CI Workflows
Release Update
Issues & Pull Requests
Attendee count (max): 18
UTC1101 - CI ci · Workflow runs · ArduPilot/ardupilot · GitHub
- RTLSpeed issues
- Test unit tests
- waf race condition
- Compiling same problems multiple times
- Corrupt output file
- Peter making a few things more reliable
UTC1105 - ArduPilot 4.6 issues list · Issue #28612 · ArduPilot/ardupilot · GitHub
- Vehicle updates
- 4.7 release?
- Randy’s been off sick for a couple of days
- 2300 PRs!
- LLMs are helping
- Normally 3 or 4 days
- Want a beta out before next partner’s call
- PH thinks we’re basically ready
- uBlox stuff
- Disabled by default on 4.7.0
- Maybe turn it on by default in 4.7.x for x != 0 ?
- The flash cost is a major impediment
- E.g. f103-GPS must compile!
- Flash cost is way too high
- X20 is probably going to be very popular
- Satellite distributed correction data, notionally
- No RTK base, no ntrip and get RTK accuracy
- Has been available on other GNSS devices for a long time…
- Would be nice if this wasn’t a pay-for service
- Everyone should be able to get cm-level precision….
UTC1117 - press release announcement
- Stephen Dade
- I’ve been working with a satellite modem manufacturer (GroundControl) for the last few months to test their products for compatibility with ArduPilot. They want to put out a press release announcing that their products work with ArduPilot, and I just wanted to check that nobody had any major issues with this line: Ground Control and the ArduPilot Development Team today announced successful testing of MAVLink telemetry performance over the Iridium Certus satellite service
- Any problems with that wording?
- All good!
- Blog post coming
- Latency of 0.7 to 2 seconds
- 2Hz stream rate
- Loss? Once antenna was mounted on house 0.5% over 10 minute test
- Can’t use VPN like wireguard or zerotier as that’s 50 to 80 percent loss
- Looking at about US$150 for 30MB of data
- So about US$40 an hour for a normal 2Hz ArduPilot streamrate
- Cheaper on higher plan
- Ethernet modem
- [10:19 AM]rmackay9: and when it comes to naming, a key red flag is if the manufacturer wants to put “ArduPilot” first in the name of a device or service. They should normally put their company name first and AP second
- [10:20 AM]rmackay9: so it’s not “ArduPilot satellite service from Certus” it would be "“Certus satellite service for ArduPilot” or “Certus ArduPilot satellite service”
UTC1119 - common.xml: add WINCH_LOAD_PAYLOAD enumeration value by peterbarker · Pull Request #477 · ArduPilot/mavlink · GitHub
- Another winch payload enumeration entry
- Merged!
UTC1126 - common.xml: rework MAV_CMD_START_RX_PAIR for type/subtype by peterbarker · Pull Request #479 · ArduPilot/mavlink · GitHub
- Make start_rx_pair look like mavlink/mavlink/master
- Merged!
UTC1128 - .github: add a template for pull requests requesting testing info by peterbarker · Pull Request #32115 · ArduPilot/ardupilot · GitHub
- Pull request template
- MergeOnCIPass
UTC1131 - set_desired_spool_state: Make setter virtual and move tests to the setter by lthall · Pull Request #32130 · ArduPilot/ardupilot · GitHub
- set_desired_spool_state: Make setter virtual and move tests to the setter
- Randy wants to read through it
UTC1137 - AC_WPNav: convert params to SI units and rename WPNAV->WP by rmackay9 · Pull Request #32147 · ArduPilot/ardupilot · GitHub
- Convert parameter to SI and rename WPNAV → WP
- Randy will update description and redo parameter conversion test
- More consistency between vehicles
- Using ACC for accelerations
UTC1144 - Tools: size_compare_branches.py: correct functionality when only buil… by peterbarker · Pull Request #32150 · ArduPilot/ardupilot · GitHub
- Tools: size_compare_branches.py: correct functionality when only building bootloader
- Merged!
UTC1145 - SITL: Fix the --param (aka `-P`) cli option by hunt0r · Pull Request #32158 · ArduPilot/ardupilot · GitHub
- Fix –param cli option
- Poked Pierre on whether we should just delete it
UTC1151 - AP_Compass: just return rather than panic when too many compasses by peterbarker · Pull Request #32197 · ArduPilot/ardupilot · GitHub
- Just return instead of dying with too many compasses
- Merged!
UTC1152 - Copter: Fix AHRS Trim scale problem by lthall · Pull Request #32233 · ArduPilot/ardupilot · GitHub
- Fix AHRS trim scale problem
- Relying on max values
- Also disallow it in mode-acro
- Merged!
UTC1157 - AP_Scripting: add EDGE Autonomy serial EFI driver by tridge · Pull Request #32235 · ArduPilot/ardupilot · GitHub
- Add EDGE Autonomy serial EFI driver
- Tridge will address some of PH’s comments
- Fix stale servo telem data issue from Periph
- Merged!
UTC0001 - Plane: available modes: populate user selectable property by IamPete1 · Pull Request #32271 · ArduPilot/ardupilot · GitHub
- Populate available modes selectable property correctly on Plane
- Merged!
UTC0005 - Plane: add Loiter to QLand and Autoland modes to `FLTMODE_GCSBLOCK` by IamPete1 · Pull Request #32272 · ArduPilot/ardupilot · GitHub
- Add qland and autotune into FLTMODE_GCSBLOCK
- Merged!
UTC0009 - AP_Terrain: check for old DAT files in pre-arm by tridge · Pull Request #32275 · ArduPilot/ardupilot · GitHub
-
Check for old DAT files in pre-arm
-
All dat files have been tweaked on the server
-
There’s a version-minor field now set to 1, was zero
-
Pre-arm failure if it reads a block off the SD card that doesn’t have the value 1
- Users with existing data will get this warning
- They will need to wipe and re-download or set the option bit
- Least worst option
- New terrain data is backwards compatible
- Byte is excluded from the checksum
-
SRTM3’s directory structure is really painful
- Continent structure
- Can’t fix it because MissionPlanner assumes there’s a directory structure there
-
FTP delete doesn’t work from MissionPlanner?
-
Some discussion on whether we should even try to reuse the old terrain data or just up and break everything
- E.g. TERRAIN2
- No terrain data when you upgrade from one to the other
- Then an option bit to use old terrain data
- Rename TERRAIN to TERRAIN.old?
- Moving between 4.6 and 4.7 and back to 4.7 would be a problem
- Iif only if you set a second option bit?
- Breakage happens at upgrade time
- Word will get out that you need to do this
-
Old hgt files cached within MissionPlanner?!
- Different bug
- Srtm3
-
Terrain server converts hgt to dat
-
Or aircraft request data from GCS, GCS gets hgt file from cache or terrain server
- Cached hgt files could repopulate the SD card with bad data
-
11:13 AM]rmackay9: Pre-Arm Safety Checks — Copter documentation
-
Let’s put the message here ^^^
-
[11:14 AM]Tim Tuxworth: Thanks - I’ll make a point to try to “spread the word” - it will be great to refer to a technical article/explanation from @AndrewTridgell !
-
[11:16 AM]Tim Tuxworth: Sure
-
[11:17 AM]Robert L.: Is there any interest in talking about the probable Stack Overflow I had?
-
[11:18 AM]Craig.Elder: We’ll have off the floor topics shortly
-
[11:27 AM]Tim Tuxworth: It could be in the prearm message “terrain2 has no valid data”
-
[11:29 AM]Peter Barker: @Thomas rules No, because currently we are relying on a run-time check to fail, which could happen in flight
-
[11:31 AM]Thomas: i don’t follow. the preflight check for the necessary blocks will run and it is guaranteed they will fail.
-
[11:31 AM]Peter Barker: The pre-flight check is incapable of finding all the bad blocks. tridge covered that when he explained just how much data there is.
-
[11:32 AM]Thomas: any block it finds is bad, and the recommended user action will delete all blocks
-
[11:32 AM]Peter Barker: e.g. guided-go-to-over-there-altitude-80m-relative-to-terrain
-
[11:32 AM]Peter Barker: Again, you’re talking probabilistic.
-
[11:32 AM]Peter Hall: That is a common use case that would always bypass the proposed pre-arm.
-
[11:32 AM]Thomas: why would you be flying terrain without any terrain blocks at all? wouldn’t then we fail because the terrain data is missing?
-
[11:32 AM]Tim Tuxworth: yes agreed @Thomas - ftp to download terrain is so bad as to be unusable. I always pre-load terrain by putting the SD card in my PC
-
[11:33 AM]rmackay9: what’s the cost of the check and do we ever get rid of it?
-
[11:33 AM]Thomas: if they did guided go over there they would see the problem once they landed
-
[11:34 AM]Peter Barker: My understanding is that we are stuck with this forever more.
-
[11:34 AM]Tim Tuxworth: For people who don’t use mission planner, using ftp rename is not a good answer.
-
[11:34 AM]Thomas: the check is very cheap
-
[11:40 AM]Bob Long: I think we’re too hasty to conclude that AP renaming the folder automatically when it detects old data (to terrain.old) is off the table. A beta tester who switches to 4.7 and then back to 4.6 and finds their terrain gone (in a renamed folder) probably should go redownload their terrain cache no matter what. Then, as long as they have new good terrain in their terrain folder, then they can switch between 4.6 and 4.7 at will without issue. With terrain2, you would need to keep two identical copies to bounce back and forth
-
[11:42 AM]Bob Long: And for those using auto download, the whole thing is totally transparent to them
-
[11:45 AM]Bob Long: Mission Planner redownloads after 7 days. We’ll be fine next week
-
Should we move the GCSs to downloading “our” files?
- Uses them for display and for autopilot data
- Maybe we should stop doing the conversion between hgt and dat and just use the dat files directly
- The format that goes between GCS an dautopilot is an efficient encoding
-
If we want to do a more extensive change then someone else will need to look at this
-
Pushing people further towards grabbing data from the terrain server is a good idea
-
[11:49 AM]Bob Long: Yeah, Planner automatically uses whatever terrain source you have set up: .hgt, dted, or geotiff. Both for the MAVLink autodownload protocol and for .DAT generation (though we need to fix its .DAT generation now)
-
[11:49 AM]Tim Tuxworth: Yes I will be testing for sure!
-
[11:51 AM]Thomas: what fix? it probably has the double step bug too
-
[11:52 AM]Thomas: (mavproxy definitely does…)
-
[11:53 AM]Tim Tuxworth: Do we know if QGC does the height file/caching thing?
-
[11:53 AM]Bob Long: Yeah, that stuff. It was ported from whatever the python scripts were doing. I need to look into the bug and fix in more detail. I’ve only been loosely following.
-
[11:54 AM]Thomas: it’s not that severe, but it would be appreciated
-
So what’s the decision?
- PR as it stands
- A TERRAIN2 directory
-
We’re going with the PR as it currently stands
UTC0114 - off the floor
- Rob had a crash running on Andy’s small-fast-copter branch
- He’s got information on it
- Maybe a stack overflow?
- Maybe we should always build with -g ?
- We can test this on the weekend at SV
UTC0033 - a hello to Hunt0r
- Trying to build a freelance software career for himself
- Just saying hi and introducing himself
UTC0033 - close
1 Like