DEV call 30 Jan 2017 3pm PST = 30 Jan 2017 2300 UTC


  • 3.5.0-rc1


  • AP_Deepstall from Michael is moving forward, he’s almost ready for a PR.
  • AP _Landing refactor is complete
  • v3.8.0beta2 is released
  • Tilt wing progress


10:03am - Randy and Copter
3.5rc is very close
Waiting for build server to sort itself out
Mission Planner is failing to download it

Patch will be required for lg320h
Move to in-tree dirvers may have borked something
EKF failsafe isn’t as sensitive as before
Copters will fly all over the place (fly-away)
Will be fixed for RC2
Noted in release notes

10:12am Tom and AP_Landing
Everything to do with Plane landing has been abstracted into AP_Landing
ATM glide-slope is pretty much the only one
Michael is working on deep-stall landing for AP_Landing
Got distracted with some weirdnesses with height concepts

10:20am: Plane 3.8beta2
Includes changes over the last couple of months
Landing changes from Tom and Michael
Binary tilt-rotor support (CL84)
Seems to be lots of interesting tilt-rotors
FireFly should fly master, but people have been reluctant to test
Two people have said they’re now willing to fly master
Simulator will be changes to reflect a FireFly (servo mapping stuff)
New servo-channel code makes this possible at all!
AT offers to send tridge a Firefly Y6, which tridge accepts

10:30am: issue - -fsingle-precision-constant
Most places in code want single-precision floats
About 4 or 5 lines need double-precision
GPS code might have quite a few more
E.g. pi and e constants need to be double-precision
Tridge things we should have a constants.h and constants.cpp
Stm32 can’t do double-precision in hardware, so if a ‘f’ is missing we end up doing software
-fsingle-precision considered evil
Tridge wants to avoid going through and changing 1.0 to 1.0f
We don’t want Linux code doing double-precision and the stm32 doing single precision; want both platforms using single-precision
Warnings in the ardupilot tree are what we’re trying to eliminate
Tridge points out that a lot of those warnings
We need the constants file because the maths is wrong in several places
Tridge really doesn’t want to force people to add “f” to force single-precision constants
So we’re going to retain the -fsingle-precision-float

10:57am: issue - libiio on Linux builds
This has gone in
Sonar should now work on Bebop

10:59am: issue - NED home vs origin
Planes can be 10m in the air according to the aircraft when they’re on the ground
We have two reference locations
EKF origin set once per boot when EKF works out where it is
AHRS “home location” which can be moved
These are not necessarily the same
Plane forced altitude to be the same between the two
Michael will be creating functions with suffixes _home and _origin to make it clear what you’re getting
Will get a function giving you a vector from home rather than EKF origin
Where do we get home altitude from?
Claim that when EKF first locks on its altitude can be wonky
Paul points out that EKF has about 10 different metrics to ensure it should be good
ATM you can take a home position based on a glitched GPS signal
For these patches we’ll retain GPS
Look at changing it later

11:17am: UAVCAN and px4-v2
fitting within 1MB flash on broken smt32s
Removed it on px4-v2
Users were unhappy
Drops out infrequently used features
We drop out some really obscure features based on that
Weird compasses
Now can fit UAVCAN into v2
But now we’re low on space again
Dropping PX4Firmware?
JC is doing some work here
IO outputs
Compiler versions
LdM has been looking at new versions
5.4.3 does still use up more space
Which is ~20% of our remaining flash ATM
Maybe some default flags have changed
What advantages?
It is supported by upstream

11:29am: Contribution of the month✓&q=label%3A"Top%20Contribution"%20closed%3A>%3D2017-01-01%20
Optical Flow for Bebop2 from JBeraud
APSync from Peter Barker
OLED display
uBlox integration should be in there

Question as to what happens when massive rework is done on a “good idea” PR
Idea is to encourage new contributors, thus the original PR is most worthy of a “Top Contribution” award

11:41am: get-together in Australia
Pow-wow with Australian developers in Canberra
If other people want to come
We can’t pay international airfares
But talk to funding committee if you really really want to come

11:42am: merging ArduSub?
Really need Rusty here
Randy can afford to put the time into it
Thanks to enRoute
Randy calls for a PR so we can get started!
11:45am: APSync
New images out fixing --mav10 issue on RPi and
Also moved to common username/password of apsync/apsyncand hostname of apsync

11:50am: New Mumble server
We’re moving the mumble server to the DO server due to issues during the call