[npp1993]

Requested by Nathaniel

• What was wrong or hard: Two things in the base's machine registry bit Analog BD over USB.
  • SRC SETUP's editor hook (mw_setup_edit6, taken from POLY8's shared hooks) leaves encoder E, slot 10, read-only for every module machine. That is POLY8's hidden slot, but it is Analog BD's HIGH, so HIGH never moved over USB (0.1.5 included).
  • A module machine without a pool opened its track's double-tap window on stock's sample list, which is empty for it. Its tracks are FLEX underneath, and stock picks the window's column from the track's type byte. The base then moved it to the machine list, which flickered; PICKUP opens straight on the list.
• What this changes: Two opt-in machine flags, so POLY8, VECTOR, FM Synth and Analog BD 0.1.5 behave exactly as before.
  • MODWERK_MACHINE_SETUP_ALL (16) lets encoder E edit the machine's slot (mw_row_setup_all).
  • MODWERK_MACHINE_LIST (32) makes the window open on the machine list. A new owned site at 0x400788c4 (12 bytes, mw_chooser_column → mw_window_column) takes stock's column choice and keeps stock's rule for everyone else.
  • Also: events.json, the headers and EVENTS.md are regenerated; machine_ports.py knows the two flag names. The beta base is regenerated: b2810a87 after merging main at d6f1882b (Improvement: the Octatrack beta base carries the sample-editor service (Beat Slicer's base) #992's sample-editor service). It was 68e6a90b on main at fe5955b2, and cc74d638 before the first merge; the five base-bound packages only change requiredBase, and the machine recipes re-pin machine_ports.py.
  • The machines guide gets a note on SETUP two-state switches and the two flags.
  • The Analog BD 0.1.6 release that uses both flags follows as its own PR, stacked on this one.
• How I checked it:
  • npm run pr:ready: Ready, after merging main (dbf75048, dd113bb7, and a9ae0de1 at d6f1882b). The last merge also re-pins the Tape Echo and Euclid standalone recipes to module-build.ts and the card-build assets Beat Slicer 0.1.66-experimental: first release, over USB on the base's sample-editor service #993 changed, without which octatrack:module-packages refuses. The base files were regenerated with their producers (events.py --write, npm run octatrack:base -- --beta --write, npm run octatrack:module-packages), not resolved by hand. Analog BD's controller hash is unchanged from main; only its source-policy hash moves (the machine_ports.py re-pin).
  • test_machine_registry.c covers both flags: a SETUP_ALL row, and the window column for a LIST machine, a pool machine, a stock FLEX track and a stock non-sample type.
  • ot_emu, card build: the double tap shows the machine list; T2's window switched to T1 shows the list; FLEX + RIGHT still opens FLEX's pool.
  • On a MKII, with this base plus Analog BD 0.1.6 (my build 9976a5e8, RAM boot, then flashed with 14 modules), reported by me (npp1993):
    • HIGH moves.
    • The double tap opens the machine list instantly with no flicker.
    • T2 → T1 shows the list.
    • The set survives a power cycle.
  • The run below is on 68e6a90b, this PR's base before Improvement: the Octatrack beta base carries the sample-editor service (Beat Slicer's base) #992 was merged in. Improvement: the Octatrack beta base carries the sample-editor service (Beat Slicer's base) #992 adds the sample-editor service and changes nothing in the machine registry; the run has not been repeated on b2810a87.
  • On a MKII, base 68e6a90b (this PR's merged base at the time, npm run octatrack:base -- --beta --out, identical to the committed package) RAM-booted over USB, with Analog BD 0.1.5 (7a1b08e8, flags off) kept, observed by me (npp1993), 27 Oct 2026:
    • Self-test passed: 1/1 module intact, longest frame 374 µs.
    • Double tap on a stock FLEX and a STATIC track opens their pool.
    • Double tap on a THRU and a PICKUP track opens the machine list.
    • Analog BD 0.1.5's double tap opens its 808/909 engine browser as before; encoder E on its SRC SETUP stays read-only.
    • Then flashed over USB (flash, written and verified into slot 2) and power-cycled: the unit started on 68e6a90b from the flashed slot (confirmed), Analog BD 0.1.5 still kept, self-test passed (1/1 intact).
    • With my other 13 modules reinstalled on top (Output Matrix and Recorder Loop Fix bound to 68e6a90b), all 14 are intact, but the self-test fails twice on CPU over 90 % (93.2 % and 93.1 %, named Air Chorus) with all 9 DSP effects running together. I have no run of this exact set on another base to compare, so I don't attribute it to this PR, which changes no DSP or tick code.
  • The earlier run above used 9976a5e8 (Analog BD 0.1.6-experimental: TYPE and FIXD switches, SETUP reorder, LPF OFF clockwise, 808 variation, leaner DSP #944's base) with Analog BD 0.1.6, not this PR's base.
  • Not checked: FM Synth, POLY8 and VECTOR on the unit (unchanged by design, flags off).
• What the guides lacked or got wrong, and where I fixed it: docs/module-guides/machines.md did not say how to make a SETUP control a two-state switch, or that encoder E is read-only for module machines over USB. Added to the Track machines checklist.

