Session Handoff [hw-cli]: hw CLI merged with dev, reviewed and hardware-tested on the fork; upstream waits on Pieter's go; #101 awaiting owner #49

Next steps, in priority order

1. Monitor rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 and answer the owner's feedback, unchanged from round 8 (Session Handoff [hw-cli]: liquidraver/ZephCore#101 awaiting owner review; #98 worktrees retired, shared Zephyr tree re-homed to r6-base #48): fix on work/rx8130ce-week-one-hot first, then re-squash the clean branch with a lease on Pieter's go. Before rebuilding in rx8130ce-week, delete its stale build folders (Session Handoff [hw-cli]: liquidraver/ZephCore#101 awaiting owner review; #98 worktrees retired, shared Zephyr tree re-homed to r6-base #48, What not to repeat). Done means rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 is merged or the owner asks for nothing more.
2. hw CLI: wait for Pieter's go before anything goes upstream. Pieter, 2026-10-05: "continue working on hw cli local branches and local pr only, not ready to file upstream yet before manual testing." The RAK4631 hardware pass is done (see What the last round did). When he says go:
   • cut a clean branch fresh from the then-current dev and squash work/hw-cli onto it, per the hub's upstream-contribution-workflow;
   • re-check line endings: git diff --shortstat must equal git diff -w --shortstat;
   • open the upstream PR only with his explicit say-so naming it.
     Done means a clean upstream PR exists, or Pieter asks for more testing first.
3. If dev moves before then, merge it into work/hw-cli (Add hw hardware-report CLI command (iteration branch) #1) and hw-cli-only (Add the hw hardware-report CLI command (rebased, hw CLI only) #16) as this round did. Keep the two trees byte-identical by taking Add hw hardware-report CLI command (iteration branch) #1's resolution into Add the hw hardware-report CLI command (rebased, hw CLI only) #16 (git checkout <#1 head> -- <paths>). Run a recorded local pass, push, and request Copilot on both. CodeRabbit allows one per hour, and Add hw hardware-report CLI command (iteration branch) #1 goes first.
4. When rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 merges, close fork rtc: write the RX8130CE weekday one-hot #45 and RTC save writes the weekday in binary; the RX8130CE's WEEK register is one-hot #39, and delete the rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 branches and worktrees on Pieter's go. Keep r6-base, which holds the shared Zephyr tree.
5. Hold companion: mark a CLI reply cut to the app frame (iteration branch) #27-companion: never treat an over-the-air CLI command as local (iteration branch) #32 until Pieter says go.

External blockers

• Pieter's go-ahead for any upstream PR (the hw CLI included), any force-push of a clean upstream branch, and any branch deletion.
• liquidraver's review of rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101.
• CodeRabbit on the fork: one review an hour per repository, and it must be requested (@coderabbitai review). Its clean pass shows only as an edit to its summary comment naming the head.

Internal dependencies

• The hw upstream PR touches ZephyrRTCDiscover.c/.h, and so does rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101. Whichever lands second needs dev merged in first. rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 only changes rtc_time_block(), which the hw branch does not touch, so a conflict is unlikely but must be checked.
• Step 4 waits on rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 merging.

State (re-derive: git worktree list, gh pr list, pr_review.py status)

• Fork dev = upstream dev = 108b96d.
• work/hw-cli (Add hw hardware-report CLI command (iteration branch) #1) is at 14e6968, in the primary west-workspace worktree. hw-cli-only (Add the hw hardware-report CLI command (rebased, hw CLI only) #16) is at 29af49c, in the hwcli worktree. Both trees are byte-identical, with dev merged in. Both have recorded local review receipts at content digest 0f64f1d1. Copilot is clean on both heads. CodeRabbit is clean on Add hw hardware-report CLI command (iteration branch) #1's head and was not requested on Add the hw hardware-report CLI command (rebased, hw CLI only) #16 (identical content).
• Build hw-cli against the r6-base Zephyr tree (ZEPHYR_BASE set to its real path). It carries dev's patch set, so the shared tree in the primary worktree is not re-patched.
• Hardware: this session's RAK4631 (fitted with GNSS, the RAK12002 RTC and one environment sensor at 0x76) runs a repeater debug build of 14e6968, handle lpxz, 927.875 / 62.5 / SF7 / CR5, tx 1, adverts off, repeat off. The other RAK4631 (two sensors) belongs to the MeshCore session.
• rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 is open; fork rtc: write the RX8130CE weekday one-hot #45 is at 44ed7e3.

The parked decision queue

0 issues carry decision on this fork.

What the last round did

• Merged dev (108b96d) into Add hw hardware-report CLI command (iteration branch) #1 and Add the hw hardware-report CLI command (rebased, hw CLI only) #16. rtc: identify RTCs by documented always-zero bits; stop skipping real ones liquidraver/ZephCore#99's rtc_identify() conflicted with where hw recorded RTC report state. dev's logic was taken whole and its verdicts mapped to report states. A new RTC_UNREAD keeps unprobed for non-NACK failures. rtc: identify RTCs by documented always-zero bits; stop skipping real ones liquidraver/ZephCore#99's all-0xFF skip became an all 0xff state, and the summary says unsettled. The hw docs moved to the renamed docs/CLI_commands.md.
• Three recorded local passes. The first raised 6 findings, fixed in 76a27a7. The second raised 3, fixed in 14e6968: discovery now publishes states and the adopted chip together under a spinlock, and zephcore_rtc_snapshot() replaced zephcore_rtc_get() and zephcore_rtc_active(). The third raised 0.
• Builds: native_sim, rak4631 repeater and companion, thinknode_m1 companion, heltec_wifi_lora32_v4_r8 repeater. All clean.
• RAK4631 hardware pass on 14e6968.
  • Every hw subcommand, paging, the out-of-range page message and junk-word rejection, and the xx| prefix behave as documented.
  • The RTC restored the time after the flash, and again after a time sync and a real reboot.
  • Evidence: Add hw hardware-report CLI command (iteration branch) #1 issuecomment-6009358478.
  • Not covered: the all 0xff and unsettled paths (this board declares only the RV3028), and hw on the M1 and the V4-R8 (built, not flashed).
• Retired the rak4631: use a fitted RAK12002 RTC, and keep its backup switchover enabled liquidraver/ZephCore#98 worktrees and re-homed the shared Zephyr tree into r6-base (Session Handoff [hw-cli]: liquidraver/ZephCore#101 awaiting owner review; #98 worktrees retired, shared Zephyr tree re-homed to r6-base #48).

What not to repeat

• Hub write scripts refuse to run from an exported copy. pr_review.py wait --request from a git archive export returned OUT_OF_SCOPE, because it derives the owner scope from its own checkout's origin. Run hub tooling from a fresh detached hub worktree at origin/main.
• The round-1 fix introduced the round-2 findings. Staging only s_state left s_active published mid-probe. When a reviewer reports a shared-state race, publish every field a reader combines, together.

New learnings

• None durable beyond this chain. Host facts (board fittings, the hub-worktree requirement) go to the maintainer's host memory.