This would continue to alert the operator of an imminent breach.
Merged!
UTC0709
Andrew: This looks like it’s a case to catch multiple begin attempts. Peter: RunCam seems to be doing those re-begin requests. Andrew: The commit seems correct, the commit message is wrong and needs fixing.
Merged!
UTC0716
Peter: The differences in binary sizes are unexpected and strange.
MergeOnCIPass
UTC0720
Peter: VRBrain v51, v52, v53 have gotten their own hwdef files.
They now get dedicated sensor descriptions.
Andrew: We could wait to get a response from Virtual Robotics. Peter: Randy pinged them and they didn’t respond.
We can try. If this breaks them, we’ll fix it agian.
Merged!
UTC0723
Peter: Due to a mistake of mine, the circle fences were being stored as integers.
Merged!
UTC0726
Peter: Still needs review from other vehicle maintainers.
Andy: TBS has another board with a corrupt bootloader.
This apparently happens on bare boards with no peripherals attached.
Matek H743Slim is also affected. By cycling battery power, but USB power might work as well.
Cycle power quickly, don’t let it boot.
Will do a PR for TBS tonight.
Randy: If it’s a significant % that is affected, then we should release a beta with all new fresh bootloaders. Andrew: The reason I’m reluctant, is that apparently there are ECC errors. So if we ignore them and boot, the vehicle can fall out of the sky when the flash sector is accessed while corrupted.
Yes, it could also be that the ECC checking engine is buggy, or that it’s the parameter space is what’s corrupted. But we have indirect proof that it’s not the parameter sectors: If it was the parameter sector, then you wouldn’t be able to re-flash the firmware, because the parameters would still be corrupt.
The normal DFU upload tool doesn’t touch the last two sectors with the parameters. But yes, it can be used differently to reset the parameters too.
This could be a new bug in 4.6, a change in code that causes the firmware corruption.
Randy: We could allow us a time limit to solve the issue, and when it lapses we can pull all of the suspect boards from 4.6.
Let’s also create a Discuss thread highlighting the issue, calling for information from the general public.
Michelle: Then a good fix would be to not erase the firmware but also not boot it. If it’s a false positive, the next power cycle will allow booting. If there’s an actual error, then the firmware won’t be booted and flown. Randy: I have created a Discuss topic: AP 4.6 Firmware Wipes On Boards with Params on Flash
UTC0734
Andrew: This adds a lot of complexity, but I don’t see the benefit. The additional precision doesn’t add any value.
The un-normalization operation assumes that the relative position components are normalized, but it’s not clear what they are normalized to.
UTC0747
Merged!
UTC0747
Andrew: Let’s not nuke Piccolo CAN out of the default build without asking first.