Dev Call Aug 19, 2019 2300 UTC

Issues & Pull Requests




1 Like

Attendee count (max): 18

UTC2301 - GSoC update

  • In finish-up phase
  • Akshath’s parameter editor is merged and works nicely
    • Flightmode editor has been added
      • Will be merged in next few minutes
  • AirSim
    • Rajat’s been working on Rover stuff
      • Physics of virtual car is problematic due to speed
    • Waiting for review from the AirSim people
      • Airsim people were expecting to control with their own API
      • Tried a java-based mavlink tool to connect
        • Tridge suspects it’s a mavlink1-only tool
          • Just because we only speak mavlink2 on serial0?
  • Peter H
    • More path planning stuff
      • Thread creation issues
    • Wiki stuff
      • Writing down what he’s learnt so far so he doesn’t forget it
    • Also bought AIS receiver
      • ADSB for ships
      • Sailboats can sail around forever - let’s do that
  • Matt K
    • Porting to master
    • Dealing with various issues to get that sorted
    • Now working on master
    • Prepping to put into a pull request
      • Autorotation
      • SITL-only
        • Entry and guide phases only
    • Fallback to get vehicle down with minimal damage - depending on energy
      • Arduino Rpm sensor wiki PR
        • Feedback on this would be good
      • PR for adding arduino RPM sensor support
    • Hall-effect sensors can be finicky
      • And wire colours don’t necessarily match
      • Tried to explain that in the Wiki
    • Touch a little on other sensor types too
    • Firmware to turn ArduPilot into a simple logic analyzer would be nice
      • Possibly get reasonable timings with a custom firmware
    • Need to get this stuff active and available so it doesn’t die on the vine
    • Matt has a loose roadmap of development going through to roughly April next year
      • Detailed in funding request for hardware

UTC2315 -

  • @readonly marker
    • Linkage with Peter’s commit?
  • For a maximum_range parameter
    • Regulations!
      • 250/300km max range
    • Afs-max-range too
  • Bitmask flag for readonly
  • Send-text vs sending back old value
  • Peter’s PR sends both back to make the GCS happy
    • Tlog will show the value changing
    • Vartype denied as a nack?
  • Peter will clean up his PR and we can get that merged

UTC2331 -

  • Peter will eliminate the method to get the parameters and just get the trivial namespace change in

UTC2337 -

  • No terrain means no scripting
  • This fix is in a generated file, so it does need work anyway
  • Will make having terrain a hard requirement for scripting to work

UTC2338 -

  • Can be pushed in once the comment is fixed

UTC2338 -

  • Can be pushed in once after Francisco’s comment responded to

UTC2338 -

  • Native i2c protocol support for lightware sensors
    • Means no configuration required
  • Max-range needs testing
  • Serial-over-i2c
    • Sharing code would be problematic
  • Use the multi-return feature in future?
    • Currently #ifdef’d out
    • Logging it only would be good
  • Tridge to address Randy’s comments
  • Main risk is that some Lightware speak some other variant of i2c

UTC2344 -

  • Thread creation fails on some builds
  • Cygwin
  • PR is in for it
  • Will be merged soon

UTC2351 -

  • Now uses same file* under SITL as it does under ChibiOS
  • Michael noted an existing issue with opendir
  • If a script fails then the opendir file descriptor can leak
    • Didn’t matter under ChibiOS
    • Accept this and fix it later?
  • MdB wants to test a little further
  • Retries should cover over some losses in flight
  • Indefinite wait time so long as volume label and last_log.txt as expected

UTC2353 -

  • No new test cases
  • Matt’s tried unplugging/retrying the i2c compass with no luck
  • Matt K
    • Had something that sounded similar
  • Why can’t we reproduce?
    • Corrupt build?
      • Very unlikely
    • Some hardware component involved?
    • ArduPilot acts the same at the sensor level armed or disarmed
    • Interrupt storm in i2c?
      • Add defensive code against interrupt storms?
      • More than n interrupts per second it will disable the device (e.g. compass)
      • Counts of interrupts in PM message
        • And in watchdog data
    • Not losing the test case next time…
    • Temperature, loose wires etc?
    • Design vulnerability?
      • Something is marginal?
    • [10:03 AM] (Channel) MK: The vehicle that I was having the issues on had 4 noisy digital servos which lends to the interrupt storm theory.
  • Smbus?
    • Tridge did check the datasheet to see if we checked all the interrupts
  • Tridge will ad interrupt storm handling
    • Will be an internal error
    • Need to know how many interrupts we can expect so we know what the threshold is
      • Probably going to be a significant break
    • How long does the storm have to last for before we trigger?
      • Maybe a tenth of a second
    • Max interrupts per transaction instead?
      • One interrupt per byte is the most you can get
      • And 2 for DMA
      • … and if you get an interrupt outside a transaction you reset the peripheral

UTC0010 - gimbal overshooting under ChibiOS

  • Buffering on the uart?
  • Peter to add more uart logging

UTC0018 -

  • GPS quality time can be measured?
  • Check_scalar is more straight forward
  • Way too many parameters
  • Check_scalar doesn’ts scale with time
  • Divide 10 seconds by the check-scalar ratio
  • MdB will close

UTC0020 - Plane update

  • 3.9.10 beta1 is out

  • Constrained change for IOMCU handling

  • Plane 3.9.10 release

  • Outputs zeroing is bad.

    • Some ESCs might not cope
  • Fixes CRC errors on IOMCU protocol

  • PR is in for Copter too

  • 4.0 release process will start soon

    • Copter release number?
      • Both copter and Rover will move forward
    • 4.0 means no NuttX support
    • … and scripting
    • Build system supporting multiple major release streams?

UTC0022 - Copter

  • 3.6.10 - beta testing soon
  • Copter 4.0
    • Major hold up is stay-out zones
      • Randy’s tested it against Rover successfully
      • Polygon stay-out zones only
        • Circular stay-out zones not supported in Dijkstra / Bendy Ruler
    • Vibration failsafe
      • Partial fix in a branch
  • Rover 4.0
    • Can start beta-testing once stay-out zones are in
  • S.BUS decoding on 3.6.10 hasn’t been followed up on
  • Parameter for turning logic analyser on
    • Something like batchsampler?
    • Maybe log at a megahertz?
    • Do it with DMA could probably get 10s of MHz
  • Randy’s not keen on carrying a logic analyzer around in our code…
    • Feature is cool
    • Maybe as a separate vehicle?

UTC0038 - discussion on vsnprintf and null termination

UTC0041 - close

1 Like