Dev: v3.10.5

AGENTS.md

@@ -33,8 +33,8 @@ It is standalone (not Eurorack), powered from a 5 V power bank through a screw t
 - KiCad exports: Export Gerbers from `hardware/pcb/kicad/patternflow.kicad_pcb`. Export STLs from `hardware/case/source/patternflow_case.blend`.
 
 ## Versioning
-- Project: v3.10.4 (current, released 2026-09-19), using unified semantic versioning across firmware, hardware, web, and docs — the firmware reports it as `PF_IMPROV_FW_VERSION` in `net_config.h`, and `CHANGELOG.md` gets a section per release. Current hardware is the v3.9 board — the v3.0 board with the USB-C footprint removed, same enclosure, pin map and guide; every v3.x release has been firmware/web on it.
-- Editions carry their own version line, independent of the project version (`PF_VARIANT_VERSION` in `firmware/bundles/<name>/overrides.h`): Audio v0.6.4, Performance v0.4.0 at the time of writing. The shelf (`web/src/app/editions/editions-data.ts`) and the flasher manifest (`web/public/flash/manifest.json`) name the images that are live; only those images are kept in `web/public/flash/bin/` — older ones live on their release tags. The one exception is a try-out image: a frozen copy of a composition that is not on the shelf, named by the features page (`tryOut` in `web/src/app/features/features-data.ts`) so people can install it to try, never bumped with the core, and retired only by a newer try-out of the same name. `clock` and `midi` are bundles the tree keeps building (CI compiles every composition) but does not put on the shelf: the clock is a feature, and the MIDI bundle is the composition that proves the USB-OTG build; each has a try-out image (`clock-v0.1.5`, `midi-v0.1.0`). Neither is in the `EDITIONS` tuples of `release.py`/`check_versions.py`; an edition joins those, the cards and the line above only when it goes on the shelf, and its maintainer cuts it with `release.py edition <name> vA.B.C` (docs/RELEASING.md).
+- Project: v3.10.5 (current, released 2026-10-02), using unified semantic versioning across firmware, hardware, web, and docs — the firmware reports it as `PF_IMPROV_FW_VERSION` in `net_config.h`, and `CHANGELOG.md` gets a section per release. Current hardware is the v3.9 board — the v3.0 board with the USB-C footprint removed, same enclosure, pin map and guide; every v3.x release has been firmware/web on it.
+- Editions carry their own version line, independent of the project version (`PF_VARIANT_VERSION` in `firmware/bundles/<name>/overrides.h`): Audio v0.6.5, Performance v0.4.0 at the time of writing. The shelf (`web/src/app/editions/editions-data.ts`) and the flasher manifest (`web/public/flash/manifest.json`) name the images that are live; only those images are kept in `web/public/flash/bin/` — older ones live on their release tags. The one exception is a try-out image: a frozen copy of a composition that is not on the shelf, named by the features page (`tryOut` in `web/src/app/features/features-data.ts`) so people can install it to try, never bumped with the core, and retired only by a newer try-out of the same name. `clock` and `midi` are bundles the tree keeps building (CI compiles every composition) but does not put on the shelf: the clock is a feature, and the MIDI bundle is the composition that proves the USB-OTG build; each has a try-out image (`clock-v0.1.5`, `midi-v0.1.0`). Neither is in the `EDITIONS` tuples of `release.py`/`check_versions.py`; an edition joins those, the cards and the line above only when it goes on the shelf, and its maintainer cuts it with `release.py edition <name> vA.B.C` (docs/RELEASING.md).
 - Firmware source lives in `firmware/patternflow/`; use release tags for versioning instead of encoding the release in the folder name. Tags are `vX.Y.Z`.
 - Conventions: filenames lowercase with underscores; commit messages start with the area (`firmware:`, `web:`, `docs:`, `hardware:`) then a short present-tense summary.
 

