Servers by jDrones

Dev Call Oct 29, 2018 2300 UTC


(Craig Elder) #1

ChibiOS

  • iofirmware,
  • timeframe for removing PX4NuttX and PX4Firmware

Issues & Pull Requests















Plane

  • Release Update

Copter

  • Release update

Rover

  • Release Update

AUVSI

  • Planning

(Craig Elder) #2

Attendee count (max): 17

UTC2300 - ChibiOS migration

  • https://github.com/ArduPilot/ardupilot/pull/9651
  • ChibiOS on IO firmware on PixHawk and related boards
    • Once in, we don’t need PX4 NuttX or PX4 Firmware
    • Needs to be flown and tested
  • When do we switch and remove those other repos
    • Next Plane will be ChibiOS only
    • Next Copter will be ChibiOS only
    • Next Rover will be ChibiOS only
    • Next Sub?!
  • MdB: plane needs to keep PX4 support until actual feature completion is present
    • Why remove it rather than just not giving it new features?
      • Maintenance cost
      • Want to evolve IO protocol
    • Need feature parity!
      • In particular, mixing support in fixed wing support, getting manual takeover
      • Also a servos-to-extreme problem present last time tested
  • People are also sticking with Makefile because it’s the easiest path
    • Not moving to waf because the easiest path is Make
    • Same thing with ChibiOS!
    • End users use stable and beta releases
      • Most people don’t fly out of master
      • Devs should be pushing the edge
    • Commercial users want to see flight time on ChibiOS
  • [10:07 AM] (Channel) DB: here’s the list of things that don’t work yet…
  • https://github.com/ArduPilot/ardupilot/issues/8109
  • Mixer failsafe is the big one
  • Current mixer failsafe isn’t great
    • We were going to do another one
    • But if it is a sticking point then we could go back and reimplement the old one
  • How do we get flight time on ChibiOS?
  • [10:08 AM] (Channel) LV: commercial users tend to use Release builds, and couldn’t care less
  • [10:09 AM] (Channel) LV: if it is chibios or nuttx they don’t care as long as it does what they want.
  • [10:10 AM] (Channel) DB: "Not respecting BRD_SAFETYENABLE and BRD_SAFETYMASK intermittently "
    • Believed fixed
  • IO outputs work regardless of safety switch (BRD_SAFETY_MASK)
    • Believed working by tridge
    • Believed non-working by others
    • Tridge will test
  • How do we get people to test ChibiOS when when they test they simply drop back to NuttX and don’t report their problems?
  • When do we remove?
    • Not today!
    • Soon!
    • If there are remaining issues people should start pushing hard on them
    • We switched to waf years ago but we have core team members who have issues with waf and stop us from moving solely to waf
    • If you have any issues, put them on https://github.com/ArduPilot/ardupilot/issues/8109
    • [10:17 AM] (Channel) LV: Now would be great. Been flying Copter and TradHeli Chibios ever since and no major issues
    • Tridge would like to do this within a month
    • MdB isn’t keen on trying to maintain a fork if we remove support for ChibiOS in master
    • MdB wants to get through a major release with ChibiOS as the main
      • MdB: Not feature complete!
      • Tridge: Neither is NuttX! Lots of features on ChibiOS not on NuttX
        • Mixer is only missing feature as far as tridge can see
        • Peter: Pwm-based rangefinder (see Peter’s PR)
    • Tridge is happy to put time in to push this forward
      • Wants an aggressive timeframe to move onto ChibiOS
      • Wants a gate - when can we remove the old one
        • The only thing out there is the fixed-wing mixer on Plane
        • A point release for 3.9 would meet that requirement?
        • Stop pins for analog stop pins
          • Can be done in hardware
          • It’s a nasty piece of code
          • Layering violations galore
          • Randy thinks we can do without it
    • Fmuv1?
      • Not done yet
      • Randy thinks we should retire this hardware
      • Tridge could do the port
        • But not a priority
        • The IO-board side is hard because of a uart/i2c weirdness / shared-telemetry-radio thing
      • There are low-cost boards out there based on fmuv1
      • This could be done if anybody ever complains
