Session Handoff [hw-cli]: hw hardware-report CLI (#1, #16) and fork issues #3/#4

Next steps, in priority order

Dispose of CodeRabbit on Add hw hardware-report CLI command (iteration branch) #1 and Add the hw hardware-report CLI command (rebased, hw CLI only) #16. Both were pushed and re-requested on 2026-10-02 but CodeRabbit was rate-limited ("Review limit reached"), so neither head is CodeRabbit-reviewed yet. Done means: CodeRabbit reviews head 0aa6b62 (Add hw hardware-report CLI command (iteration branch) #1) and 78b486a (Add the hw hardware-report CLI command (rebased, hw CLI only) #16), and every finding is closed per pr-review-conduct (fix through local-strict-review, or decline with evidence). Partial or unstated coverage is acceptable on this fork.
Re-run hw on hardware. Neither branch has run on a board since dev was merged and the hw CLI was ported. Done means: on the RAK4631 and the Heltec V4-R8, hw, hw N, hw board, hw rtc, hw i2c, hw i2c scan, hw gps, hw sensors and hw all answer as docs/Repeater_CLI_commands.md says, locally and over remote admin, and on a companion over the app CLI (paged to 175 bytes, next:N intact). On the RAK4631 with a RAK12002 fitted, hw rtc shows the RV3028 adopted. Blocked on the boards being attached to WSL (see External blockers).
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 once its quota is back (about 2026-10-06), and close its findings the same way.
Upstream, only with the maintainer's explicit go-ahead naming each one: Add the hw hardware-report CLI command (rebased, hw CLI only) #16 is the presentation of the hw CLI; fork issues Companion CLI replies over 175 bytes are truncated silently #3 (marked reply truncation) and RAK4631 never uses a fitted RAK12002 RTC #4 (RAK4631 RTC) each want their own clean branch cut from the synced mirror tip. Add hw hardware-report CLI command (iteration branch) #1 stays open as the iteration log, never merged, per the hub's upstream-contribution-workflow.

External blockers

Boards not attached. Only the ThinkNode M7 is attached to WSL. The RAK4631 and Heltec V4-R8 need usbipd attach by the maintainer. Air rules for any flash: SoCal parameters, 1 dBm, short random name, minimal traffic.
CodeRabbit OSS review limit (resets within the hour) and Copilot quota (about 2026-10-06).
The maintainer's go-ahead for every upstream write (step 4).

Internal dependencies

Step 2 should follow step 1 if CodeRabbit's findings change code, so the hardware run covers the final head.
Add the hw hardware-report CLI command (rebased, hw CLI only) #16's seven files are byte-identical to the same files on Add hw hardware-report CLI command (iteration branch) #1; a change to either must be made on both, and re-checked with cmp.

State (re-derive at resume)

ptr727/liquidraver-ZephCore: Add hw hardware-report CLI command (iteration branch) #1 work/hw-cli → dev, OPEN, head 0aa6b62; Add the hw hardware-report CLI command (rebased, hw CLI only) #16 hw-cli-only → dev, OPEN, head 78b486a. Both have the fork's dev (= upstream/dev 3c7a7c8) merged in. Add hw hardware-report CLI command (iteration branch) #1's diff against dev is the hw CLI plus fork issues Companion CLI replies over 175 bytes are truncated silently #3 and RAK4631 never uses a fitted RAK12002 RTC #4; Add the hw hardware-report CLI command (rebased, hw CLI only) #16's is the hw CLI alone.
Worktrees (git worktree list): Add hw hardware-report CLI command (iteration branch) #1's is the zephcore-hw checkout, which also holds the west workspace (Zephyr updated to the pinned 1ee3b93134b this round); Add the hw hardware-report CLI command (rebased, hw CLI only) #16's worktree symlinks its zephyr, modules and .venv into it, so the two share one Zephyr tree and must build one at a time.
Reviews: three local-strict-review passes recorded on each head (11, 2, then 0 findings on Add hw hardware-report CLI command (iteration branch) #1; 9, 2, 0 on Add the hw hardware-report CLI command (rebased, hw CLI only) #16). Add the hw hardware-report CLI command (rebased, hw CLI only) #16's Copilot thread (header-only scan line) fixed and resolved. CodeRabbit pending (step 1).
Issue Tracking: hw hardware-report command — state and handoff #10 is the older free-form tracking issue for this work; this chain replaces it for state.

