Dev Call Sept 23, 2019 2300 UTC

Issues & Pull Requests




1 Like

Attendee count (max): 19

UTC1102 -

  • Trimming for Rovers
  • Manual, armed and moving
  • Only changing the input side
  • Need to change servo trim
  • Trim then re-neutralize trim on transmitter
  • Won’t work it non-manual modes
    • Will be seen as consistent input
  • Need a line of code for each active mixed output
  • Probably not a good idea for skid steering
  • Connecting real transmitter to SITL is a useful debugging technique

UTC1112 -

  • Add new CPU class 250
  • F7 and H7 small log buffer size
  • Soak up that extra memory!
  • Using CPU speed for memory?
      • Need to set this for SITL….
    • 250-class is defined as having at least 512kB of memory
      • Fast CPUs not much RAM?
      • 75MHz for Maple stuff
    • Proportions of memory would be bad
  • ESP32?
  • davidbuzz: esp32 240mhz, Memory 520 KiB SRAM
  • [9:20 AM] (Channel) MdB: I’m just pointing out that CPU class of speed doesn’t translate well across everything that exists, and when it’s been annoying before

UTC1122 -

  • EKF RAM reduction using common variables
  • Inspired by PR from JC
    • Uses stack
    • 5kB stack frames
    • Loses only static monitoring of stack size we get from compiler as we need to set the number very high
  • This sets up one set of transient data store for all cores
    • They use it in turn
  • Tridge wants this reviewed so he can merge it
    • Peter will look at it later today

UTC1125 -

  • Add loiter gain
  • Thoroughly tested on lake
  • Critical on displacement hulls
  • Cruise on displacement hulls doesn’t work so well
  • Allows option of bypassing calculation and putting true maximum speed of craft in
  • Linear responses to stick inputs in many modes is what this is attempting to achieve
  • Speed_max rather than max_speed
  • Randy will merge after a few tweaks from Henry

UTC1126 -

  • Peter still needs to fix the gtest stuff
  • Want to get the gtest stuff fixed before merging this

UTC1136 -

  • Merged!
  • Eliminates on AP_Buffer class

UTC1136 - safe-disarm

  • Zero outputs can mean ESCs aren’t initialised when you finally send pulses
    • You get beeping ESCs
    • Half-pulses (e.g. on disconnect) can cause a bit of spin
  • Adds parameter to add minimum delay
  • Comment is out of date
    • Only does enable side rather than disable thing
  • These default times are very long
    • Reboot recovery?
    • Blheli needs 2/3rds of a second!
      • Maybe we have a bug in our output layer?
        • Tested with logic analyser to be on the blheli side
        • Arming-in-auto is rare
          • Combined with zero-pwm is even rarer
  • Need to make sure this doesn’t affect spin-up time when not using zero-pwm

UTC1145 -

  • Option bitmask in hold/loiter and at the end of RTL
  • Currently refuses failsafes in hold
  • This adds an option so your failsafes can continue to work in hold
  • Add something so you don’t activate failsafes within some distance of home?
  • If you RTL in surface vehicle you hold
    • In a boat you will circle in a very tight circle at full throttle
    • Option to loiter at end of rtl
  • Separate parameters for accel vs decel?
  • May want to split this into two PRs
    • To match MISS_DONE_BEHAVE

UTC1153 -

  • Copter - fix RC failsafe for no RC receiver
  • Stops fast spinning
  • Could do better with using the failsafe counter
  • Merged!

UTC0000 -

  • Publish nav data to OSD
  • Reviewed already
  • Don’t reference explicit flight modes
  • Where do we put this data?
    • Screen real-estate issue
    • Reusing xtrack error space for loiter distance
  • Use Flightmode object for distance and destination

UTC0011 -

  • AP_LTM_Telem
  • Need to reduce scope on semaphore
  • Need to resolve the issues raised by glacomo892

UTC0017 - Matt and autorotate

  • PR for SITL-only stuff
    • Doesn’t appear on a normal board
  • Large PR
  • Bill has followed up and done testing
    • With enable/disable on autorotation
  • Using IDs above 255 is problematic
  • Can’t go backwards too far
    • Never see anything before 3.2.1
      • Except on apm boards which aren’t going to get updated
    • December 2015 change went in to master
    • Maybe December 2016 for stable release
    • Probably 3.4
    • Anybody running early 3.4 or 3.3 might have problems when upgrading
  • Put it into g2 instead?
    • Yes
  • Maybe in 4.0 we can start using extra parameter IDs

UTC0022 -

  • Extra failsafe bitmask options
  • Replaces continue-mission options
  • GCS failsafe fixes?
    • Out-of-step with documentation?
    • Parameter conversion takes care of GCS and radio failsafe
    • MP updates needed
    • Parameter conversion works
    • Works in SITL
  • Parameter conversion not quite right
    • Need to use “configured”, not check against zero
  • Behavioural change on throttle failsafe needs to be corrected
  • Blog post to announce changes coming

UTC0036 - Plane update

  • New beta release this-morning
    • Out later today
    • Beta2
    • Smallish set of changes
    • Plane 4.0.0beta releases
    • Includes SCHED_LOOP_RATE fixes
      • Not checking for RC inputs in stabilize mode is a problem
      • Oneshot-125 timing problem also found
      • Delay_microseconds to ensure gap
      • 125 delay is only 1/8 th normal length
      • Throttle above certain level would mean running out of time for main loop
    • Leonard is now running @1KHz
    • Sched_debug has achieved loop rate now
    • Uniitalised value in mag fusion in EKF3
    • Enable EK3 as a ride-along
    • track-origins option available for
    • Perhaps we should run valgrind on an autotest?
    • Fill-nanf
      • Catch uninitialised values without running Valgrind
      • Trying to prove Jon’s change was safe
        • Ensuring No reuse of variables between update calls in EKF
      • Need to not prefill when running Valgrind
    • Another couple of weeks of testing before 4.0 is released
    • Backports project added
      • Please add to this things that should come back into 4.0 from master

UTC0049 - Randy and Copter

  • 3.6.11 went out
  • Copter-3.6.11-rc1 available for beta testing
  • EKF fixes
  • Baro spikes issue
  • I2c storm mitigations
  • F405 wing Mark showed it happened when disconnecting a compass. Level shifter
  • Really need to publicise the need to move to the most recent stable firmware
  • Parity checking for commercial S.Bus device broken (inverted!)
  • Add option bit to ignore parity
    • Need to propagate option to IOMCU
  • New Rover and Copter releases
    • Status only moved forward a little bit
    • Complex fences PR needs to be merged
    • Vibration failsafe patches have been PRd
  • New point release for Rover?
    • One with JUST a patch for the i2c storm fix
    • Tridge is glad he did it for Plane
    • Tridge offers to do it for Rover
    • One-line-fix

UTC0103 - clos