Dev Call Nov 16, 2020

Issues & Pull Request

Moving snprintf




1 Like

Attendee count (max): 21

UTC1100 -

  • Fix nullptr dereference on boot
  • Trivial fix

UTC1101 -

  • Fix 3.6 -> 4.0 changes due to spool-up changes
  • Allow user to spool up in guided mode
  • API changes will be discussed with tridge

UTC1113 -

  • Tailsitter bug fixes
  • Use stopping point for target altitude
  • Merged!

UTC1115 -

  • Run Replay with EKF2 against EKF3 data even if EKF2 wasn’t configured (and vice-versa)
  • Peter will merge later today if he doesn’t spot anything wrong

UTC1117 -

  • Fix script scheduling
  • If you say 1000ms then you probably want to be called at 1000ms intervals
    • ATM you would get 1000ms + your script’s consumed time
  • Merged!
  • Would be nice to have a hook for “next IMU data update”
    • Currently you have to poll the last-update-ms

UTC1122 -

  • Really useful when using dual-ported-USB
  • Passthru lets you pass one serial port through to another
  • This fixes the baudrates
  • Merged!

UTC1126 -

  • Moves buffer handling to common classes
  • Template class for each buffer type separately
    • Same complex code 20 times
  • Saves a vast amount of flash
  • Lots of scope for naming arguments
  • Initialisation fixes need to be unfixed and changed to use designated initialisers
  • Peter will look through it some more
  • Reset_history might go away
  • Returns pointers into buffer so you need alignment
  • Isbaseof std type-traits
  • Renamed pop to get oldest element

UTC1140 -

  • Watchdog resets due to out of memory on omnibus when mounts enabled
    • Tridge to work with Henry on that
      • Perhaps it was the RSSI thing that’s already fixed?
  • Tridge wants to cut down memory in other ways

UTC1143 -

  • Memory guard
    • Only on compile-time option to help find problems
  • External navigation was causing overwrites early in uptime
    • H7 64kB of memory at address 0
    • Low-memory is valid on H7
    • We offset the heap to 1024
    • Buffer overwrite would run past the 1kB
    • Starting to obliterate core structures
  • This PR also checks the first 1kB to make sure it hasn’t been written to
    • Internal error if it is found
    • H7 only
  • Can’t zero first four bytes…
  • Get Sid to review….
  • Use a non-zero fill value?
  • Backport?
    • Could happen with external position
    • Could happen with OpticalFlow
    • Is marked for backport
    • 200kB log buffer change recently might have made this more likely

UTC0001 -

  • Ignore yaw input in automatic modes on Copter
  • New RTL_OPTIONS input
  • Stick-mixing….
    • Coming after S-Curves?
  • Can be merged after comment is added

UTC0007 -

  • Load fonts from filesystem
  • FONT3.bin on SD card replaces font3
  • Linux would be relative to where you started ArduPilot
  • Peter will merge after reading through

UTC0015 -

  • Add safety limit for VTOL transition battery
    • Prevents stalls
  • Hover-throttle based
    • No battery compensation
  • Not using AP_Motors so no hover throttle compensation
    • PH has a patch for that
  • This is still a valid patch
  • Entire back-transition uses bigger of hover and trim throttles
  • Don’t want the thing to climb…

UTC0021 -

  • Stop hard-coding GCS timeout
  • There’s a pattern in SITL which allows for having multiple tables in the one object
  • Doesn’t have to be the same name when you make the macro
  • merged!

UTC0024 -

  • NMEA GPS with vertical velocity
  • L1/L5/Multiband
  • Sacc, vacc
  • Dirt-cheap
  • $PHD - private extensions
    • Mimics uBlox structures in NMEA format
    • Tridge has the datasheet
      • Can do the safe-config stuff now
      • Can set the rate etc etc
  • Amazing value for money
  • Comparison flight coming up
  • Roughly M9 in static testing
  • MdB will approve or merge or let tridge know

UTC0026 -

  • Airspeed auto cal only in loiter or circle

  • Probably just keep the enable/disable switch

  • Autocal doesn’t build models of the airspeed

    • Assumes constant wind
    • Copes with you flying straight ahead
    • Doesn’t cope with systematic changes e.g. updraft on a hill on one side of circuit

UTC0038 -

  • Try to protect about bad sensor calibrations
  • Giving users more options can be bad
  • Disable airspeed cal at boot
  • Use the existing don’t-offset-cal-at-boot parameter
  • Add an rc option to do offset calibration
  • Might need a skip-but-require-prearm enumeration option
    • Perhaps fail to arm if you have cal-on-switch and don’t use it
  • Check delta from previous boot?
  • Check delta from an absolute value?
  • We’ll get an expected-offset and a margin
    • Will need to set it to a non-zero value

UTC0055 -

  • Tell people to contact devteam on internal error
  • Link to how-to-open-an-issue

UTC0107 -

  • Binding dependency for Plane
  • MdB is looking at this
  • PH will chase this some more

UTC0111 -

  • PWM rather than boolean truth on button input
  • Invert button input
  • Attach RC options to

UTC0123 -

  • EKF data source selection
  • Several outstanding review questions
  • Two little real things:
    • Turning off fusing velocity from the GPS
      • Tridge has a PR in for fixing that
    • Slightly resolved / slightly made worse
    • Switching between sources
  • Want to put in the force-EK2 / force-EK3 before merging this one
  • Particular interest for CanberraUAV for flying into/out of buildings

UTC0129 - Copter

  • Another report from a 4.0.5 user
    • RC inputs getting shifted again?!
    • Frsky4b
    • No IOMCU
    • $90 for a crappy receiver
    • CPPM
      • Sum of the channels might be really high
  • S-curves underway still
    • Passes Copter tests
    • QuadPlane and Sub still need to pass
  • Parameter resets
    • Orange backs up 16kB storage onto the SD card in storage.bak directory
      • Tridge wants a zip of that directory
    • Master only
    • Next stable storage should get the backup code
    • Race condition?
  • H7 is just fast?

    • No reason to think this is a hardware fault
      • Happens on multiple different boards

UTC0134 - Plane

  • Nothing to report
  • 4.1 soon

UTC0141 - recording calls

  • Still considered problematic for usual reasons
  • We have searchable meeting notes…

UTC0146 - messing with AC_AttitudeControl

  • Quat-body-to-ned fixes
  • Changes in tailsitter.cpp
  • Probably don’t need an AC_Attitude_Control subclass…
  • Const-reference-vector arguments
  • Accessors are probably not a problem

UTC0157 - close

1 Like

UTC0024 -
What GPS is this referring to?

1 Like