drivers/uavcan: subtree merge last working libuavcan (preserving history) #23819
Conversation
Missing features for LPC11C24
…uilds, but unit tests are failing horribly
The server will now attempt to provide files from the primary
path RootPath and then the alternate AltRootPath. This alows
a layered approch to file service.
Setting the RootPath to an SD card and the AltRootPath to a
ROM FS will allow a user to provide an updated uavcan firware
file that will be used instead of the baked in one without
any copying.
The path as been removed from the interactions with nodes.
There by supporting Non-compliant nodes that could not
handle the specified path length.
The vitrual pathing has been moved into the FDCache.
|
|
dagar
force-pushed
the
pr-libuavcan_subtree_merge
branch
from
fb6b189
to
d800496
Compare
dagar
changed the title
dagar
force-pushed
the
pr-libuavcan_subtree_merge
branch
from
d800496
to
d85ab9e
Compare
…ory) - upstream libuavcan was broken and then marked deprecated, this fully absorbs the submodule (renamed libdronecan to deconflict) up to our last good working commit and all commit history is kept - fixes #23727 (regression introduced in #23113) - this puts us in a much better position to evolve the library as needed now that we have full control
dagar
force-pushed
the
pr-libuavcan_subtree_merge
branch
from
d85ab9e
to
e59cc9a
Compare
|
@dagar will this break CubeID? I guess I can check. And how does this relate to https://github.com/dronecan/libcanard? Should we move over to that one? |
This includes the latest DSDL, so I believe it should work, but yes please check.
I believe there's some interest in that, but to be honest I don't think it's actually going to be any better. In fact I would guess it's probably a lateral move after quite a lot of effort. I don't think the hate for libuavcan is completely justified, there's actually quite a lot it does very well, it's just somewhat inaccessible template heavy c++ code. By bringing it in tree we could potentially keep some of the good parts while dropping unnecessary extras and duplication. For example libuavcan needs its own dedicated timer for monotonic time and we could more tightly integrated it with our own subscription callback and scheduling mechanisms. |
|
@dagar you're right, Remote ID still works. I tested both CubePilot CubeID and the Holybro's AP_Periph based remote ID. |
|
This also saves about 3kB of flash. Thanks everyone. |
Upstream libuavcan was broken and then marked deprecated. This pull request fully absorbs the libuavcan submodule (renamed libdronecan to deconflict) up to our last good working commit and all commit history is kept.
Other than fixing the regression introduced in #23113 the goal here is to put us in a much better position to easily evolve the library as needed now that we have full control.
Fixes #23727