AirborneCodeReorg

From PaparazziUAV
Revision as of 04:51, 1 November 2010 by Flixr (talk | contribs) (clean up a little, to reflect current status a bit better)
Jump to navigation Jump to search

Our Challenge

The current airborne sourcecode has gone a little bit out of control. Everyone was working on it enthusiastically, which is great! However, we feel now the moment has arrived where deep reorganization of the code would benefit us all.

It is really appreciated if you want to help out in this quest. Therefore remarks and suggestions can be placed on the page designed specifically for this. Of-course feel free to use the mailinglist also.

Problem areas

Fixed wing airframes have makefile directly in them - we can't change anything in the airborne code without breaking everybody's airframe

--> use the subsystems

There is also the "module" facility which is very good for optional features.

There are 3 kinds of code in the airborne directory

  • generic code (that doesn't depend on any architecture or plateforme) - there's a lot of it and it should probably be sorted by type of functionalities
    • fixedwing/main.c
    • thrust_vectoring/main.c
    • gps.[ch]
  • processor specific code (stm32, lpc21, avr, etc...)
    • arch/lpc21/uart_arch.[ch]
  • plateforme specific code (tiny, booz, lisa, etc...)
    • blaaaaa

Coding Style

  • 2 spaces for indent
  • please no trailing whitespaces (e.g. fix with M-x whitespace-cleanup in emacs)

Proposals and RFC

Directory structure proposals

relative to sw/airborne

 firmware/
     rotorcraft/
         rotorcraft.makefile
         rotorcraft.xml // list of subsystems: maybe autogenerate as this should reflect the directory structure
         foo/
             foo.c
             foo_2.c
     fixedwing/
         control/
             control.c
             control_adaptive.c
 subsystems/
     imu/              // a general subsystem with several implementations and arch depended code (initialization, interrupts, ...)
         imu_xsense.c
         imu_b2.c
         calibration_routines/    // might be useful to have subdirectories
     gps/              // another general subsystem
         ....
     radio_control/    // another general subsystem
         ....
     actuator
         stm - lpc - arm - omap - ...
 boards/
     test_tiny
     test_lisa
 modules/
     ....
 arch/
     subsystems/
         imu/
             lcp21/
                 imu_xsense_arch.[ch]
                 imu_b2_arch.[ch]
             stm32/
                 imu_xsense_arch.[ch]
                 imu_b2_arch.[ch]
             sim/
     firmware/
         fixedwing/
             control/
                 lcp21/
                     control_arch.[ch]
                     control_adaptive_arch.[ch]
                 stm32/
                     control_arch.[ch]
                     control_adaptive_arch.[ch]
                 sim/


each general susbsystem having several implementations gets a subdirectory under subsystems

 imu.[ch]
 imu/
   imu_xsense.[ch]
   imu_b2.[ch]

Rationale for the subsystem directory:

  1. to have less directories in the airborne root
  2. to help auto-documentation tools: "all subfolders in airborne/subsystems/ are subsystems. if you put them directly in airborne it becomes: all directories in airborne are subsystems except arch/firmware/modules/... etc
  3. it's less obvious how the imu code is used (is it standalone/firmware/modules) to newcommers than with the extra dir.

Test Code

Testing code is specifically for UAS of utmost importance. To test we could opt to make small cases in Testlink for everyone to try and report the outcome.

Replace makefile section with xml - [READY]

of course leave the possibility to have a makefile section, but the normal user would have something like

airframe.xml => Makefile.ac
 <firmware name=rotorcraft>
   <target name="ap" board="booz"/>
   <target name="sim" board="sim" />
   <subsystem name="radio_control" type="ppm"/>
   <subsystem name="actuators" type="mkk"/>
   <subsystem name="imu" type="b2v1"/>
 </firmware >
 <firmware name="setup" >
   <target name="tunnel" />
   <target name="test_actuators" />
 </firmware>
 ifeq ($(TARGET),ap)
 include $(PAPARAZZI_SRC)/conf/board/booz.makefile
 include $(PAPARAZZI_SRC)/conf/autopilot/rotorcraft.makefile
 include $(CFG_BOOZ)/subsystems/booz2_radio_control_ppm.makefile
 include $(CFG_BOOZ)/subsystems/booz2_actuators_mkk.makefile
 include $(CFG_BOOZ)/subsystems/booz2_imu_b2v1.makefile
 endif
 ifeq ($(TARGET),sim)
 include $(PAPARAZZI_SRC)/conf/board/pc.makefile
 include $(PAPARAZZI_SRC)/conf/autopilot/rotorcraft.makefile
 include $(CFG_BOOZ)/subsystems/booz2_radio_control_ppm.makefile
 include $(CFG_BOOZ)/subsystems/booz2_actuators_mkk.makefile
 include $(CFG_BOOZ)/subsystems/booz2_imu_b2v1.makefile
 endif
 ifeq ($(TARGET),tunnel)
 include $(PAPARAZZI_SRC)/conf/board/booz.makefile
 include $(PAPARAZZI_SRC)/conf/autopilot/setup.makefile
 endif


?? put defines in here that get written in airframe.h ??

cdewagter I'm not really sure this is needed as otherwise you might put all other data of the airframe file in here too next to the submodule it belongs too. I think it is more readable with less defines in here. I would suggest to only have global compile defines that need to be in the Makefile.ac here...

