monkey-w1n5t0n 68d4cc4017 build(firmware): migrate to PlatformIO and vendor memllib (plan §5)

One cut, no dual path. Closes ALIGNMENT defect 3 ("Arduino-CLI build
machinery is actively hostile") and vision bullet 4.

platformio.ini carries 16 [env:], one per variant, each passing
-DMEMLNAUT_MODE_TYPE; selftest passes -DNISPS_SELFTEST=1 instead. The env list
IS the registry now — the .ino comment-registry and the NISPS_ST_* token-paste
table are deleted rather than migrated. L12 noted that table was already
silently missing the currently-shipped SLPWorkshop variant, which is the whole
argument against having a second list.

Also deleted: the Python/sed machinery that rewrote the COMMITTED .ino on every
build, the sketch symlink forest, the global TFT_eSPI User_Setup.h mutation
(now -D flags — TFT_eSPI's own documented PlatformIO recipe), the UF2
boot-mount detection stack (upload_protocol=picotool talks to the bootloader
directly), and build-firmware-arch.sh entirely. Scripts 683 -> 435 lines.

memllib is vendored at lib/memllib/ from upstream e291192; no submodules
remain. VENDORED.md records provenance and the re-sync procedure.

S9: a firmware-build CI job compiles three representative envs against a cached
toolchain and reports per-variant flash/RAM. Firmware is in an automated gate
for the FIRST time. The old ci.yml comment justified excluding it as "low
verification value" — an assessment that did not survive contact, since the
SelfTest variant sat broken for an unknown period calling a DisplayDriver
method that did not exist at the pinned memllib commit, and nothing noticed
because nothing built it.

Verified: all 16 envs build from an empty cache, each within ~520 bytes of the
arduino-cli binary it replaces, flash and RAM. Measured as .text+.rodata /
.data+.bss+vector+uninitialized — NOT PlatformIO's console line, which
double-counts .data on this board. This does not prove the hardware boots; no
flash+smoke test was possible and that stays an operator chokepoint.

  slpworkshop 248232/145028   pafsynth 256880/149716   selftest 216228/17960
  (all 16 in the CI log format; none exceeds 2% of a 16 MB flash)

Two traps recorded so nobody rediscovers them: vendoring memllib's subdirs
without a src/ wrapper makes PlatformIO's library builder silently compile
NOTHING while still linking; and project build_flags land BEFORE the
framework's own -std=gnu++17 -Os, so build_unflags is required.

CORRECTION carried in this commit: the firmware sizes in c19d846's message and
the first version of the memllib recon doc were wrong — SLPWorkshop 145348,
PAFSynth 145300, SelfTest 141840. They came from building variants in sequence
through a SHARED incremental arduino-cli build directory, which reused stale
objects and under-reported by ~75 KB. Clean-cache rebuilds of the identical
commit give 216736/18492 for SelfTest. The real cost of the memllib upstream
bump is +216 bytes flash, not +316. Never measure firmware size through a
reused build dir.

HISTORY NOTE: this commit and the docs commit before it were rebuilt (force-push,
2026-07-21) so that each contains only what its message describes. The first
versions had the firmware deletions stranded in the docs commit by a shared-index
race between concurrent agents; content is byte-identical to the originals.

Gates: run-all-tests.sh ALL GREEN (nisps/ untouched by this change beyond
include paths); 16/16 pio envs build.

2026-07-21 20:17:58 +02:00

7 KiB

Raw Blame History

kind	date
finding	2026-07-21

memllib usage inventory — the prerequisite for the §7.5 ownership decision

Prerequisite named by ../plans/simplification-plan.md §5: "inventory which memllib surface (AudioDriver, MEMLNaut board, display/menu, MIDI) is load-bearing" before choosing between a fork-pin and vendoring. The operator's §7.5 decision is vendor the parts we use ("everything should be self contained in this repo"). This document is the evidence that decision needs, plus one consequence that was not visible when it was made.