The parked decision queue

This repository has no decision label; nothing is parked.

What the last round did

The maintainer chose to merge dev into Add hw hardware-report CLI command (iteration branch) #1 per the skill (2026-10-02). Merged dev into both Add hw hardware-report CLI command (iteration branch) #1 and Add the hw hardware-report CLI command (rebased, hw CLI only) #16. On Add hw hardware-report CLI command (iteration branch) #1 every conflict took dev's version where upstream had already merged that content (reset cause, the V4-R8 board, its docs note); the two unsubmitted changes (Companion CLI replies over 175 bytes are truncated silently #3, RAK4631 never uses a fitted RAK12002 RTC #4) were re-applied where dev moved the code; the hw files were rebuilt as dev plus Add the hw hardware-report CLI command (rebased, hw CLI only) #16's delta.
Ported the hw CLI to dev (consolidated mesh/MeshCore.h, per-device sensors, removed GPS callback) and fixed Add the hw hardware-report CLI command (rebased, hw CLI only) #16's Copilot finding.
Fixed everything the local review found: replyCap(), paging to the companion's app frame, a resumable summary (hw N), an honest not compiled in, the autodiscovery-disabled report, and docs that disagreed with the code.
RAK4631: replaced the full rtc-i2c.dtsi include with one RV3028 node. The include let the probe adopt an IMU at 0x68 as the clock and write time into it. Commented the correction on RAK4631 never uses a fitted RAK12002 RTC #4; status on Companion CLI replies over 175 bytes are truncated silently #3.
Builds clean on rak4631 companion and repeater, heltec_wifi_lora32_v4_r8 repeater, and the rak6421 native preset.

What not to repeat

A bare native_sim/native/64 build of dev fails in main_companion.cpp; native builds go through the boards/linux_native/ presets (-DEXTRA_CONF_FILE=boards/linux_native/rak6421.conf;...). The cross toolchains are not installed here, so build the preset for the host.
Trusting that #16 still builds on a moved base: it was rebased onto 251269d and broke against 3c7a7c8 on three API changes nothing flagged. Build after every base move.
Trusting a code comment's safety claim: the RAK4631 overlay comment said the probe rejects an IMU, which the probe's code contradicts.

New learnings

None beyond the above; nothing here belongs in governance.

Progress after this link was filed (2026-10-02):

CodeRabbit on Add hw hardware-report CLI command (iteration branch) #1 (head 0aa6b62): one finding. A line too long for an empty page was never emitted, so next:N pointed back at it forever. Fixed in 24728a0, replied to and resolved.
Add the hw hardware-report CLI command (rebased, hw CLI only) #16: carries the same fix as 7dbe59d.
Local review: a fourth pass on each found nothing.
Heads now: Add hw hardware-report CLI command (iteration branch) #1 24728a0, Add the hw hardware-report CLI command (rebased, hw CLI only) #16 7dbe59d. Both are pushed and build clean on the four targets.
Step 1 remaining: CodeRabbit allows one review per hour on this repository. Add the hw hardware-report CLI command (rebased, hw CLI only) #16 has not been reviewed yet (rate-limited), and Add hw hardware-report CLI command (iteration branch) #1's new head is unreviewed. Re-request each when the limit resets.
Step 2 (hardware): the maintainer chose to skip it this session.

CodeRabbit on #16: it reviewed the full diff at 7dbe59d and reported "No actionable comments were generated", through its summary comment rather than a review object. #16 is therefore at its stopping point until Copilot's quota returns. #1's head 24728a0 is next, in the following hourly slot.

CodeRabbit on #1 at 24728a0: one outside-diff finding. A failed RTC read was reported as absent. It is fixed in 69001b7 and answered in #1's conversation. #16 carries the same fix as 0057730, and a fifth local review pass on each found nothing. Heads are now #1 69001b7 and #16 0057730, both pushed. A CodeRabbit pass on these heads is still owed, at one review per hour.

CodeRabbit on #16 at 0057730: two findings, both on the RTC probe's labels.

The -EIO ambiguity is documented in 5108d4d. Treating -EIO as unprobed instead was declined with driver evidence: every driver here reports an address NACK as -EIO.
The PRESENT label after a failed status read was declined. It is dev's unchanged acceptance rule, and the hazard behind it is filed as RTC probe adopts a chip whose status-register read failed, then writes time into it #23.
Two more local passes followed: one found doc inaccuracies, which are fixed; the next found nothing.
Heads now: Add hw hardware-report CLI command (iteration branch) #1 69ae19f, Add the hw hardware-report CLI command (rebased, hw CLI only) #16 5108d4d, both pushed. Each owes one more CodeRabbit pass on these docs-only heads, at one review per hour.

CodeRabbit on #1 at 69ae19f: one outside-diff finding. A re-probe could keep a stale adopted RTC. It is fixed in 99f8cf8 (on #16 as 75c5c52) and answered in #1's conversation. It isn't reachable today: restore runs once at boot. Local review round 8 found nothing.

Heads now: #1 99f8cf8, #16 75c5c52, both pushed. CodeRabbit's last four rounds have each found one smaller issue in the same RTC probe file, mostly in code that came from upstream dev. The loop continues at one review per hour: #16 next, then #1.

CodeRabbit on #16 at 75c5c52: one outside-diff finding. A local xx| prefix wasn't subtracted from the companion page cap. It is fixed in 5231627 (on #1 as 4c8e906) and answered in #16's conversation. Local review round 9 found nothing.

Heads now: #1 4c8e906, #16 5231627, both pushed. The hourly CodeRabbit loop continues with #1 next.

Upstream moved to 5b77947 (2026-10-02): it adds the Seeed nRF54LM20A board and bumps Zephyr to ac03a4a9085.

Merges: the fork's dev was fast-forwarded and merged into Add hw hardware-report CLI command (iteration branch) #1 (35ea3e2) and Add the hw hardware-report CLI command (rebased, hw CLI only) #16 (01a03b1), both pushed. Local review round 11 found nothing new, and the rak4631 overlay merge is correct.
RTC probe adopts a chip whose status-register read failed, then writes time into it #23 is fixed on both branches (581fe6c / 0212c7a) at the maintainer's direction: a failed status read no longer adopts a chip. CodeRabbit's two findings on Add hw hardware-report CLI command (iteration branch) #1 are answered (one fixed, one declined with driver evidence).
ESP32-S3 builds are layout-dependent on this base (ESP32-S3 builds fail to link on dev 5b77947: l32r literal after use in the Xtensa double-exception vector #24, Zephyr f43453290df). Plain dev fails to link heltec_wifi_lora32_v4_r8, _v43 and xiao_esp32s3. Add the hw hardware-report CLI command (rebased, hw CLI only) #16's V4-R8 build fails, while Add hw hardware-report CLI command (iteration branch) #1's links by luck.
Owed: one CodeRabbit pass on each new head, at one review per hour.

Round status, 2026-10-02 17:30 UTC: #17 56fba85, #16 01a03b1 and #1 35ea3e2 each have a clean CodeRabbit pass on the current head ("No actionable comments"), 0 unresolved threads, quick green and merge state CLEAN. The fork's dev equals upstream/dev 5b77947. What remains is Copilot once its quota returns (about 2026-10-06), and the maintainer's go-ahead for any upstream PR. Partial or unstated coverage is acceptable on this fork.

Round status, 2026-10-02 (Copilot is back): #1 work/hw-cli at 515dd95 and #16 hw-cli-only at 47c0649. Copilot's round on each head reports no findings, both have 0 unresolved threads, quick is green, and both merge cleanly. Every Copilot finding from the earlier rounds was fixed or declined with evidence in the PR conversation. Both branches have the fork's dev merged in, which equals upstream/dev f9d01cf. #16's seven files are still byte-identical to the same files on #1.

Fixed this round, on both PRs:

hw dispatches only on the whole word, so hwboard no longer runs hw board.
hw i2c scan stops a bus that times out or is busy, after printing what it had found.
The report is compiled only where CommonCLI is (observers excluded).
A disabled GNSS node is reported as disabled.
A line longer than the 160-byte render buffer is marked with ....
Room is left for a local xx| prefix on every role.
The docs list a bus that was not ready as an unprobed RTC case.
A start index past the last line now answers no line N, the report has M. Before, the reply was empty, and the companion app showed "Unknown command".
hw i2c scan names every declaration at a shared address. The common sensor list puts bosch,bme280 and bosch,bmp388 both at 0x77, and a fitted BMP388 read as BME280.

Declined with evidence: requiring a supported GNSS compatible in HW_GNSS_OKAY. The available: line already reports the GPS manager's verdict, and a second copy of its compatible list would drift.

Issues filed from this work:

RTC probe adopts a chip whose status-register read failed, then writes time into it #23: RTC probe adopted a chip whose status read failed. Fixed on both branches; open until it reaches dev through upstream.
ESP32-S3 builds fail to link on dev 5b77947: l32r literal after use in the Xtensa double-exception vector #24: ESP32-S3 links break on dev (an l32r literal in the double-exception vector). This is upstream's, kept as a fork issue that may need to go to Zephyr or ZephCore.
Local CLI reply behind an xx| prefix gets the full CLI_REPLY_SIZE on repeater/room server #26: a local xx| reply on a repeater or room server gets the full CLI_REPLY_SIZE. This is the pre-existing class in dev that the hw fix exposed.

Next steps, in order:

Re-run hw on hardware (deferred by the maintainer for now). It needs the RAK4631 and the Heltec V4-R8 attached to WSL. Air rules apply to any flash: SoCal parameters, 1 dBm, a short random name, minimal traffic. Done means each subcommand answers as docs/Repeater_CLI_commands.md says, locally, over remote admin and through the companion app.
Optional: CodeRabbit on the current heads. Its last clean passes were on 35ea3e2 and 01a03b1. The OSS plan allows one review per hour per repository, and partial coverage is acceptable on this fork.
Upstream only with the maintainer's explicit go-ahead naming each target. Add the hw hardware-report CLI command (rebased, hw CLI only) #16 is the hw CLI. Companion CLI replies over 175 bytes are truncated silently #3 and RAK4631 never uses a fitted RAK12002 RTC #4 (on Add hw hardware-report CLI command (iteration branch) #1 only) and RTC probe adopts a chip whose status-register read failed, then writes time into it #23 each want their own clean branch from the synced mirror. Add hw hardware-report CLI command (iteration branch) #1 stays open as the iteration log, never merged.

What not to repeat: a review summary headline with Findings: None was twice a real bug (the empty page, the shared address). Read the headline against the code before treating the round as clean.

Round status, 2026-10-03 PDT. The hw CLI is now separated from the upstream dev defects.

The hw CLI. #1 (work/hw-cli, ad70d15) and #16 (hw-cli-only, cc54670) now hold byte-identical trees containing the hw CLI alone.

The three defect fixes Add hw hardware-report CLI command (iteration branch) #1 carried were moved out by append-only commits: fork issues Companion CLI replies over 175 bytes are truncated silently #3 and RAK4631 never uses a fitted RAK12002 RTC #4 on Add hw hardware-report CLI command (iteration branch) #1, and RTC probe adopts a chip whose status-register read failed, then writes time into it #23 on both.
Several local passes and a Copilot round followed. Their findings were documentation-vs-code mismatches, all fixed and answered.
Copilot coverage is unstated, which is acceptable on this fork.
CodeRabbit is queued for Add the hw hardware-report CLI command (rebased, hw CLI only) #16 and Add hw hardware-report CLI command (iteration branch) #1.

Defects found in upstream dev (f9d01cf), one iteration PR each, all into the fork's dev:

Fork issue	PR	Branch	Review state
#3, companion CLI reply cut silently	#27	`work/companion-cli-reply-truncation`	Copilot and CodeRabbit clean; verified on a RAK4631
#26, local `xx\|` prefix overruns the server reply buffer	#28	`work/cli-local-prefix-reply-room`	Copilot and CodeRabbit clean
#13, orphaned CPU-clock comments (widened to 5 boards)	#29	`work/esp32-orphan-cpu-clock-comments`	Copilot and CodeRabbit clean
#23, RTC probe adopts a chip whose status read failed	#30	`work/rtc-probe-failed-status-read`	Copilot and CodeRabbit clean
(no issue, by choice) remote CLI echoes overrun the remote buffer	#31	`work/cli-remote-reply-echo-cap`	Copilot and CodeRabbit clean
(no issue, by choice) companion over-the-air CLI with timestamp 0 treated as local	#32	`work/companion-ota-cli-never-local`	Copilot and CodeRabbit clean
#4 + #33, RAK12002 RTC declared and its backup switchover kept	#35	`work/rak4631-rak12002-rtc`	Copilot round on the head with no new findings; four documentation points listed open in the body; CodeRabbit queued; power cut survived on a RAK4631 + RAK12002

Closed or resolved this round:

ESP32-S3 builds fail to link on dev 5b77947: l32r literal after use in the Xtensa double-exception vector #24 closed: an artifact of symlinked worktrees, not a dev defect.
tcp_work stack overflow under TCP load: Ethernet fixed in #17, WiFi companion unmeasured #25 closed: the ESP32 WiFi driver already gives tcp_work 2048 bytes.
First CLI command after opening the USB serial port returns "Unknown command" #6 does not reproduce on dev; commented, left open for the maintainer.
TX power accepts -9..0 dBm everywhere, though 1 dBm is the floor #19 left open at the maintainer's choice.
RTC probe assumes 24-hour mode: an RV3028 left in 12-hour mode is misread or never adopted #34 filed: the RTC probe assumes 24-hour mode.
RV3028: backup switchover and trickle charging written to RAM only, so they revert at the next midnight; 0xB4 also overwrites a calibration bit meshcore-dev-MeshCore#10 filed: the MeshCore RV3028 RAM-only write.

Next:

On the maintainer's go-ahead for each one: cut a clean squashed branch from the synced dev and open the upstream PR from it, following the "Proposed upstream PR" text in each fork PR body.
rak4631: use a fitted RAK12002 RTC, and keep its backup switchover enabled (iteration branch) #35's four open documentation points go in the next round.

Hardware left as:

The RAK4631 runs rak4631: use a fitted RAK12002 RTC, and keep its backup switchover enabled (iteration branch) #35's debug repeater build: SoCal, 1 dBm, repeat and adverts off, handle lpxz.
The M7 is back on Elecrow ThinkNode M7: board, CH390 ethernet driver, wired TCP companion (iteration branch) #17 a7e2698.
The V4-R8 is untouched.

Continued in #36, round 2 on track hw-cli.

Navigation Menu

Repository navigation

Session Handoff [hw-cli]: hw hardware-report CLI (#1, #16) and fork issues #3/#4 #22

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Session Handoff [hw-cli]: hw hardware-report CLI (#1, #16) and fork issues #3/#4 #22

Description

Activity

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 2, 2026

ptr727 commented on Oct 4, 2026

ptr727 commented on Oct 4, 2026

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions