@TalesFromTheArmchair@hachyderm.io
Post #983119
2026-03-04 18:48 UTC
@jripley We did exactly the thing described for Electric Imp crash telemetry: if the faulting address in the report was just one bit off from what would be a valid address (and especially if the same unit had done this repeatedly), then the default assumption would be that it wasn't a software issue. On embedded the usual culprit was badly-designed power supply circuitry -- or, in one case, badly-designed example power supply circuitry in the microcontroller datasheet, which was significantly amended after we'd already gone to market :(
Replies (1)
-
@TalesFromTheArmchair@hachyderm.io 2026-03-04 19:03
@jripley (In our case we were pretty sure it wasn't actual bad RAM cells because -- on a system with no MMU -- the actual faulting address and backtrace would be different each time on the same unit.)