CHANGELOG.md

@@ -4,10 +4,12 @@ All notable changes to Patternflow will be documented in this file, newest first
 
 ## [Unreleased]
 
+## [3.10.5] - 2026-10-02
+
 ### Firmware
 
 - **A panel that crashed says where.** `resetReason` in `/api/status` said `"panic"` or `"task_wdt"` and nothing else - not where, and not which of forty installed patterns was on. The SDK had been writing a core dump to its own flash partition on every panic all along, and nothing read it. At boot the firmware now reads that dump's summary - the task, the cause, the faulting address, the PC and up to sixteen return addresses, and the hash of the image that crashed - and it keeps a breadcrumb through the reset: the pattern's slug, which of its calls was running (loading, constructors, `setup`, `update`, `draw`, or none of them) and where the module's code was executing from. Status gains a `crash` object with both. With the two together an address inside a module comes out as an offset (`"+0x4e"`) that the `.pfm`'s own symbol table resolves, whether the code ran from PSRAM or from internal RAM. `DELETE /api/crash` clears it, and the same record is printed on serial at boot (`[CRASH] ...`). It reports and does nothing else: no pattern is skipped, no reboot is forced, and what the panel does after a crash is what it did before. The breadcrumb covers the boot that follows a panic or a watchdog and is gone with the power; the dump stays in flash, so a board that panicked on an older firmware shows that dump, marked `fromThisReset: false`, until it is cleared. Checked on a board: a module stored through a null pointer in `draw()` while running from PSRAM, and after the reboot status named the pattern, the phase, `pc` `+0x4e` and the caller at `+0xb6` - the offsets the module's own symbols give for those two functions - with the dump read in 62 ms; a dump left by an earlier firmware appeared as `fromThisReset: false`. It costs 296 bytes of internal RAM. `docs/rest-api.md` ("The crash record") says how to read one and decode it, and `firmware/toolchain/check_crash.py`, a new host check, boots the logic through every reset reason against every state of the breadcrumb and the dump; CI runs it.
-- **Every pattern is 2.45 ms a frame faster.** Copying a finished frame to the panel's bit-planes was the largest fixed cost in the loop: 5.96 ms of every frame, nearly half of Origin's. The kernel was one long loop with more live values than the CPU has registers and a plane count it only learned at run time; it is now three short passes with the 8-bit case written in, 390 instructions a column pair where there were 635, emitting exactly the same words (`check_blit.py` compares 70 million of them). Measured on two boards, alternating images: `presentUs` 5,960 -> 3,496 µs on the default and Audio editions alike, so Origin goes from 12.1 to 9.6 ms a frame and a 30 ms pattern to 27.6. With white balance or gamma tuned away from the defaults it is 3.8 ms. It also takes 1 KB less internal RAM than the loop it replaces. `/api/status` gains `loopStackMin`. Not yet checked by eye on a panel.
