There is a check to see if it is no longer able to push more
messages into a full queue. When these is another CPU consuming
messages, the queue would not be full at that point. A new message
can be pushed into the queue and thus failing the "full queue"
check. So limit this to 1 CPU only so this check's assumption
can be fulfilled.
Fixes#79319
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Use an explicit callback name so that multiple instances of this do not
result in a:
redefinition of '_input_callback__longpress_cb'
error. This used to work when it was using unique generated wrappers,
but now it needs an index in the callback name.
Use it in one of the API tests as well, just in case.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Add a void *user_data pointer to the input callback structure. This is
useful for driver to get back the driver data structure and avoid
defining wrapper functions.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
The default 512 bytes stack size is a bit tight for some architecture
and leads to samples running out of stack. Let's default to 1024 and let
the user tweak it down if necessary.
Suggested-by: Benjamin Cabé <benjamin@zephyrproject.org>
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Looking back at the current INPUT_LISTENER_CB_DEFINE api naming, it
feels like it's a bit overloaded. Rename it to a simpler
INPUT_CALLBACK_DEFINE.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Add some initial tests for the input subsystem, covering all the APIs in
both synchronous and thread mode.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
