Repository navigation
Investigate rising UART framing errors during acknowledged strip streaming #696
Description
With the AVR register-allocation fix in #693, the user physically confirmed the full 100-pixel rainbow works on PD6. Both the original host and updated pacing host can stream and receive ACKs, but the firmware framing-error counter continues rising. CRC errors stay zero and reset count stays stable in the observed stream. This prevents claiming long-run transport reliability or precise concurrent media/effect timing.
Configuration: ATmega328P, 16 MHz, 115200 serial, 100 WS2811 pixels. Each frame uses seven acknowledged staging chunks and one show. Complete frame commands measured 66–76 ms on the observed transport; #693 removes additional idle time after a late transfer.
Next: compare counter growth with strip stopped/running and media-clock activity disabled/enabled, capture bounded sanitized request/ACK evidence, distinguish malformed input from receive-ring overflow/interrupt blackout, and add the focused regression before changing UART pacing. Pealayer-to-host strip RPC and board connectivity were verified; no physical relay/motion or synchronized playback acceptance is claimed.
Activity
atomicdeploy commentedon Oct 30, 2026
PR integration checkpoint: #701 now retains #693 strip operand/waveform/pacing, #697 verified-write restoration recovery/backend propagation, #700 deferred RF decoding, and #694 Urclock-only nometadata policy as ancestor commits, alongside mandatory namespace admission from #704. The deferred-RF fixture,137 tooling/firmware tests and nine targeted Go suites pass locally; no final-image flashing or physical verification was performed in this source-only pass. The50ms/250ms playback limits remain unchanged. Keep this issue open for final-image physical RF/strip/endurance acceptance and distinguish historical ACK/cache/software receipts from new hardware evidence. Fresh exact-head CI and normal merge are still pending in #701; updater work #695/#562 is not claimed fixed.
atomicdeploy commentedon Oct 30, 2026
Source integration completed: #701 merged normally as b8cd35e with24 passing final-head checks, retaining exact #693/#694/#697/#700 heads and their functional tests. Dependency integration #688 then merged as441d89873e5294fb6963c3a2768afc4ee78c3e47 with24 passing final combined-head checks. All10 previously open PRs are now integrated; canonical main is clean and its tree equals tested3310050c. Local combined validation passed48 Go package suites, go vet,32 dependency tests and75 Web files/371 tests; AVR default/menu-label budgets and linked waveforms plus25 Windows VirtualBoard CTest cases passed. No board actuation, flash, runtime update, configuration change or Cafe deployment occurred. Keep this issue and the separate physical/Android/native/endurance/updater gates open; source integration is not fresh physical or production acceptance. Details: docs/verification/pr-integration.md and the merged PR receipts.