20210120 Dev Meeting
Hamish Willee edited this page Jan 20, 2021
·
8 revisions
General:
- Update from relevant previous calls
- Tagged: issues, PR.
Agenda
-
Doc outdated: Camera Protocol #298 - generic request message - can we drop support for existing cameras. Can I add a protocol bit for this? If not, revert/add note to docs?
- JamesP does not impact APM but likely it is worth making the break.
- Lorenz, probably this is amenable to fixing.
- DONE AO Hamish to ping Lorenz and Gus to find a way forward (added a note here and in slack here)
-
Proposal for more robust COMMAND_LONG response #1558 - command response might be an information type.
- Not an acceptable solution as every single command requires bespoke handling.
- The problem of possibly dropping messages is a direct consequence of the fact that MAVLink is deliberately not a reliable transport.
- The solution is to resend in the application layer. If this is a significant problem we could define a "microservice" for this - effectively just send command, set timeout on ack, wait for second response.
- DONE AO Hamish to post response.
-
DONE 1556 Move docs into mavlink/mavlink repo - opinions?
- All - no compelling argument presented.
- AO JamesP to close with reasons.
-
Component information and events interface #1531
- APM review positive; overall can proceed with development with confidence.
- Two questions for Beat to answer through any channel:
- CompInfo would be useful for configuring whole device, not just MAVLink stuff. Any thought on configuring UAVCAN etc?
- Concern that events definitions might disappear off public servers, resulting in vehicles not working/updating properly. Not mavlink "problem" but would be perceived as such and affect confidence. Possible suggestions are to mandate on-device presence of definitions, or MAVLink hosting. Thoughts please?
- DONE AO Hamish to post above to Beat
- AO Hamish to provide draft docs of the component information microservice.
-
- ArduPilot makes heavy use of mission updates from companion computer, so AP will have to find some way to turn this off in certain networks/devices. But, notes that this is done for historical reasons, and is probably not the "right" way to do these kinds of things
- OK with this to proceed. First implementations will provide a better idea of what works/does not.
- AO James to do final check.
- DONE AO Hamish to update issue with "please proceed".
-
Lorenz did you do Travis fix for building C libraries?
- No
- DONE AO HamishW to add issue.
-
- Looking good - move into the RFC process.
-
Criteria for inclusion into supported language list as discussed here.
- Mavgen languages are supported. In practice the C and Python libs are "best supported" because of demand. However there are increasing numbers of tests and will take bug reports, fixes on others.
- There are non-mavgen languages that have varying levels of support.
- James suggests that we should engage more with them rather than fire and forget.
- AO hamish to add the second Java implementation to table.
- AO Hamish to add a "support statement" to the list.
-
common: add MAV_CMD_CAMERA_TRACK_SCENE (JonasV)
- Adds tracking of whole image. Needed because when tracking a whole image you don't want the highlight rectangle displayed.
- Lorenz. This is really optical stabilisation - semantically different. Rename.
- Requires implementation before can be merged.
-
FTP protocol incompatibilities - append vs overwrite + upload in fewest steps question
- JamesP not yet had a chance to follow this up. Will continue to do so.
Attendees: HamishW, JamesP, Lorenz, JonasV