Dev Call Mar 05, 2018 2300 UTC


  • Update


  • Update

Dev wiki

  • Update


  • STM32F7 progress
  • docs for porting to ChibiOS
  • single-file porting
  • cygwin waf build
  • Vote results
  • Contributor of the month
  • New developers
  • looking for QGC developer


Dev call

  • Revising the Dev call structure to make it more efficient

Attendee count (max): 28

UTC2306 - Randy and Rover

  • Beta release for Rover later today
    • Fixes when-desired-speed-0-can’t-steer issue
    • Some testing to do to make sure it doesn’t kill boats
  • Randy now has a boat

UTC2306 - Randy and Copter

  • Countdown to beta-testing
  • A few things to go in
  • Hopefully before end of March for beta testing
  • New loiter from Leonard needs to go in
    • Making progress
    • Held up by QuadPlane changes required
      • Weird problem with landing
  • MdB is hoping to get battery stuff in

UTC2306 - Randy and Wiki

  • RealFlight testing instructions are being improved
  • Wiki editing guide
  • Interactive rebase guide
    • MdB: something that tells them how to split a commit into multiple commits would be great
    • [10:11 AM] (Channel) MdB: We get lots of monolith commits then confusion about how they actually split them up
  • Full list of boards we support
    • E.g. revo mini
  • Interfaces area

UTC2319 - tridge and stm32 porting

  • Focus on making new boards easy
  • Minimise conflicts when porting to a board
    • Reduce entry points into the source code
    • Should only need to add a single file now!
  • Write-up courtesy of Randy
  • Ifdefs in Compass baro etc are getting difficult to manage
  • Support list of drivers to load in hwdef.dat?
    • Include orientations in board file too
  • Philip’s F7 board is on the way
    • And has already put hwdef.dat in place
  • Tridge bought a cheap F745 Omnibus v2 board
    • USB is up
    • SPI is partly working
      • Connects with mavlink over USB
      • Tridge can see one of the Invensense IMUs
        • Sporadic errors
        • Does get accel and gyro samples
      • Probably a clocking issue
    • Abstractions for different MCUs are going in
      • No longer F4-centered
      • Can later do non-stm32-based-boards
    • No ccm
    • Many different classes of memory on an F7
      • Only using main memory ATM
    • 745 has no double-precision
      • 769 does
    • Bootloader
      • Currently loading AP at first sector
        • Slow DFU!
      • Considering making a bootloader based on hwdef.dat
      • Non-trivial effort!
  • F1?
    • Not for main flight code (not big enough)
      • Would need to go back to DCM
        • Not enough flash for KF
      • Wouldn’t really be AP any more
      • Betaflight has deprecated
    • Might be able to do F3 for main flight code
      • Would need a reduced-state KF
        • Without GPS, for example
      • Racing community is moving away from F3
      • Maintenance burden
    • But need F103 for the PX4 IO microcontroller
      • On top of AP_HAL or not?
        • Firmware size….
        • F103 has only 64kB of flash
          • 4kB for bootloader
          • Minimal AP_HAL is fw ~40kB
          • 20kB for mixer logic seems tight!
            • So probably won’t use AP_HAL
            • Still use hwdef.dat
      • Cube Red has F103 / 64kB of flash
  • Getting logs off Cube Red is problematic
    • WiFi?
      • Transmitter on-board leads to issues
    • USB-C allows sealing of unit
      • SD card is a problem!

UTC2334 - CygWin

UTC2334 - new compiler?

UTC2240 - speeding up compiling

  • Disable header checks on Windows?
  • Command-line option already
    • Change default
  • Remove -g on ChibiOS build by default
  • Turn off anti-virus under Windows
    • Or add exclusions
  • Need a Wiki page on how to speed up compilation
    • Randy will add to his list
  • Amilcar is very interested in fixing things!

UTC2300 - contribution of the month

  • Should get nominations for February
  • Tom nominates himself for bad merge
    • Big impact!
  • Olivier will put up a request for nominations

UTC2302 - nominations for Bill and Chris for devteam

  • Joint maintainer role for Heli?
  • Tridge was interim maintainer and (in his words) “did an awful job”
  • Amount of testing has been fantastic
  • Wiki work
  • Support work
  • Need to get heli re-consolidated
    • Randy’s got some stuff on his plate to make that happen
  • Luis’ run through the CoC with them
  • Randy will put two vote together
    • One form two questions
  • Is it differentiated in the Wiki?
    • Yes, there’s a “traditional helicopter” section
    • Developer process for tradheli could be made easier
      • ./waf copter builds heli as side-effect but no way to upload
        • Uploads a random firmware
  • RealFlight8 wiki page has been updated for tradheli

UTC2310 - sponsoring a developer for 3-4 months for 20 hours/week on QGC

  • If you would like to work on QGC for money, contact Craig E
  • And there’s a separate one - contact Randy or Craig