  • Flight time hours on ChibiOS?
    • We have a lot!
    • Mostly due to the Mission Planner interface!
  • Tridge will be putting effort into the mixer stuff
    • If MdB can put the testing into it
    • Not via old protocol
    • New protocol for mixer support
  • [10:29 AM] (Channel) DB: oreo LEDs for Solo reported as not working on chibios
  • [10:29 AM] To Weekly devcall: @DavidBuzz - wasn’t Matt looking at that?
  • [10:29 AM] (Channel) DB: no idea. the issue is here: https://github.com/ArduPilot/ardupilot/issues/9076
  • [10:30 AM] (Channel) DB: matt “assigned tridge” to the issue, so I think it’s on tridge at the moment.
  • Turning off mixer when rebooting?!
  • Randy: biggest issue is the dev environment
    • Bit like the Make thing
    • Eclipse environment and waf
      • Tridge has asked several times on that
      • Happy to put time into working on his own laptop if someone can help with Eclipse install
      • Or help on someone else’s machine
      • Much, much faster when it is working
    • Randy is happy to work on this any time
  • Tridge’s IO firmware needs testing on a Copter
  • Randy raised the issue with developers that could not set up a build environment after the move to waf
  • [10:35 AM] (Channel) JW: @all I’m fine with chibios, In my situation another stable Sub release is low priority
  • JW: 3.6 will be chibios
  • https://github.com/ArduPilot/ardupilot/pull/9651
  • [10:34 AM] (Channel) DB: “update documentation in wiki to remove all PX4 and/orNuttx references, especially dev build howtos.”
  • [10:37 AM] (Channel) MdB: Are we deleting VRBrain? Because without it I don’t see the difference
  • [10:37 AM] (Channel) MdB: just under 50% of PX4 features are also VRBRain features…
  • Make wrapper
    • Just call waf to do stuff
  • Developers are averse to changes in their development env
  • VRBrain doesn’t build in master
    • Chased Roberto and co for a long time to get them to do the ChibiOS
    • 6 or 8 emails
    • They’ve said they intend to do it
    • Only way this is going to move forward is to delete it
    • [10:39 AM] To Weekly devcall: I know I broke stuff in there :frowning:
    • [10:40 AM] To Weekly devcall: I promised Francisco I’d fix it if anyone noticed.
  • [10:33 AM] (Channel) PR.mobile: Please do not delete PX4!!!
    • Not enough testing
    • Wants a 3.7 release before removing NuttX
    • Randy points out 3.7 is a long, long way off
      • Probably a year
    • There was 6 months of 3.6 beta testing
    • Tonnes and tonnes of flight testing before 3.7 goes out
  • “Guinea pig has already left the barn” – tridge
    • Majority of logs tridge sees are ChibiOS
    • Thanks MO for the MP interface!
  • All new flight boards - our most active testers - use ChibiOS
  • Big proportion of 3.6 testers were ChibiOS
  • [10:42 AM] (Channel) LV: Sorry Copter -> Copter 3.5.0 17-July-2017
  • “We’re on ChibiOS now” - tridge
  • Take-aways
    • Mixing support
    • Peter’s rangefinder
    • Randy and tridge on eclipse
    • Michael on mixer testing
    • Each dev call we should have an update on what’s remaining
    • Peter will flight-test
      • But can really only do straight-up PWM on his vehicle
      • More flight-testing on things like one-short will be required
        • Logic-analyzer says the outputs are fine…
    • Want to get this done!
  • Should this be a 4.0?
    • Tends to lead to scope creep
      • “If we’re doing a major release we should change X as well”
    • [10:47 AM] (Channel) DB: leave the decision on the number change till the very last minute. then maybe make it 4.0 without telling anyone.
  • [10:45 AM] (Channel) eosbandi: My two cents:If Nuttx remains in 3.7, users (mostly professional ones) who like to play safe will continue to use just because it is older than chibios. The only way to make them change is to delete/disable it in the next release.
  • [10:47 AM] (Channel) MdB: @tridge: can you start pinging me everytime you have to do PX4 maintenace work? I just haven’t seen the effort happening (outside of safety switch) so I’d like to just be tracking that better. (I genuinely just haven’t seen the maintance burden at this point and would like to see it better)

