LL_APPEND is called just before the loop spins up but various error conditions can cause the task to exit before then.
When that happens Mavlink::start_helper calls delete on the instance which tries to prune it from the global list.
If this is the first Mavlink instance to attempt starting the list head is null and we hardfault in the Mavlink dtor.
Only call LL_DELETE after checking the list head for a null pointer.
QuRT does not support UDP so moved the mavlink specific code
to a new file that is not built for the qurt target
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
The inclusion of px4.h requires C++ features not supported in the
Hexagon toolchain. The features are not required so the required
headers are used instead.
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
Most of the current POSIX builds will not use the UART and it
is just a unnecessary dependency to satisfy when running in a
test environment.
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
QuRT does not have a filesystem, so creating a virtual filesystem
that could be implemented as an in-memory file or a remote file
over fastRPC.
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
The calls to task_spawn_cmd, kill_all, and systemreset were wrappers
around the px4_{task_spawn_cmd|kill_all|systemreset} implementations.
Removed the wrappers and changed all calls to the px4_ equivalents.
NuttX specific code was moved into px4_tasks.h
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
The latency_test used to pass an object pointer as argv which
won't work in the posix port because it expects argv to be a
null terminated array of character pointers (which it makes a
copy of).
The test was refactored to use a singleton pattern and avoid
having to pass the object pointer to the thread.
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
GCC was more picky about prototypes for inlines being required.
The generate_listener.py script used incorrect printf formats and
was casting %f params to float, but printf casts all %f params to
double per the spec.
Signed-off-by: Mark Charlebois <charlebm@gmail.com>