UTC2313 - job postings

UTC2314 - Tom and Nate

  • Flightmode Wiki document
  • Nate’s available for testing binaries
  • Saving stuff from travis so we can test it?
    • So we could say that a PR has been flight-tested
    • Delete binary once PR is closed
    • MdB: too dangerous!
    • Philip: would love to be able to do that
  • Added a feature so own builds were built on autotest
    • Based on tags
    • Repos / tags
    • Was never used
  • E.g. ArduHeli project
    • External patches
    • Use ability to push to specific branches
    • Grant CO ability to push to a branch and have autobuild system build it and make binaries available
      • Jacob for sub, too
  • Ardupilot-next
    • Huge amount of work
    • MdB: The number of not working PR’s is high…

UTC2317 -

  • Jacob wants to add arbitrary rotations
    • Any interest?
  • Have a trim option?
    • So AHRS_TRIM_X, AHRS_TRIM_Y but for compasses
  • roll/pitch/yaw instead?
  • Interactions with other sensors?
  • JP: Matrix would enable cal to set the orientation

UTC2317 -

  • Battery failsafes
  • Don’t want to fall back to a lower-level failsafe
  • Enumeration for actions?
  • [11:37 AM] To Weekly devcall: subclass a failsafe-action library per-vehicle?
  • AFS_Terminate?
    • Kills the vehicle
  • FF: Peter: that’s an option but when I talked to Michael, it seemed like too much for just having a map of priority to action number
  • Unified failsafe would be nice
    • But should probably wait for scripting
  • Failsafes in motortest for Copter needs looking at
  • Weekly devcall: Is “scripting” our new “punt down the road” statement? :slight_smile:
  • [11:41 AM] (Channel) FF: I hope not or I’ll need to learn Lua properly :smiley:
  • [11:42 AM] (Channel) MdB: Peter: Yup :stuck_out_tongue:
  • [11:42 AM] (Channel) MdB: no pressure or anything
  • Jacob will review this PR this week
  • Tridge is happy with Plane changes
  • Some odd constants in the Copter code

UTC2345 -

  • Camera feedback stuff
  • Flags should be a bitmask
  • Change XML to change meanings?
    • Change enum
      • Not part of hash so shouldn’t be a problem
    • Discuss with Amilcar with merging impacts
    • It’s in ardupilotmega.xml
  • Video vs photo
    • Collapse into one bit
    • Bad-exposure flag?
      • Drop it?

UTC2350 - using both Python2 and Python3 in CI!

  • Need something in Wiki about how to set this up on your own branch

UTC2352 -

  • Remove devcall topic
  • No consensus on changing the bit meanings in ArduPilot

UTC2353 -

  • Joypad support
  • HID support based on APM for joystick support into mission planner
  • Not ArduPilot-specific
  • Doesn’t work on boards we support
  • Not tested
  • Separate repo?
  • In the same way that is not in ArduPilot
  • JW: We have some customers that make custom joysticks to connect to computers
  • [12:00 PM] (Channel) JW: They buy these potentiometer and button to HID converters
  • JW: That seems like this serves the same purpose
  • JW:

UTC0001 - revising devcall structure

  • Should continue this on the email list
  • Communications wasn’t great
    • Two people thinking they were hosting the call wasn’t great
  • Rotating host?
    • Craig’s been very, very busy
    • Scheduler is now back to something normal again
    • Hosting call is not a time burden
  • CE was first person to be paid to work on AP
    • His job was to build a business/community so others could be paid
    • Now not paid
      • AP doesn’t have the budget to do it
      • Has been doing it as a volunteer for quite some time
      • No intention to stop
      • No expectation to get paid
  • Do we need to change things up?
    • Make sure we push this project forward
  • Expose more people to the way things work
  • Set of people willing to run call
  • Roster system
  • Common way of getting in what devcall topics are
  • Is there general support for rotating the call chair
  • Chairman gets very involved in call
    • Had four people volunteer
  • Should we share the workload?
    • Continuity?
    • Defer to Craig?
  • Split the call into admin and technical?
    • Complaints about getting into the weeds?
    • Just put technical stuff up the front of the call?
    • Issues at beginning of call?
  • Separate “org” meeting?
    • Two calls a week not really going to work
      • Timezones etc
      • Tridge is going to run out of slots for calls!
  • Admin at start, max 45 minutes?
  • JP: As an idea
  • State of the union start of call should probably be at start of call
  • PRs linger because we run out of time
    • We never discuss them
  • First half of meeting presentations rather than discussions?
  • Dead air?
    • (Channel) MdB: There is a lot of dead time of people not killing their topics once done (or thinking it was self evident it was done but others unsure)
  • Issues and PRs at beginning?
    • Fixed period is a problem
  • JW: Just put the prs in front
  • LV: 175 PR’s
  • TP: 891 issues
  • As soon as something gets the devcall topic it goes on the agenda ATM
  • Peter is the best note taker in history
  • Peter to create a dev team mailing list email to continue this discussion