v3.1.3
Updates no longer give up on the first hiccup, and when one does fail it now says
why.
Fixed
- A download was abandoned after ten seconds without a byte. That is
aggressive for a TLS transfer on a device that is also weighing, serving a web
interface and talking to the cloud: a transfer that would have recovered was
reported as a failed update. The window is now thirty seconds. - A failed image is retried once, automatically. The stalls seen so far were
transient. Asking someone to start a two-minute update over by hand, for
something the scale can retry itself in two seconds, is a poor trade. - A stall now reports where it happened and how much memory was left. The one
failure captured before this happened with the heap down around 31 KB, which is
a real lead — and it was only visible because a serial cable happened to be
attached at the time.
Known issues
- The underlying cause of the stalls is not yet proven. The evidence points at
memory pressure during the TLS handshake and transfer rather than at the
network. Updating shortly after a restart, while the heap is least fragmented,
is still the most reliable moment.
Updating
Over the air from Settings > Update, or over USB with
bash scripts/flash.sh --fs. Both keep your WiFi credentials, TigerTag session
and calibration.