+- **Every pattern is 2.45 ms a frame faster.** Copying a finished frame to the panel's bit-planes was the largest fixed cost in the loop: 5.96 ms of every frame, nearly half of Origin's. The kernel was one long loop with more live values than the CPU has registers and a plane count it only learned at run time; it is now three short passes with the 8-bit case written in, 390 instructions a column pair where there were 635, emitting exactly the same words (`check_blit.py` compares 70 million of them). Measured on two boards, alternating images: `presentUs` 5,960 -> 3,496 µs on the default and Audio editions alike, so Origin goes from 12.1 to 9.6 ms a frame and a 30 ms pattern to 27.6. With white balance or gamma tuned away from the defaults it is 3.8 ms. It also takes 1 KB less internal RAM than the loop it replaces. `/api/status` gains `loopStackMin`.
 - **A pattern's code no longer competes with the network for memory.** A loaded module's code used to need one contiguous block of internal executable RAM - the same RAM the console, lwIP and every feature run on - because "the S3 cannot execute from PSRAM". That is true only of the address the heap hands back: on the S3 the instruction and data buses share one MMU table, so the same PSRAM page can be fetched at that address + `0x06000000`. The loader now places code in PSRAM, writes and relocates it through the heap's pointer, and runs it through the alias (`src/core_module_memory.h`, `execAddress()` in `src/core_module_loader.h`). On the Audio edition a module with 23 KB or 32 KB of code was refused and now loads, and internal heap with a module resident is about 27.8 KB whatever the size of its code, where a 10 KB one used to leave 21-23 KB. On a second panel a community pattern with 8 KB of code was being refused one load in four for want of 528 bytes; it now loads every time and leaves 29 KB free instead of 22. Measured on two boards, same boot, alternating: catalogue and community modules within -0.1..+1.2% of their old frame time; modules built to be the worst case (10-32 KB of code, all of it executed every frame) 2.6-5.3% slower; 400 switches between modules of different sizes with no reset and no refusal. The code block is given whole cache lines to itself, the load reads it back through the instruction bus before calling it, and a unit on which that ever fails goes back to internal RAM until reboot (`moduleMemory.codePolicy`). `load.code` in `/api/status` says where the code is. Module files and the ABI are unchanged.
 - **A request that stops half-way can no longer reboot the panel.** A multipart `POST` that ended after its first boundary - to any address, not only an upload - left the web server's form parser reading empty lines for ever with nothing to wait for, and the Core-0 watchdog reset the board five seconds later. The parser now knows a line that never ended from a blank one and gives the form up, tells an upload handler whose file had already arrived that it was aborted, and leaves none of the form's fields behind to answer for the next request (`src/webserver/VENDORED.md`, Fix 4). Reproduced on a panel before, gone after.
 - **Three more requests that could reboot it, found on a PC first.** `curl -X POST` to `/api/patterns` or `/update` with no form in it called the upload handler with no upload to read and panicked the board; a multipart request naming a 9,000-character boundary overflowed the network task's stack; and a raw `PUT /update?size=N` never saw its own `size`. All three are fixed in the vendored server (`VENDORED.md`, Fix 5), and all three came out of `firmware/toolchain/check_parser.py`, a new host check that replays cut-off, stalled and malformed requests - 13,953 of them - through the real parser and fails if it ever spins or holds; CI runs it.