🤖 Generated with Claude Code

[repeat98]

Thanks, this reads well. Both flags are opt-in, nothing in machine-ports.json sets them here, the regenerated base matches its sources, and the host test covers the new paths. Two things before I merge:

1. The unit run was on a different base. The MKII results are from 9976a5e8, which is Analog BD 0.1.6-experimental: TYPE and FIXD switches, SETUP reorder, LPF OFF clockwise, 808 variation, leaner DSP #944's base (it also carries Analog BD 0.1.6's controller policy). This PR publishes cc74d638, which no unit has booted yet. The new site at 0x400788c4 runs on every track's double tap, not only on module machines, so I'd like one short run on the base this PR finally publishes, with Analog BD 0.1.5 (flags off):

   • double tap on a stock FLEX and a STATIC track opens their pool
   • double tap on a THRU or PICKUP track opens the machine list
   • Analog BD 0.1.5's engine browser still opens, and encoder E on its SRC SETUP stays as it was
   • a power cycle keeps the set

   I can run this on my own unit if you can't; just say so. Then put the result in the description.

2. Merge order. Analog BD 0.1.6-experimental: TYPE and FIXD switches, SETUP reorder, LPF OFF clockwise, 808 variation, leaner DSP #944 contains this commit, so this one goes first and Analog BD 0.1.6-experimental: TYPE and FIXD switches, SETUP reorder, LPF OFF clockwise, 808 variation, leaner DSP #944 right after it. Other base pull requests are ahead in the queue (Improvement: Octatrack base events for MIDI tracks' KEY (ev_midi_note_key, ev_arp_setup_knob, ev_arp_setup_text) #857 is ready now), so expect one merge of main and a rebuild of the base before this goes in. The unit run above should then be on that rebuilt base, not on cc74d638; I'll tell you its hash.

Everything else looks fine: the changelog entry is there, no event or table index is added (only one owned site in ot_machines, 48 to 49, and flag bits 16 and 32), and the other machines' recipes only re-pin machine_ports.py.

[repeat98]

#857 is in now, so this is the merge of main and the base rebuild I mentioned. Please merge main and regenerate the base files with their producers rather than resolving them by hand (base-binding.json, the elemod, the variants and the module packages conflict now). Then the short run from my earlier comment on that rebuilt base, with Analog BD 0.1.5. If you'd rather I do it on my unit, just say.

[repeat98]

Thanks for the run on 68e6a90b and for regenerating everything with the producers. One small thing: the new checklist line in docs/module-guides/machines.md points at EVENTS.md, which main renamed to HOOKS.md (sdk/machines/octatrack/HOOKS.md). The PR description has the same old name. Could you fix both? #944 picks the fix up when you merge it in there.

[npp1993]

Thanks! This landed in #1017 (its bundle contains this head), and main's docs/module-guides/machines.md already points at HOOKS.md, so I'm closing this one.

---

Generated by Claude Code