fix: Heltec V4 board config for Detector and Gateway (VBAT_Read + 16MB flash)

[popshark1]

Problem

The sdkconfig.detector and sdkconfig.gateway files ship with CONFIG_BATEAR_BOARD_HELTEC_V3=y and CONFIG_ESPTOOLPY_FLASHSIZE_8MB=y, but the Heltec WiFi LoRa 32 V4 boards have different hardware that requires updated settings.

Root Cause

On V4 hardware with V3 config:

• GPIO37 (VBAT_CTRL) is misconfigured at boot, pulling the LDO enable line LOW — this prevents boot from battery
• battery_v always reads 0 mV because the ADC control pin is not initialized correctly for V4
• 8MB flash size leaves the upper 8MB of the 16MB chip inaccessible, shrinking available OTA partition space

Changes

sdkconfig.detector

• CONFIG_BATEAR_BOARD_HELTEC_V3=y → CONFIG_BATEAR_BOARD_HELTEC_V4=y
• CONFIG_ESPTOOLPY_FLASHSIZE_8MB=y → CONFIG_ESPTOOLPY_FLASHSIZE_16MB=y

sdkconfig.gateway

• CONFIG_BATEAR_BOARD_HELTEC_V3=y → CONFIG_BATEAR_BOARD_HELTEC_V4=y
• CONFIG_ESPTOOLPY_FLASHSIZE_8MB=y → CONFIG_ESPTOOLPY_FLASHSIZE_16MB=y

Verification

Tested on Heltec WiFi LoRa 32 V4 hardware:

• ✅ Device boots from battery
• ✅ vbat reads correctly (~4090 mV on full charge)
• ✅ LoRa TX packets sent successfully with correct battery voltage in telemetry
• ✅ Serial log: battery monitor ready (adc=GPIO1 ctrl=GPIO37 ratio=4.90)

[TN666]

Thanks for tackling V4 support and adding the build matrix — the 16MB flash change is correct (V4 is 16MB external + 2MB PSRAM, per the Heltec V4 hardware update log). A few things need addressing before this is mergeable, and the stated root cause is inaccurate.

Must fix (blocking)

1. docs/flasher/index.html hardcodes your fork. const REPO = "popshark1/batear", the footer GitHub/Issues/Datasets links, and the datasets link were all changed from batear-io/... to popshark1/.... This is the upstream page — revert all of these.

2. The Gateway and Wired Detector role-info panels are broken. Both lost their <p> description paragraph and gained a stray <span>Heltec WiFi LoRa 32 V3 / V4</span> floating outside .specs. Also: Wired Detector runs on the LILYGO T-ETH-Lite S3, not Heltec — that label is wrong.

3. Duplicate sdkconfig variants. You added both sdkconfig.detector_v3/sdkconfig.gateway_v3 (used by CI) and sdkconfig.detector.v3/sdkconfig.gateway.v3 (unreferenced). Drop the .v3 ones.

4. CI size-report regression. The size-vs-firmware-latest delta comparison was removed, leaving only a flat current-size table. Please keep the delta logic and loop it over the 5 roles instead of deleting it. Also stat ... || echo 0 silently masks a missing binary — make it fail the job.

Root cause is inaccurate (important)

The PR description says GPIO37 "pulls the LDO enable line LOW, prevents boot from battery" and that battery_v=0 is fixed by switching the board flag. Verified against Heltec datasheets, this is not right:

• GPIO37 is ADC_Ctrl, not an LDO enable. It does not gate boot.
• V3 and V4 have opposite ADC_Ctrl polarity. V3 schematic: GPIO37 pulled LOW connects the divider to GPIO1. V4.2.0 official datasheet: "the ADC_CTRL(37) pin needs to be pulled HIGH" (confirmed by MeshCore's HeltecV4Board::getBattMilliVolts() which drives it HIGH to read).
• main/battery.c hardcodes the V3 polarity (drive LOW to read, HIGH to disconnect), and main/pin_config.h shares one #if block for V3 and V4. This PR changes neither. So per the datasheets, flipping only the sdkconfig flag should still read 0 mV on V4 — the real fix belongs in battery.c (invert GPIO37 polarity for CONFIG_BATEAR_BOARD_HELTEC_V4), not in the sdkconfig label.

Could you confirm with a scope/measurement which V4 hardware revision you tested, and whether battery.c was modified out-of-band? The reported ~4090 mV reading contradicts the official V4 datasheet polarity.