README.md

@@ -175,7 +175,7 @@ Patternflow is built around a standalone ESP32-S3 driving a HUB75 RGB LED matrix
 
 ## Repository & documentation
 
-> **Moving fast.** [v3.10.4 is out](https://github.com/engmung/Patternflow/releases/tag/v3.10.4): the panel films properly on a phone. It now refreshes at exactly 300 Hz — the one rate the shutter speeds a phone actually uses (1/50, 1/60, 1/100, 1/30) all divide — so the rolling bands are gone at those; 1/120 and slow motion still band, and [the firmware README](firmware/README.md#refresh-rate-anti-flicker-for-video) says why and what to set. [v3.10.3](https://github.com/engmung/Patternflow/releases/tag/v3.10.3) made the panel's colours come out right: bright edges had been growing purple and sky-blue fringes and greys read pink; the cause was in the display driver — which bit plane each brightness window actually lights — and part of it was 3.10.1's own dark-end fix. The planes are exactly binary now, and because the fix is in the driver every installed pattern gets it without a rebuild. It also brings a [Knobs page](docs/rest-api.md#get-apiknobs-and-post-apiknobs) on the device (which way each encoder counts, set from the browser) and the [features catalogue](https://patternflow.work/features). Both editions — Audio v0.6.4, Performance v0.2.9 — are re-cut on this core. [v3.10.2](https://github.com/engmung/Patternflow/releases/tag/v3.10.2) made a Format that did not take stop reporting success, and a storage that will not mount say why. [v3.10.1](https://github.com/engmung/Patternflow/releases/tag/v3.10.1) fixed an upload that could reboot the board, a failed settings write that could make the next boot forget a good pattern, a mic-less Audio panel holding the knobs, and MIDI notes that would not let go. [v3.10.0](https://github.com/engmung/Patternflow/releases/tag/v3.10.0) made a heavy pattern come on every time. [v3.9.5](https://github.com/engmung/Patternflow/releases/tag/v3.9.5) taught the web console to survive a slow Wi-Fi link and a heavy pattern, and put thumbnails under the knob. [v3.8.0](https://github.com/engmung/Patternflow/releases/tag/v3.8.0) split the firmware into [editions](docs/EDITIONS.md) — core, Audio (on-board microphone, OSC, network MIDI) and Performance (sequences, MQTT, weather) — installable in one click and switchable without losing your patterns. [v3.2.0](https://github.com/engmung/Patternflow/releases/tag/v3.2.0) made patterns install over Wi-Fi as `.pfm` modules, no reflash. On v2.x hardware? Everything you need stays bundled at [v2.1.0](https://github.com/engmung/Patternflow/releases/tag/v2.1.0). Follow the [changelog](CHANGELOG.md) and the [journal](https://patternflow.work/journal) for what's current.
+> **Moving fast.** [v3.10.5 is out](https://github.com/engmung/Patternflow/releases/tag/v3.10.5): every pattern runs 2.45 ms a frame faster, and a pattern's code no longer takes the memory the network runs on - it executes from PSRAM, so patterns that were refused for size now load and the console has more room while one is running. The panel raises its own Wi-Fi hotspot when it knows no network in reach, with the whole console on it. A panel that crashed says which pattern it was in and where, a pattern that hangs no longer takes the console down with it, and five requests that could reboot the board cannot any more. Audio is re-cut on this core as v0.6.5; Performance stays at v0.4.0 until its maintainer re-cuts it. [v3.10.4](https://github.com/engmung/Patternflow/releases/tag/v3.10.4) made the panel film properly on a phone. It refreshes at exactly 300 Hz — the one rate the shutter speeds a phone actually uses (1/50, 1/60, 1/100, 1/30) all divide — so the rolling bands are gone at those; 1/120 and slow motion still band, and [the firmware README](firmware/README.md#refresh-rate-anti-flicker-for-video) says why and what to set. [v3.10.3](https://github.com/engmung/Patternflow/releases/tag/v3.10.3) made the panel's colours come out right: bright edges had been growing purple and sky-blue fringes and greys read pink; the cause was in the display driver — which bit plane each brightness window actually lights — and part of it was 3.10.1's own dark-end fix. The planes are exactly binary now, and because the fix is in the driver every installed pattern gets it without a rebuild. It also brings a [Knobs page](docs/rest-api.md#get-apiknobs-and-post-apiknobs) on the device (which way each encoder counts, set from the browser) and the [features catalogue](https://patternflow.work/features). Both editions were re-cut on that core. [v3.10.2](https://github.com/engmung/Patternflow/releases/tag/v3.10.2) made a Format that did not take stop reporting success, and a storage that will not mount say why. [v3.10.1](https://github.com/engmung/Patternflow/releases/tag/v3.10.1) fixed an upload that could reboot the board, a failed settings write that could make the next boot forget a good pattern, a mic-less Audio panel holding the knobs, and MIDI notes that would not let go. [v3.10.0](https://github.com/engmung/Patternflow/releases/tag/v3.10.0) made a heavy pattern come on every time. [v3.9.5](https://github.com/engmung/Patternflow/releases/tag/v3.9.5) taught the web console to survive a slow Wi-Fi link and a heavy pattern, and put thumbnails under the knob. [v3.8.0](https://github.com/engmung/Patternflow/releases/tag/v3.8.0) split the firmware into [editions](docs/EDITIONS.md) — core, Audio (on-board microphone, OSC, network MIDI) and Performance (sequences, MQTT, weather) — installable in one click and switchable without losing your patterns. [v3.2.0](https://github.com/engmung/Patternflow/releases/tag/v3.2.0) made patterns install over Wi-Fi as `.pfm` modules, no reflash. On v2.x hardware? Everything you need stays bundled at [v2.1.0](https://github.com/engmung/Patternflow/releases/tag/v2.1.0). Follow the [changelog](CHANGELOG.md) and the [journal](https://patternflow.work/journal) for what's current.
 
