This is the first cleanup pass through makefiles and pios.
Probably it is difficult to track changes due to the nature of them.
I would recommend to look at resulting files and compiled code instead.
NOTE: original branch was rebased and lot of conflicts were fixed on
the way. So do not expect that every commit in this series will be
buildable (unlike original branch). Only final result was tested.
The main goal was to remove as much duplication of code (and copy/paste
errors) as possible, moving common parts out of Makefiles. It still is
not perfect, and mostly no code changes made - Makefiles and #ifdefs only.
But please while testing make sure that all code works as before, and no
modules/options are missed by accident.
Brief list of changes:
- Moved common parts of Makefiles into the set of *.mk files.
- Changed method of passing common vars from top Makefile to lower ones.
- Some pios cleanup, mostly #ifdefs, and all pios_config.h files.
- Many obsolete files removed (for instance, AHRS files, op_config.h).
- Many obsolete or unused macros removed or fixed/renamed (ALL_DIGNOSTICS).
- Unified pios_config.h template. Please don't remove lines for board
configs, only comment/uncomment them. Adding new PIOS options, please
propagate them to all board files keeping the same order.
- Some formatting, spacing, indentation (no line endings change yet).
- Some cosmetic fixes (no more C:\X\Y\filename.c printings on Windows).
- Added some library.mk files to move libs into AR achives later.
- EntireFlash target now uses cross-platform python script to generate bin
files. So it works on all supported platforms: Linux, OSX, Windows.
- Top level packaging is completely rewritten. Now it is a part of top
Makefile. As such, all dependencies are checked and accounted, no
more 'make -j' problems should occur.
- Default GCS_BUILD_CONF is release now, may be changed if necessary
using 'make GCS_BUILD_CONF=debug gcs'.
- GCS build paths are separated into debug and release, so no more obj
file clashes. Packaging system supports only release builds.
- New target is introduced: 'clean_package'. Now 'make package' does not
clean build directory. Use clean_package instead for distributable builds.
- Targets like 'all', 'opfw_resource', etc now will print extra contex
in parallel builds too.
- If any of 'package', 'clean_package', 'opfw_resource' targets are given
on command line, GCS build will depend on the resource, so all fw_*.opfw
targets will be built and embedded into GCS. By default GCS does not
depend on resource, and will be built w/o firmware (unless the resource
files already exist and the Qt resource file is generated).
- fw_simposix (ELF executable) is now packaged for linux. Run'n'play!
- Make help is refined and is now up to date.
Still broken:
- UnitTests, should be fixed
- SimPosix: buildable, but should be reworked.
Next planned passes to do:
- toolchain bootstrapping and packaging (including windows - WIP)
- CMSIS/StdPeriph lib cleanup
- more PIOS cleanup
- move libs into AR archives to save build time
- sim targets refactir and cleanup
- move android-related directories under <top>/android
- unit test targets fix
- source code line ending changes (there are many different, were not changed)
- coding style
Merging this, please use --no-ff git option to make it the real commit point
Conflicts:
A lot of... :-)
The flashfs layer is now entirely decoupled from the UAVO
data types. All UAVO related calls are now properly
contained within uavobjectmanager.c.
This will make the flashfs layer more portable as well as
more testable in a standalone environment. Eventually,
this layer could be used to store arbitrary files.
Also fixed all of the doxygen headers for the flashfs API.
The only downside is that it needs to update on the TaskInfo object which means
since htat updates periodically this code runs more often that really needed since
the modules only start at first. However, leaving that connection out means it
misses the initial set of modules.
flash so it says completed. However, it still blocks the system for a long
time. During an erase the heartbeat will flash at 10 Hz to indicate what's
happening.
This still blocks telemetry even after lowering hte system priority (and there
is a vTaskDelay) which makes me think that the SPI bus being locked is blocking
Sensors or somethign else. This should not be permited when the system is
armed.
The reason the system locks up during the erase is that the file system
operations occur within the event dispatcher thread. It is very bad practice
for anything to block this (i.e. callbacks should never take very long). We
should probably move the object persistence handling into the system thread or
something but that can be a separate issue.
OpenPilot platform (and thus sim too) was missed when the
DIAG_TASKS macro was broken out from the DIAGNOSTICS macro.
This allowed accesses to the TaskInfo UAVO even though it hadn't
been initialized.
After 3 failed warm start attempts, the init sequence
will force the RAM version of the HWSettings object
to its defaults. This should allow a user to regain
connectivity to a board that is continually faulting
during init.
This is accomplished by:
- Incrementing a boot counter that is stored in the
STM32 BKP registers. These registers survive a
warm start but are cleared on a cold start (ie. powerup).
- On multiple failures, force hwsettings to defaults
and raise the (new) BootFault alarm to prevent arming.
- Resetting the boot counter whenever the system manages
to successfully run the System Module task.
NOTE: This does not actually change the hwsettings object in
flash. That's up to the user.
This is intended to catch ONLY faults during early initialization.
It should not be used to recover from faults after the application
is up and running.