Heads-up: V4.3.1 GPIO5 reassignment

The detector uses GPIO5 as I2S WS (PIN_I2S_WS). Heltec V4.3.1 (2026-02-25) reassigned GPIO5 as the FEM control pin (KCT8103L). Advertising a generic "V4" detector will break on V4.3.1 boards. Please scope the claim to V4.2 (GC1109) or reassign I2S WS.

Minor

• fwUrl() indentation is off and the closing brace is on the return line; prefer role === 'wired_detector' ? '' : '_' + boardVer so future LoRa roles don't get missed.
• Keep the ENDOFBODY/ENDOFTAG heredoc markers — EOF is more likely to collide with report content.
• 22 commits with repeated "fix CI / fix flash size / fix env syntax" churn; please squash into ~3 logical commits.

[TN666]

Pushed two commits onto this branch addressing the blocking review items (battery polarity fix you added is kept as-is). Summary:

0c61064 — flasher + sdkconfig

• Reverted REPO const + footer GitHub/Issues/Datasets links from popshark1/* back to batear-io/* (this is the upstream page).
• Restored the <p> descriptions on the Gateway and Wired Detector role-info panels and removed the stray floating <span> that was breaking the markup. Wired Detector keeps its correct LILYGO T-ETH-Lite S3 label.
• fwUrl(): fixed indentation and switched to a wired_detector-exclusion (role === 'wired_detector' ? '' : '_'+boardVer) so future LoRa roles also get a board suffix.
• Dropped the unreferenced sdkconfig.detector.v3 / sdkconfig.gateway.v3 duplicates (CI uses the _v3 underscore variants).

c969178 — CI + battery cppcheck

• Restored the firmware-size delta vs firmware-latest, now looped over all 5 roles. Added set -euo pipefail + bare stat so a missing build artifact fails the job instead of silently reporting 0 B.
• Restored the ENDOFBODY/ENDOFTAG heredoc markers (EOF risks colliding with report content).
• Restored the cppcheck-suppress knownConditionTrueFalse on cali_init() that the polarity commit dropped — without it cppcheck CI fails again.

Please re-flash and confirm on your V4 board that detector + gateway still boot and vbat reads correctly after these changes (none of them touch battery logic beyond the comment, so it should be unchanged — just want a sanity check on real hardware).

Still needs your call (not changed): the V4.3.1 GPIO5 reassignment. The detector uses GPIO5 as I2S WS, but Heltec V4.3.1 reassigned GPIO5 to the FEM control pin. A generic "V4" detector firmware will break on V4.3.1 boards. Which do you want:

1. Scope the V4 detector to V4.2 (GC1109) in the flasher label, or
2. Reassign I2S WS to a free GPIO for V4.

Once that's decided and the V4 hardware check passes, this should be mergeable. (For history, recommend "Squash and merge" given the commit count.)

[TN666]

Firmware Size Report

Posted manually — fork PRs run with a read-only GITHUB_TOKEN, so the CI job can't auto-comment (see note below). Figures are from the green run 27621184521 (4c6023e); firmware is unchanged from c969178.

Role	Current	Previous	Delta
detector_v3	490528 B	—	new
detector_v4	490528 B	—	new
gateway_v3	1089696 B	—	new
gateway_v4	1089696 B	—	new
wired_detector	905104 B	906096 B	-992 B

Compared against firmware-latest. The _v3/_v4 artifact names are new, so detector/gateway show "new".

idf.py size details

detector_v3: Total image size: 424875 bytes (.bin may be padded larger)
detector_v4: Total image size: 424875 bytes (.bin may be padded larger)
gateway_v3: Total image size: 1024047 bytes (.bin may be padded larger)
gateway_v4: Total image size: 1024047 bytes (.bin may be padded larger)
wired_detector: Total image size: 839450 bytes (.bin may be padded larger)

---

Note on automated commenting: the build job runs in the fork-PR context, where GitHub forces GITHUB_TOKEN to read-only regardless of declared permissions, so actions-comment-pull-request can't post (this is what was reddening the check). The proper fix for auto-commenting on fork PRs is a separate workflow_run-triggered workflow that runs in the base-repo context — but workflow_run only uses the workflow file from main, so it has to be merged to main before it takes effect on fork PRs. Happy to add that as a follow-up.