 | Folder | Contents |
 | :--- | :--- |

firmware/bundles/audio/overrides.h

@@ -46,7 +46,7 @@
 // v0.4.0 shipped still believing it was v0.3.1 because nothing tied the two
 // together; shelf.sh now refuses an image that does not contain its version.
 #define PF_VARIANT "audio"
-#define PF_VARIANT_VERSION "v0.6.4"
+#define PF_VARIANT_VERSION "v0.6.5"
 
 // ── The on-board microphone drives the knobs ────────────────────────────
 //

firmware/patternflow/net_config.h

@@ -109,7 +109,7 @@
 // Firmware version string reported to the flasher (Improv device-info RPC).
 // Keep in sync with web/public/flash/manifest.json.
 #ifndef PF_IMPROV_FW_VERSION
-#define PF_IMPROV_FW_VERSION "3.10.4"
+#define PF_IMPROV_FW_VERSION "3.10.5"
 #endif
 
 // ── Variant identity (RFC: docs/rfc-core-and-variants.md) ────

web/public/flash/manifest.json

@@ -1,26 +1,26 @@
 {
   "name": "Patternflow",
-  "version": "v3.10.4",
+  "version": "v3.10.5",
   "new_install_prompt_erase": false,
   "new_install_improv_wait_time": 20,
   "builds": [
     {
       "chipFamily": "ESP32-S3",
       "parts": [
         {
-          "path": "bin/core-v3.10.4/patternflow.ino.bootloader.bin",
+          "path": "bin/core-v3.10.5/patternflow.ino.bootloader.bin",
           "offset": 0
         },
         {
-          "path": "bin/core-v3.10.4/patternflow.ino.partitions.bin",
+          "path": "bin/core-v3.10.5/patternflow.ino.partitions.bin",
           "offset": 32768
         },
         {
-          "path": "bin/core-v3.10.4/boot_app0.bin",
+          "path": "bin/core-v3.10.5/boot_app0.bin",
           "offset": 57344
         },
         {
-          "path": "bin/core-v3.10.4/patternflow.ino.bin",
+          "path": "bin/core-v3.10.5/patternflow.ino.bin",
           "offset": 65536
         }
       ]

web/src/app/editions/editions-data.ts

@@ -163,8 +163,8 @@ export const EDITIONS: Edition[] = [
       'The largest block a pattern can claim — 92 KB, against 74 KB elsewhere',
     ],
     hosted: {
-      version: 'v3.10.4',
-      url: 'https://patternflow.work/flash/bin/core-v3.10.4/patternflow.ino.bin',
+      version: 'v3.10.5',
+      url: 'https://patternflow.work/flash/bin/core-v3.10.5/patternflow.ino.bin',
     },
     source: 'https://github.com/engmung/Patternflow',
     note:
@@ -236,8 +236,8 @@ export const EDITIONS: Edition[] = [
     // Served from here, so the panel's own /update page can fetch it. Under
     // /flash/bin, which already sends the CORS header that fetch needs.
     hosted: {
-      version: 'v0.6.4',
-      url: 'https://patternflow.work/flash/bin/audio-v0.6.4/patternflow.ino.bin',
+      version: 'v0.6.5',
+      url: 'https://patternflow.work/flash/bin/audio-v0.6.5/patternflow.ino.bin',
     },
     source: 'https://github.com/engmung/Patternflow/tree/main/firmware/bundles/audio',
     note: