Peter Linux boards get hwdefs now. Instead of boards.py.
Some boards have small binary output changes.
The Aero board now compiles again.
For new boards you still have to modify RCInput.
Strangely some compilation symbols have been renamed.
Andrew Approved!
UTC0714
A We are getting a DMA timeout on the DShot. - We then call the VT, and eventually RCOutput.cpp350 sets dshot_waiter to null. But VT is not reset.
This is the only place the waiter could be null.
Andy Oh, it’s the only place we don’t check if the lock has expired.
Peter The dma_cancl() should be resetting the lock, no? A Perhaps the (!mask) check if failing.
Let’s wait for Feruccio to test my PR.
Andy I think I can see the race condition. A Andy, can this happen to 4.5? Andy Probably, we haven’t done changes in BDShot in 4.6. A For the record, Ferrucio was getting the issue after arming, IIRC.
UTC0736
A Approved. P Merged!
UTC0740
A That looks like an actual bug. Can we reproduce it and test this PR?
Approved but needs testing.
UTC0754
Andy I’ll ping Randy offline.
UTC0755
Sid I still have some review comments to merge.
UTC0800
Andy Fixed what was requested. P There’s one more request to fix a comment.
We’ll merge after these short fixes.
UTC0804
Andy Your suggestion seems to work, tridge. A Please also remove the now-unused code.
Would be nice to have some performance figures for the distance-from-fence calculations. For many fence points, it might bog down other parts of code.
They could also be on IOCallback.
UTC0815
P This is a step towards removing AP_HAL/RCinput.
Let’s ask Thomas to test.
A Let’s make sure he makes use of RCInput, not overrides.
UTC0822
Andy The board picture has been updated. A Let’s ask Henry to approve.