luke/esp32-poe-scrypted-viewport/main/app_main.c, branch v1.0.0

streaming pivot: raw TCP data plane, drop per-frame HTTP entirely

2026-06-20T00:59:14+00:00

Replaces the per-frame HTTP POST loop with a single long-lived TCP connection on port 81. The HTTP control plane (/state, /config, /frame for snapshot) stays unchanged. Wire protocol (big-endian, repeating until connection close): [4 bytes jpeg_len][4 bytes seq][jpeg_len bytes JPEG] Why --- The HTTP path hit a measured floor of ~37ms p50 with intermittent ~230ms p95 spikes that survived every Nagle/keep-alive fix attempt. Each frame paid: TCP setup (or pool churn), HTTP parsing, body recv, decode, paint, response write, response ACK. Streaming removes everything except recv + decode + paint. Firmware (main/stream_server.[ch], new) --------------------------------------- - TCP listen on configurable port (81) in its own FreeRTOS task, one client at a time (matches the one-stream-per-device model). - Per accepted socket: TCP_NODELAY on, then loop reading 8-byte header → jpeg body → through the existing jpeg_decoder + display paths. Same stale-seq guard as the HTTP /frame handler (reset per connection so each session starts at seq 1). - Frames received while asleep are still drained (to stay framed) but not painted. The HTTP control-plane POST /state {wake} resumes painting on the next frame. - Every 30 painted frames a structured serial log shows per-stage timing + sustained MB/s — replaces the cross-side Server-Timing header (no HTTP response to attach it to anymore). Scrypted side ------------- - net.createConnection({ host, port: 81, noDelay: true }) opened once per stream session. On disconnect/error/close we auto- reconnect after 500ms. - ffmpeg stdout demux writes [header][body] directly to the socket. sock.write() returning false sets a backpressured flag that drops incoming ffmpeg frames until 'drain' fires — natural TCP backpressure handles "firmware can't keep up" without us modeling it manually. - Stripped the entire fetch-based timing infrastructure (pushStreamFrame, fetchSamples, parseServerTiming, depth histogram, Server-Timing parser). Replaced with one stream-shaped log every 10s: fps + MB/s + socket.write p50/p95/max + drop count + backpressured flag. - pushSnapshot (one-shot first-paint) still uses the HTTP /frame endpoint — small and infrequent, not worth reworking. - postJSON still uses node:http for /state and /config. State machine status flags -------------------------- Extended boot-time flags array from 6 → 7 slots so the stream server's bring-up shows up alongside ETH/MDNS/HTTP. Layout is now E M H S D J T (was E M H D J T).

unified end-to-end per-frame instrumentation + version stamps

2026-06-20T00:07:55+00:00

Cross-side timing was opaque: script saw "req=70ms" but couldn't split TCP/dispatch overhead from firmware decode time, and firmware serial logs and Scrypted console logs couldn't be correlated. Firmware /frame handler now emits Server-Timing on every response with per-stage breakdown: Server-Timing: recv;dur=X, dec;dur=Y, paint;dur=Z, post;dur=W, handle;dur=total Script parses it, joins by X-Frame-Seq implicitly (one POST per seq), derives net_up = req − fw_total, and logs a multi-line p50/p95 breakdown every 10 fetches: fetch "kitchen" #10 (jpeg=137KB) wall p50=65ms p95=140ms emit→post p50=0ms p95=2ms (queue wait) req p50=62ms p95=135ms (fetch → Response headers) net_up p50=43ms p95=110ms (TCP + body wire + dispatch) fw_recv p50=12ms p95=18ms (body off the wire) fw_dec p50=6ms p95=8ms (hardware JPEG) fw_paint p50=0.1ms p95=0.2ms (backbuffer flip) fw_post p50=0.4ms p95=0.6ms body-read p50=1ms (drain — empty body) inflight d0=8 d1=2 d2=0 stale-drops=0 Plus version stamps on both sides for the "is the user on the right code" question: - CMakeLists: PROJECT_VER = git short hash + -dirty marker if dirty. esp_app_get_description()->version exposes it at runtime. Boot log: "Scrypted Viewport boot (v0.1.0 build=4cf36e2-dirty)". - TS: SCRIPT_VERSION const at the top, bumped per commit, logged at script-eval: "Scrypted Viewport up (script=4cf36e2). Callback ..."

Default name = MAC; brief wake on boot; drop dead touch defences

2026-06-15T01:11:21+00:00

- Seed viewport_name with the full base MAC, colons stripped (e8:f6:0a:e0:90:94 → "e8f60ae09094"). Stable across reboots, globally unique, 12 alphanumeric chars — well inside the mDNS hostname limit even with the "viewport-" prefix. - Add a mac_str field to viewport_state and expose it as the new "mac" key in GET /state and as a "mac" line on the info screen. - "Configured" no longer requires a viewport_name (it always has the MAC default); the scrypted URL alone gates outbound POSTs and the loading-vs-info screen choice. - Strip viewport_name[0] fallbacks in http_api / mdns / local_screens now that the field is never empty. - Replace the wake-on-boot-stays-on behaviour with a ~600 ms flash followed by sleep — long enough for the FT5426 touch IC to come out of its initial unresponsive state, short enough that an idle device doesn't burn the backlight. - Drop the touch defensive cruft added when we were chasing a PoE-power theory: the DEV_MODE=0x00 write at init, the touch_reset_pulse() call at init, and the runtime wedge-self-heal in the polling loop didn't actually fix anything — the wake-on-boot flash did. Also delete the display_touch_reset_pulse() helper since it now has zero callers.

touch: wake on boot — FT5426 only reports while LCD is streaming + lit

2026-06-15T00:58:45+00:00

Symptom: after a few cleanup cycles + power cycles, fresh boots left touch dead — the FT5426 ack'd, polling read TD_STATUS=0x00 cleanly, but taps never registered. Curl-waking the screen made touch work again instantly; tapping the dark backlight-off panel did nothing. Root cause: on this panel the FT5426 firmware only reports contact points while the DSI link is actively streaming with the backlight on. Booting straight into ASLEEP per the spec leaves taps silently unread, and there's no built-in way to wake without touch (no hardware button on this board — see prior commit). The fix is to wake on boot; the idle timer cuts back to sleep after idle_timeout_ms if nothing happens. Also defensively write DEV_MODE=0x00 (Working Mode) at touch_init, and add a runtime self-heal: if the polling loop sees stuck TD_STATUS above the 5-touch ceiling for ~3 s, pulse PC_RST_TP_N via the panel MCU and retry. Combined with display_touch_reset_pulse() the chip recovers from cold-boot wedges without manual intervention.

Info screen: full /config + /state dump, replaces identity screen

2026-06-14T22:37:51+00:00

Previously the 4-line "identity" screen showed only viewport name, mDNS host, IP, and state. Expanded to 15 lines covering the full GET /config + GET /state output (name, host, ip, state, configured, scrypted, orientation, brightness, idle, fw, uptime, frames, errs, free heap, free PSRAM), label/value pairs left-aligned with auto-scaled font. Renames "identity" → "info" throughout — symbol, log messages, README/TESTING references. Also: - Move the info-screen render's big locals (~1.6 KiB: lines[16][80] + scrypt[256] + vp_name[64]) to BSS. The touch task's 3 KiB stack was overflowing on every long-press, leaving whatever frame was previously on the panel — gave the appearance of a "blue screen" after the M5 test pattern. - Drop the ≥5 s touch factory-reset gesture and remove all stale references to it in docs and comments. NVS wipe is now a USB-side `idf.py erase-flash` operation only. - M5 follow-up TODO in jpeg_decoder.c: solid green renders ~black and solid blue renders green; not a byte-order issue (BGR setting turns red into blue). Tracked as a known M5 gap.

M3 done: drop boot test, fold BOOT button into touch long-press

2026-06-14T21:41:03+00:00

The Waveshare ESP32-P4-ETH board exposes BOOT on GPIO 35, but at runtime that pin is owned by the EMAC TXD1 signal. No usable GPIO is wired to a separate user button, so the BOOT-button task could never fire. Move both BOOT-button behaviours onto the touch panel: - ≥1.5s hold → 15s identity overlay (was BOOT short-press) - ≥5s hold → factory reset (was BOOT long-hold) Short tap (<500ms) still toggles wake/sleep. Long-press fires while the finger is still down so the user gets immediate feedback at each threshold. Also strip the R/G/B 6 s boot test sequence — the panel now renders correctly, so it's no longer useful diagnostically. Boot subsystem flags drop the trailing B column: [EMHDJT].

M3 WIP: DSI bring-up + state machine + identity overlay

2026-06-14T20:53:33+00:00

DSI / panel: - LDO_VO3 acquired at 2500 mV (VDD_MIPI_DPHY) — without this the PHY PLL busy-waits forever inside esp_lcd_new_dsi_bus. - PSRAM bumped to 200 MHz (CONFIG_IDF_EXPERIMENTAL_FEATURES) to keep the DPI fed without underruns. - Pi-7"-style 800x480 panel init (REG_POWERON, PORTA/PORTB/PWM/PORTC) + 16-bit RGB565 framebuffer with portrait-mode rotation buffer in PSRAM. Currently shows garbled output — known issue, next commit switches to the TC358762-bridge-aware init sequence. State + UX: - viewport_state simplified to AWAKE/ASLEEP only; "unconfigured" is a flag, not a state. Tap always toggles; content choice (identity vs frame) is driven by the configured flag. - BOOT button arms a 15 s identity overlay via local_screens_overlay; expired callback returns to prior state. - Boot-done indicator: two backlight flashes (no usable LED GPIO on the Waveshare board — GPIO 35 BOOT is shared with EMAC TXD1). - Best-effort subsystem init in app_main; display init deferred to its own task so a panel hang can't block networking. Display test pattern: solid R/G/B for 2 s each on boot to characterize the garble independent of text rendering.

M1+M2 hardware bring-up: 3 fixes + elegant subsystem degradation

2026-06-14T17:11:43+00:00

First flash to real hardware (Waveshare ESP32-P4-ETH on USB power, no LAN cable, no panel attached) exposed three bugs and the need to degrade gracefully when peripherals are missing. 1. net_eth.c: RXD0 / RXD1 GPIO swap. ESP32-P4's EMAC iomux table (components/soc/esp32p4/emac_periph.c) fixes RXD0 to GPIO 29 and RXD1 to GPIO 30. The Waveshare-wiki/ESPHome research had them transposed. Symptom was: E (esp.emac.gpio): invalid RXD0 GPIO number E (esp.emac): esp_eth_mac_new_esp32 failed -> ESP_ERROR_CHECK abort Fix: swap the two #defines. CRS_DV / TXD0 / TXD1 / TX_EN / REF_CLK pinout was already correct. 2. mdns_service.c: mdns_service_add() rejects with INVALID_ARG when the hostname isn't set yet (see mdns_responder.c:771 — first guard in mdns_service_add_for_host). We were setting hostname AFTER service_add inside apply_records(). Restructure mdns_service_start to: init -> hostname -> service_add -> txt. apply_hostname() and apply_txt() helpers reuse the same snapshot under viewport_state's mutex; mdns_service_refresh() re-applies both. 3. app_main.c: every subsystem now best-effort instead of ESP_ERROR_CHECK abort. A single missing peripheral can't take down the rest of the firmware. End-of-boot summary line: boot complete — subsystems [EMHdJ-B] ip=(no link) Uppercase letter = up, lowercase = down. E=Ethernet M=mDNS H=HTTP D=Display J=JPEG T=Touch B=BootButton. Touch shows '-' when display didn't come up (it shares the panel I2C bus). On a stripped board (just ESP32-P4-ETH on USB, no panel, no LAN) the line above prints and the device serves /state, /config, and mDNS over the loopback; plugging Ethernet in later picks up DHCP without a reboot. Side-effects: the DHCP timeout in app_main shortened from 30s to 15s so a no-cable boot finishes quickly. Reaching the "ready" state with no link is fine — the driver keeps the link-up event handler armed and gets the IP whenever a cable appears. Verified on hardware (bare ESP32-P4-ETH, no LAN, no panel): - ESP-IDF v5.4.1, esp32p4, 32 MB PSRAM detected, 16 MB flash config (actual 32 MB silkscreen — keeping 16 MB until access-beyond-16MB support lands in flash driver). - Ethernet driver started — MAC e8:f6:0a:e0:90:94. - mDNS up advertising viewport.local on _scrypted-viewport._tcp:80. - HTTP up listening :80. - JPEG decoder ready. - Display reports "panel MCU @0x45 unreachable" — exactly correct for no-panel state. - Touch skipped as designed. - BOOT button registered on GPIO 0 (still a guess; harmless if wrong). This is enough to flip M1's "code" status to ✅ on hardware once Ethernet is plugged in. M2 (mDNS browse + GET /state) ready to verify.

M8: local screens (IP + Loading) + BOOT button

2026-06-14T04:15:27+00:00

local_screens.{h,c}: - Embedded 8x8 bitmap font sized to a 95-char table; only the glyphs used by today's strings (viewport.local, the IPv4 string, Loading...) are populated. Unsupported chars render blank. - 768 KB PSRAM scratch FB at panel-native dims. - local_screens_show_ip() — two centered lines: "viewport.local" Scale 3x portrait / 4x landscape; centered vertically. - local_screens_show_loading() — centered "Loading..." at scale 4x/5x. - local_screens_restore_for_state() — repaint after BOOT-overlay expiry: UNCONFIGURED -> IP screen; AWAKE/ASLEEP -> black FB (next /frame or sleep handles the rest). - All paths go through display_present_rgb565() so orientation rotation is automatic. button.{h,c}: - Polling task at 30ms, active-low with internal pull-up. - Short press (<5s, released) → backlight on, IP overlay for 15s via esp_timer one-shot, then restore. - Hold ≥ 5s → nvs_config_reset() + esp_restart(). - PIN_BOOT_BUTTON = GPIO 0 is a TODO placeholder. ESP32-P4 strap pin GPIO35 is owned by RMII TXD1 at runtime so Waveshare must expose a separate user button on a free pin; confirm against the schematic. Fail-soft if mis-wired: input reads stuck-high and the task never fires. state_machine on AWAKE transition now calls local_screens_show_loading() right after display_wake() — so every wake (tap or POST /state) flashes a clear "Loading..." until the next /frame paints over it. app_main on boot: - Calls local_screens_init() after display_init(). - UNCONFIGURED → paint IP screen (replaces the M3 test pattern). - ASLEEP → backlight off (unchanged). - After all subsystems: button_init(). Build clean against ESP-IDF 5.4 (binary ~870 KB; the font/glyph data is under 800 bytes). TESTING.md M8 documents the visual checks, the BOOT-pin placeholder caveat, the AWAKE-overlay-overwrite behavior, the open follow-up about post-overlay backlight when ASLEEP, and the font fallback policy.

M7: touch + outbound /state POST to Scrypted

2026-06-14T04:01:23+00:00

state_client.{h,c}: - Worker task drains a depth-1 queue (xQueueOverwrite gives the replace-on-full semantics from the spec — in-flight POST is never cancelled; the next queued entry is overwritten by newer state). - esp_http_client POST to /state with Content-Type application/ json, User-Agent ScryptedViewport/, Connection: close, 1s timeout. Body: {"viewport":"","state":"wake"|"sleep"}. - Any non-2xx or transport error increments state_post_failures and is otherwise ignored. Silently drops if no Scrypted URL is configured. touch.{h,c}: - FT5426 capacitive touch on the shared I2C bus at 0x38. - 30ms polling task; tracks down/up transitions, detects taps as down-then-up within 500ms with a 150ms debounce. - On tap, toggles wake/sleep via state_machine_set_local(), which drives the local transition AND fires state_client_post() at Scrypted. state_machine adds state_machine_set_local(): runs the same transition as state_machine_set() then enqueues an outbound POST. Idle-timer expiry now uses this path so Scrypted sees idle-driven sleeps too (the spec's "tighten the race with /frame 409" path stays in place as the fallback). display.h exposes display_i2c_bus(); touch.c uses it instead of re-initializing the same I2C port. app_main starts state_client right after the state machine and starts touch after display init (touch is skipped if display isn't up since they share the bus). CMakeLists.txt: add esp_http_client to REQUIRES. Build clean against ESP-IDF 5.4 (binary ~860 KB; jumped ~210 KB from M6 because esp_http_client + cJSON path pulls in tcp_transport, mbedtls, http_parser). TESTING.md M7: flask-based test receiver, tap dispatch verification, failure-path check (kill receiver, confirm counter increments), queue-coalescing behavior, no-POST-on-Scrypted-initiated, no-POST- when-unconfigured.