Steps to debug:
* Run gazebo (or any other sim) server and client viewers via the terminal: `make posix_sitl_default gazebo_none_ide`
* In your IDE select `px4_<mode>` target you want to debug (e.g. `px4_iris`)
* Start debug session directly from IDE
This approach significantly reduces the debug cycle time because simulator (e.g. gazebo) is always running in background and you only re-run px4 process which is very light.
<systemlib/err.h> --> "systemlib/err.h" that fix_headers.sh
would miss because it comes after code it doesn't understand.
Effectively remove the '__EXPORT extern perf_counter_t perf_alloc('
line, because currently perf_alloc is defined to be NULL, and
after running fix_headers.sh that happens *before* this header
is included (the order of headers will be changed).
Do not define NULL to be (void*)0: that only works for C.
In fact, the conversions needed for NULL are so full of exceptions
that standard C++ introduced a new *keyword*: nullptr.
That's what we really should be using for C++ code.
In this case I just include the correct header to define NULL
the correct way.
Not really related to the header line:
Removed an #include <time.h> because I noted that px4_time.h
was already included... and moved a #include <sys/time.h>
to the top of the file (not really a fan of including headers
in the middle unless absolutely necessary).
Removed a include of queue.h because I noted it wasn't used.
Using cout drags in std::localeconv which isn't defined
on QURT. While this file is also used for POSIX, it doesn't
seem to harm much to use printf there as well.
This patch reorders px4_defines.h to make it more readable (I think)
but more importantly, cleans up the #include <math.h>/<cmath>
and [std::]isfinite stuff.
My main goal was to completely get rid of including math.h/cmath,
because that doesn't really belong in a header that is supposed to
define macro's and is included in almost every source file (if not
all).
I'm not sure what it did before ;) (pun intended), but now it does
the following:
PX4_ISFINITE is only used in C++ code (that was already the case,
but hereby is official; for C code just use 'isfinite()') and is
defined to be std::isfinite, except on __PX4_QURT because that uses
the HEXAGON toolset which (erroneously) defines isfinite as macro.
I would have liked to remove PX4_ISFINITE completely from the code
and just use std::isfinite whereever that is needed, but that would
have required changing the libecl submodule, and at the moment I'm
getting tired of changing submodules... so maybe something for the
future.
Also, all includes of <math.h> or <cmath> have been removed except
for __PX4_NUTTX. Like the HEXAGON toolset NuttX currently defines
isfinite as macro for C++. So, we could have solved this in the
same was as __P4_QURT; but since we can fix NuttX ourselves I chose
to add a kludge to px4_defines.h instead that fixes this problem,
until the time that NuttX can be fixed (again postponing changing
a submodule). The kludge still demands including <cmath>, thus.
After removal of the math header file, it needed to be included
in source files that actually need it, of course.
Finally, I had a look at the math macro's (like M_PI, M_PI_F,
M_DEG_TO_RAD etc). These are sometimes (erroneously) defined in
certain math.h header files (like both, hexagon and nuttx).
This is incorrect: neither the C nor the C++ standard defines
math constants (neither as macro nor otherwise). The "problem"
here was that px4_defines.h defined some of the M_*_F float
constants in terms of the M_* double constant, which are
sometimes not defined either thus. So, I cleaned this up by
defining the M_*_F math constants as float literals in px4_defines.h,
except when they are defined in math.h for that platform.
This means that math.h has to be always included when using those
constants, but well; not much difference there as those files
usually also need/use the macro NAN (which *is* a standard macro
defined by math.h).
Finally finally, DEFAULT_PARAM_FILE was removed as it isn't
used anymore.
All in all I think the resulting px4_defines.h is nice, giving me
much less the feeling of a nearly unmaintainable and over time
slowly growing collection of kludges and hacks.
These are some changes that I needed to compile
most of the unsupported targets.
After this all (make list_config_targets) compile
for me except: posix_eagle_muorb, posix_sdflight_default
and qurt_eagle_legacy_driver_default.
This is the equivalent to Greg's upstream solution for ctype.h &
cctype with the necessary shadow waning fix.
It also deletes the Hack in pa4_defines.h
app.h, generated from app.h_in, has unnecessary code duplication
and isn't a header file (it defines globals, static functions
and doesn't have a header guard, moreover, it has a 'using namespace
std;'). Because of this, a real headerfile that declares the stuff
defined in apps.h was missing leading to even more code duplication:
scattered forward declarations in .cpp files and an often repeated
type of std::map<std::string, px4_main_t>.
This patch moves cmake/qurt/apps.h_in to src/platforms/apps.cpp.in
(with some changes) and removes cmake/posix/apps.h_in.
Then src/platforms/apps.cpp.in is split into src/platforms/apps.cpp.in
and src/platforms/apps.h.in, splitting declarations from definitions.
A typedef is defined for the map (apps_map_type).
The main difference between cmake/posix/apps.h_in and
cmake/qurt/apps.h_in was that the first defined a global 'apps',
while qurt stores the apps in QShell. I opted to get rid of
the global variable (which are in general evil for various reasons)
and used the API of cmake/qurt/apps.h_in where a provided 'apps'
map is initialized with a call to init_app_map. Thus removing
the existing code duplication.
The thread name is used from within the entry of the new thread, but the
provided name could live on the stack of the caller thread. Thus we need to
copy the name.
The simulated timer interrupt always adds a new scheduled work task, which
is called from the work queue thread. Sending the signal creates measurable
overhead (~5% of the px4 CPU runtime) and is unnecessary, since the thread
is not sleeping anyway.
It makes more sense to set the optimization flags on a platform basis
instead of individually for each module. This allows for different
optimization options for SITL, NuttX, Snapdragon, etc.
When running a simulation with, for example,
make posix jmavsim
px4 crashed almost 100% reproducable near start up.
This turned out to be a stack overflow. On gitter
it was suggested that the main reason for this could
be stack sizes, as currently used, assume a 32bit pointer
size and that doubling the stack size for everything
but NuttX would be the Right Thing.
This is the solution that I came up with (it makes
my core dumps disappear).
- ulog file message rate limited to 50Hz
- queuing with size 2
- this replaces the mavlink log message in the ulog
(but the mavlink warnings & errors still go to the ulog)
David Sidrane