
Introduced a new flag 'disconnect_display' separate from
valid, as adding bits to valid would change its scope and
entail changes to all usages of it.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
| |
* simplify implementation.
* only pass the data that we actually need.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
| |
* Use the userdata pointer in surface listener methods.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
| |
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
| |
* No more child/main, only one remains and we do not need a suffix
anymore.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
|
| |
* Actually use the same code for maina dnchild listeners, where
possible.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|
|
|
|
|
|
| |
* Pulled the check for existing layer out of the function.
Signed-off-by: Marcus Fritzsch <marcus.fritzsch@xse.de>
|