START_RX_PAIR and REQUEST_AUTOPILOT_CAPABILITIES do not send COMMAND_ACK

[DonLakeFlyer]

This causes QGC to go into its command retry logic since it thinks the command was lost on the wire. There may be other COMMAND_LONGs with this problem as well that I haven't hit with QGC yet.

[DonLakeFlyer]

Also just hit the problem with MAV_CMD_PREFLIGHT_CALIBRATION

[DonLakeFlyer]

Two options:

1. If someone can do a code inpsection of firmware and give me the list of command which have this problem I can add a hack to QGC to not run retry logic for these commands.
2. If printf() for %f is broken since one of the last merges #1 is not possible I'm going to need to turn off retry logic for all commands when running PX4 firmware. Kind of a big loss from a hardening over lossy connection standpoint.

[mhkabir]

I think we should just make sure we properly ACK all commands.

[DonLakeFlyer]

I agree but QGC needs to support firmware version which will not have the fix.

FYI: I already implemented #2 in QGC since at this point I had no other choice and things were fairly broken without it.

[LorenzMeier]

@DonLakeFlyer Can you revert QGC and test this instead?
#6154

Either revert QGC completely or if you want to support old FW versions do a version check. Don't keep things broken just to support old Firmware builds - that's not our strategy for PX4.

Also please fix QGC to ask for an ACK. The commands QGC sends are not asking for one yet.

[DonLakeFlyer]

Also please fix QGC to ask for an ACK. The commands QGC sends are not asking for one yet.

What is the deal with COMMAND_LONG.confirmation? Docs say:
0: First transmission of this command. 1-255: Confirmation transmissions (e.g. for kill command)

Which says nothing about it controlling whether you want an Ack back or not. According to current doc setting confirmation=0 would seem like the right thing.

If you want to force people to latest master and next release then I can revert QGC otherwise I'll need version check. Up to you.

[DonLakeFlyer]

Also FYI: Not running retry logic doesn't really break things. It just is less user friendly to lossy connections. Up until a few weeks ago that was the way QGC always ran.

[LorenzMeier]

Sorry, we had repurposed the confirmation field somewhere along the path by changing the comment. Yes, all commands should be confirmed.

Regarding version check: It would be great if you could condition it on ver 1.5.2 - anything including 1.5.2 does not use it, anything above 1.5.2 and dev should use it.

[LorenzMeier]

This is fixed.