UTC2347 - https://github.com/ArduPilot/ardupilot/issues/9649

  • Too many constellations!
  • User finally tripped over it!
  • Should we genetically catch and handle this?
    • Low-latency GPS PR detects if we’re getting the 5Hz
  • DavidBuzz: every single gps configuration thats needed is automatically done , why not drop the third constellation at configure time…?
  • Dropping constellations could lead to a position shift?
  • User actually had fiddled their config in this case

UTC2353 - https://github.com/ArduPilot/ardupilot/pull/9638

  • Tridge is OK with this

UTC2358 - https://github.com/ArduPilot/ardupilot/pull/9634

  • What should the style be?
  • Use the square of the distance where possible
  • Case-by-case
  • There are cases which are the opposite - should be pre-calculating and using that precalculated value
  • Could also cache/expire cache

UTC2359 - [10:59 AM] https://github.com/ArduPilot/ardupilot/pull/9611

  • People are happy for this to go in

UTC0005 - https://github.com/ArduPilot/ardupilot/pull/9610

  • Tridge can see how it lines up
  • Merged!

UTC0007 - one for Leonard… no Leonard

UTC0008 - https://github.com/ArduPilot/ardupilot/pull/9497

  • MdB hasn’t done anything with this
  • Whitespace changes
  • Tridge-requested changes have been done
  • Tridge has approved it
  • merged!

UTC0011 - https://github.com/ArduPilot/ardupilot/pull/9391

  • Tridge will try to get to look at as soon as he can

UTC0014 - Leonard

  • https://github.com/ArduPilot/ardupilot/pull/9569
  • Slew rate has to be done elsewhere
  • Problem is triggered when digging motor out of low-rpm state
  • To do on PWM output is not as straight forward as doing on thrust output
  • On back-burner ATM, needs to be reworked
  • Will remove devcalltopic until it is worked on

UTC0016 - https://github.com/ArduPilot/ardupilot/pull/9288

  • Happy with the concept
  • Can be merged once tests have passed

UTC0022 - https://github.com/ArduPilot/ardupilot/pull/9167

  • More for tridge to work on

UTC0023 - https://github.com/ArduPilot/ardupilot/pull/9121

  • Can be merged once a one-parameter-minimum goes in

UTC0031 - https://github.com/ArduPilot/ardupilot/pull/9062

  • Make it a bitmask or enumeration

UTC0033 - https://github.com/ArduPilot/ardupilot/pull/8816

  • MdB hasn’t found time to test it

UTC0033 - more quadplane issues marked for tridge’s attention

UTC0034 - Plane update

  • No real updates
  • Tridge has been concentrating on system stuff lately
  • Takeoff mode concept
  • MdB quadplane landing changes / improvements / making it easier
  • New stable point release to address some board issues
    • Particularly PixRacer issue Peter discovered
    • Px4 Pro support too

UTC0036 - Copter update with Randy

  • Thanks to everyone for helping with the release!
  • 3.6.1 coming
    • PixRacer LED fix
    • LIDAR fixes
    • Garmin Lidarlite v3 support
      • Autodetect instead of enumeration?
      • Horrendously slow update rate without this fix
      • Read register to see if has the low value?
      • There’s the v3 high-range
    • Beta next week

UTC0039 - Rover update with Randy

  • 3.5 coming
    • Beta soon
    • Sailboat!
    • Objects for sailboat and balancebot
  • Non-gps navigation
    • See Randy video
    • HectorSLAM
    • RPiLIDAR
    • Kinda-works
      • Needs a lot of work!
    • https://www.youtube.com/watch?v=P0Xblybi0aw
    • Controlled-environment is much better than non-controlled environment
    • Attitude-stabilized mount for LIDAR?
    • Move to Google cartographer?
    • Need the message-interval stuff in!

UTC0045 - Cube LEDs

UTC0049 - AUVSI

  • Two candidate booths
    • Good placement
  • Hand-sketch coming for booth
    • To be presented at partner’s call next week
  • Hopefully can get several partners in
  • Only pay-for-play talks available now

UTC0053 - Canberra

  • 23rd and 24th of March as previously discussed
  • Not aligning with Avalon as nobody international said they cared
    • Jeff was thinking of doing both….
    • Craig also finds it interesting
    • Higher risk of hot weather
    • Didn’t fly both previous years due to heat
    • Already three weeks later….
  • Tridge wants to make booking SOON!
  • If you want input on this decision, make sure tridge knows it!