airframe.xml => Makefile.ac
 <firmware name=fixedwing>
   <target name="ap" board="tiny_2.11">
     <define name="ALT_KALMAN"/>
   </target>
   <subsystem name="led" />
 </firmware >
  FLASH_MODE = IAP
  include $(PAPARAZZI_SRC)/conf/board/tiny_2.11.makefile
  include $(PAPARAZZI_SRC)/conf/autopilot/fixedwing.makefile
  ap.CFLAGS += -DALT_KALMAN
  include $(SRCS_FIXEDWING)/led.makefile
airframe.h
 #define USE_LED

set makefile variables

airframe.xml => Makefile.ac
 <subsystem name="gps" type="ubx">
   <param name="GPS_UART_NR" value="1"/>
   <param name="GPS_BAUD"    value="38400"/>
   <param name="GPS_LED"     value="2"/>
 </subsystem> 
GPS_UART_NR=1
GPS_BAUD=38400
GPS_LED=2
include $(CFG_FIXEDWING)/subsystems/gps_ubx.makefile
  • Note that now a subsystem can be enabled for all targets in that firmware by using
 $(TARGET).srcs += navigation.c
  • Note that you only 1 time in the board file (e.g. /boards/tiny_2.11.makefile)
 $(TARGET).arch = arm7

instead having this in the firmware/subsystem makefiles

 tunnel.ARCH = arm7
 ap.ARCH = arm7
 fbw.ARCH = arm7
 ...


Include guards

All headers have to be checked for missing include guards, style template:

#ifndef <MODULE>_<TARGET>_<FILENAME>_H
#define <MODULE>_<TARGET>_<FILENAME>_H
[code]
#endif /* <MODULE>_<TARGET>_<FILENAME>_H */

Example for sw/airborne/csc/airspeed.h:

#ifndef CSC_AP_AIRSPEED_H
#define CSC_AP_AIRSPEED_H
[code]
#endif /* CSC_AP_AIRSPEED_H */


Here is a list of headers know not to include guards please remove any that you update from the list.

sw/in_progress/videolizer/spook/conf_parse.h
sw/in_progress/videolizer/spook/rtp.h
sw/in_progress/videolizer/spook/stream.h
sw/in_progress/videolizer/spook/log.h
sw/in_progress/videolizer/spook/pmsg.h
sw/in_progress/videolizer/spook/global_config.h
sw/in_progress/videolizer/spook/outputs.h
sw/in_progress/videolizer/spook/jpeg_tables.h
sw/in_progress/videolizer/spook/frame.h
sw/in_progress/videolizer/spook/inputs.h
sw/in_progress/videolizer/spook/event.h
sw/in_progress/videolizer/spook/mpegaudio.h
sw/in_progress/videolizer/spook/encoders.h
sw/in_progress/videolizer/spook/base64_table.h
sw/in_progress/videolizer/spook/rtp_media.h
sw/in_progress/videolizer/wis-go7007-linux/kernel/go7007-priv.h 
sw/in_progress/videolizer/ws-go7007-linux/kernel/wis-i2c.h 
sw/in_progress/videolizer/wis-go7007-linux/include/go7007.h 
sw/in_progress/inertial/C/tilt_display.h sw/airborne/fms/packet_header.h 
sw/airborne/arm7/interrupt_hw.h 
sw/airborne/arm7/lpcusb/usbhw_lpc.h
sw/airborne/arm7/lpcusb/usbdebug.h 
sw/airborne/arm7/lpcusb/examples/console.h 
sw/airborne/arm7/lpcusb/examples/startup.h 
sw/airborne/arm7/lpcusb/examples/msc_bot.h 
sw/airborne/arm7/lpcusb/examples/spi.h 
sw/airborne/arm7/lpcusb/examples/msc_scsi.h
sw/airborne/arm7/lpcusb/examples/blockdev.h
sw/airborne/arm7/lpcusb/usbapi.h
sw/airborne/arm7/test/bootloader/console.h
sw/airborne/arm7/test/bootloader/usbhw_lpc.h
sw/airborne/arm7/test/bootloader/startup.h 
sw/airborne/arm7/test/bootloader/usbdebug.h 
sw/airborne/arm7/test/bootloader/usbapi.h 
sw/airborne/arm7/test/bootloader/lpc21iap.h 
sw/airborne/arm7/test/welcome.h 
sw/airborne/csc/csc_vane.h 
sw/airborne/csc/csc_airspeed.h 
sw/airborne/vor/lo.h 
sw/airborne/coaxial/tl_gps_configure.h 
sw/airborne/flightzone.h 
sw/airborne/booz/arch/stm32/radio_control/booz_radio_control_ppm_arch.h 
sw/airborne/sd_card/sd_card.h 
sw/airborne/ism/uart_hw.h 
sw/airborne/sim/sim_uart.h 
sw/airborne/sim/ivy_transport.h 
sw/airborne/sim/led_hw.h 
sw/airborne/osam_imu_ugear.h 
sw/airborne/wind_tunnel/wt_baro.h

Misc

  • All headers are defined relative to sw/airborne (limit the -I to -Isw/airborne and the generated code)
 #include "foo/whatever.h"
  • Generated headers are indicated by their directory
 #include "generated/airframe.h"

Proposed methodology

[6:50:19 PM] Christophe De Wagter: "builing a new road is easy: nobody can ride until it is finished." ... "for road-works on an existing road there are 2 options: a) a non-important road can be closed for a while during roadworks b) but for roads that are too important: you need to keep a lane open: that is more work than closing it down but in the end keeps the total traffic problem minimal" [6:51:05 PM] Christophe De Wagter: I think that a gradual approach might be slightly more work, but keeps the road open and developments running.

I could not agree more!


things discuss or cleanup/remove

  • remove sw/include/std.h? was for old avr stuff
  • cleanup TODO file
  • there is wind_tunnel under sw/airborne and under sw/in_progress
  • make a bin and/or build dir where all the ground sw build files and binaries go (paparazzicenter, gen_tools) to only have source files under sw