ArduPilot EU Dev call 2024-10-09

Attendees (max): 8

UTC0700

Thomas has been diving into CAN protocols.
Apparently the change doesn’t have byte changes.
Would be nice to make KDECAN part of the custom build server.
It’s a non-negligible flash size save.
To be merged once no size-change is verified.


UTC0708

MergeOnCIPass
Won’t do for 4.6, as Cube Red reports >2M flash, we need a workaround for this.
Possibly via Pr/hal program size limit kb by peterbarker · Pull Request #28362 · ArduPilot/ardupilot · GitHub.
Andrew: Ensure the bootloaders aren’t affected by this.
Peter: This PR overflows some RAM regions, weirdly.
A: Ensure that RAMfuncs are placed on the first section of the map.

  • Too many functions are loaded in RAM and they overflow it.
  • Or it could be too much RAM data.
  • We should check an object dump and see what is there.

Andy: That particular SPRacing H730 is quite unique in that it’s the only one that’s causing trouble and is not widely used.

  • We could turn it into an FPV board and save feature space.

UTC0730

Merged!


UTC0732

A: Would be nice to test with option not set, to prevent false positives.


Andy’s PRs on fast rate: They’ll have to wait for review until after the conference, unfortunately.


UTC0746

P: Saves some space and should make it a lot easier to modify/remove existing flags/states.
A: There is a much more compact way to achieve the same goal, see comment.


UTC0801

A: Perhaps we need to restore the default to nonzero.

  • Asking the reporter to provide a log.

George: The gain is not zero, it’s using the normal IGAIN.
A: Nice. But please check again if we missed an obvious bug.


UTC0812

A: We do have this for ICE.
G: I think the reporter wants this for folding propellers.
A: The option bit is not needed, the parameter can be defaulted to zero.


UTC0815

Merged!


https://github.com/ArduPilot/ardupilot/pull/28389
UTC0823

A: Good one. Please add a backport to 4.5.7 for testing.
P: Some reformatting for legibility would be nice.


UTC0829

A: Not happy with it yet, would like to discuss with Pete.

  • In the case of an inverted configuration, the current suggestion would nose down and lose a lot more height.
  • We might need to move towards what Paul suggested.

UTC0835

A: 0xDD71C118: eyesore of the year award!

  • It would be better as a string compare.

P: The checksum test is more all-encompasing.
A: Then strcmp the format and the labels.

  • Ah no, it won’t work because other non-related fields could have changed and we still want the conversion.