Session Handoff [hw-cli]: dev 75543d4 merged, hw reworked to upstream AGENTS.md and hardware-tested; waiting on fork sync and Pieter's go #58

Next steps, in priority order

1. Pieter: Sync fork. The fork's dev is at 290b0c2, 43 commits behind upstream dev 75543d4. Until it is synced, Add hw hardware-report CLI command (iteration branch) #1 and Add the hw hardware-report CLI command (rebased, hw CLI only) #16 show upstream's newer commits in their diff, and Copilot reviews those. The agent's fast-forward push of dev is refused, so this step is his.
2. After the sync, request Copilot on Add hw hardware-report CLI command (iteration branch) #1 and Add the hw hardware-report CLI command (rebased, hw CLI only) #16 (pr_review.py --repo ptr727/liquidraver-ZephCore --request wait <n>), and CodeRabbit on Add hw hardware-report CLI command (iteration branch) #1 (@coderabbitai review, at most one an hour). Answer every finding.
3. Fix the one known remaining finding on the next push: HardwareReport.cpp:148. The comment on HW_I2C_CHIP says "enabled", but the status test is in the DT_FOREACH_STATUS_OKAY_NODE loops, so drop the word. That push gets a fresh recorded local pass, and the change goes into Add the hw hardware-report CLI command (rebased, hw CLI only) #16 the same way (git checkout <#1 head> -- <paths>).
4. hw CLI upstream: wait for Pieter's go. When he gives it, cut a clean branch from the then-current dev, squash work/hw-cli onto it, check git diff --shortstat equals git diff -w --shortstat, and open the upstream PR only with his explicit say-so.
5. If dev moves, merge it into Add hw hardware-report CLI command (iteration branch) #1, then copy the result into Add the hw hardware-report CLI command (rebased, hw CLI only) #16, as this round did.
6. 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: Sync fork, and his go for anything upstream, any force-push of a clean branch, and any branch deletion.
• CodeRabbit: one review an hour per repository, and it must be requested.

State (re-derive: git worktree list, gh pr list, local_review.py status --target upstream/dev)

• Upstream dev = 75543d4. Fork dev = 290b0c2 (behind).
• No upstream PR of ours is open. rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101, rtc: write the RTC time only for years 2000-2099 liquidraver/ZephCore#105 and rtc: clear the RV3028 12-hour bit at boot liquidraver/ZephCore#106 are merged; rtc: write the RV3028 time with an A0h year mark, and only 2000-2099 liquidraver/ZephCore#103 and rtc: select 24-hour mode before a time write, and take no time in 12-hour mode liquidraver/ZephCore#104 are closed and replaced.
• work/hw-cli (Add hw hardware-report CLI command (iteration branch) #1) is at 06de0ca in ~/repos/zephcore-hw. hw-cli-only (Add the hw hardware-report CLI command (rebased, hw CLI only) #16) is at 8c800f6 in the hwcli worktree. The two trees are identical.
• Recorded local passes on both at content digest 2a1d2c6a, against upstream/dev (three passes this round, raising 22, 5 and 1 findings).
• Build base: the r7-base worktree holds Zephyr 11808e9 (upstream dev's manifest) with its own modules, patched. ZephCore patches <app>/../zephyr, not ZEPHYR_BASE, so build from a checkout whose own zephyr/ is that tree. This round checked out the commit under test, detached, in r7-base. r6-base (Zephyr ac03a4a) is kept but no longer matches dev.
• Hub tooling ran from a fresh detached hub worktree, ProjectTemplate-zc-hwcli, at origin/main 59e0ec0.
• Hardware: USB-A (RAK4631 with RAK12002 RTC, GNSS and one environment sensor) runs a repeater debug build of 06de0ca. Its prefs are unchanged: the agreed test frequency, tx 1, and adverts off.
• Fork issues filed this round: Companion contact load takes a stored out_path_len unchecked; an invalid one sends replies as zero-hop direct #56 and CMD_ADD_UPDATE_CONTACT stores the app's out_path_len unchecked #57 (unchecked stored out_path_len, found for MeshCore parity). Nothing to do on them without Pieter.
• Closed this round: rtc: write the RX8130CE weekday one-hot #45, RTC save writes the weekday in binary; the RX8130CE's WEEK register is one-hot #39 (rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 merged); RAK4631 never uses a fitted RAK12002 RTC #4, RV3028 loses the time on a power cut: ZephCore never enables its backup switchover #33 (fixed by rak4631: use a fitted RAK12002 RTC, and keep its backup switchover enabled liquidraver/ZephCore#98); RTC probe: field checks skip real RTCs, and only the RV3028 has an identity mask #38 (fixed by rtc: identify RTCs by documented always-zero bits; stop skipping real ones liquidraver/ZephCore#99). The rtc: write the RX8130CE and YSN8900 weekday one-hot liquidraver/ZephCore#101 branches and worktrees are deleted.

The parked decision queue

0 issues carry decision on this fork.

What the last round did

• Merged upstream dev 75543d4 into Add hw hardware-report CLI command (iteration branch) #1 and Add the hw hardware-report CLI command (rebased, hw CLI only) #16 (8bcc79c and c72582b). The conflicts in ZephyrRTCDiscover.c/.h came from rtc: write the RTC time only for years 2000-2099 liquidraver/ZephCore#105 and rtc: clear the RV3028 12-hour bit at boot liquidraver/ZephCore#106 and were resolved keep-both. hw records a candidate as present, then runs dev's 12-hour clear.
• On Pieter's decisions: hw i2c scan now probes with a one-byte read rather than a zero-length write (upstream AGENTS.md: never write to an unidentified peripheral). The code was trimmed to upstream's comment and size rules (c167178, 06de0ca).
• Builds with 0 warnings: native_sim, rak4631 repeater and companion, thinknode_m1 companion, heltec V4-R8 repeater (sysbuild). Host tests: 114/114.
• Flash against dev: repeater +4,436 B, companion +4,548 B, RAM +0.
• Hardware pass on 06de0ca: evidence in Add hw hardware-report CLI command (iteration branch) #1 issuecomment-6103127062.
• Answered and resolved 8 Copilot threads: 7 on upstream code that is in the diff only because of the stale fork dev, and 1 on upstream rtc: identify RTCs by documented always-zero bits; stop skipping real ones liquidraver/ZephCore#99's identity rule.

What not to repeat

• Building one checkout's source against another tree's ZEPHYR_BASE. ZephCore's patch step targets <app>/../zephyr. Building zephcore-hw/zephcore against r7-base's Zephyr compiled an unpatched tree and pointed the patch step at the shared tree.
• A size claim without a baseline build. The first version of the round comment gave an unmeasured flash delta, and it was corrected in place. Build dev first.
• cd X && (a) & (b) & wait. The cd binds only to the first background job.

New learnings

• usbdev nrf52 touch reached USB-A's bootloader on the first try on server-3 (it missed 13/13 on the old WSL host). The note is in Pieter's host runbook and memory.