Method

Not grep. The firmware was built for real (arduino-cli, rp2040:rp2040:solderparty_rp2350_stamp_xl, SLPWorkshop variant) and the linker map was read: MEMLNaut-NISPS.ino.map distinguishes translation units that merely compiled from those contributing sections to the final image. This is the same method that would have caught the daisysp finding (S8) mechanically instead of by reading.

Result: the compile surface is 24 TUs, and all 24 are linked

subdir	.cpp linked / present	headers	size
`audio/`	4 / 4	12	132K
`hardware/`	7 / 7	21	176K
`interface/`	8 / 8	8	108K
`synth/`	3 / 3	10	1.3M
`utils/`	2 / 2	9	80K
total	24 / 24	60	~1.8M

Every translation unit reachable through the sketch-tree symlinks contributes kept sections to the image — no discarded-only TU exists. memllib is therefore not a daisysp situation (41 .cpp compiled, zero consumers). It is genuinely load-bearing.

The library has 30 .cpp in total; the 6 not counted above live outside the five symlinked subdirs (e.g. examples/) and are never compiled by the firmware build. Those are the only part safe to drop outright.

Firmware's direct include surface is just six entry points — audio/AudioDriver.hpp, hardware/memlnaut/MEMLNaut.hpp, hardware/memlnaut/Pins.hpp, hardware/memlnaut/display/{View,SingleSelectView}.hpp, interface/MIDIInOut.hpp, PicoDefs.hpp, utils/perf.hpp — but the transitive closure is the whole of those five subdirs.

Consequence for vendoring: "the load-bearing subset" is ~1.8 MB / 84 files, i.e. essentially all of memllib minus examples/. There is no small subset to lift. synth/ alone is 1.3 MB (mostly maximilian.cpp) and all three of its TUs link.

The fork divergence — and the correction to this document's first version

src/memllib is not a third-party dependency. It is the lab's shared library (MusicallyEmbodiedML/memllib), and our fork was 3 commits ahead / 31 behind.

The first version of this document (commit 7a30da9) drew the wrong conclusion from that, and recommended a rebase-then-vendor on the strength of it. Two facts checked afterwards overturn it:

All three of our commits touch only examples/ — bf9691c (swap memlp includes for nisps/core, inline RL utilities), 32cc831 (newlib RNG fix), b37fc53 (seed-helper rename). examples/ is not in the sketch symlink forest (firmware/MEMLNaut-NISPS/src/ symlinks exactly audio hardware interface synth utils PicoDefs.hpp), so it is never compiled. Those commits existed to let the RL code build against nisps/core while it was being ported, and that port is finished: nisps/ml/{jolt,ou_noise,feedback,geo_push}.hpp cite the upstream sources directly. So there is nothing of ours to carry forward, and no rebase to perform — vendoring the five linked subdirs drops examples/ and our three commits with it.
Two of the "31 commits of work we are missing" were already absorbed. nisps/ml/jolt.hpp:3 cites 9fcd459 "jolts" and nisps/ml/ou_noise.hpp:3 cites d0d8a72 "noise" as their sources. The alarm in the first version — "the 31 commits contain work that sounds directly relevant" — was therefore overstated for exactly the two commits it named.

So the real question was never "rebase or not" but "which snapshot do we vendor: the pinned 188496d we build today, or current upstream?"

The staleness was already costing us

Upstream main has DisplayDriver::NavigateToView (display/DisplayDriver.hpp:54); the pinned commit does not. The SelfTest firmware variant called it and had been failing to compile — fixed in b953681 by routing around the missing method. That variant was not written against broken code; it was written against a newer memllib than the pin. The gap is not theoretical.

Upstream also carries e291192 "l r input swap", a hardware bug fix: the physical L/R input sockets are wired to the opposite codec ADC channels, so AudioDriver.cpp now swaps them at the lowest level. Every mode on the pinned commit sees its stereo input backwards.

Verified: current upstream builds, and costs almost nothing

Submodule moved to e291192 (upstream main).

CORRECTION (2026-07-21, same day). The first version of this table, and the numbers in commit c19d846's message, were WRONG — SLPWorkshop 145348, PAFSynth 145300, SelfTest 141840, with a "+316 uniform delta". They were produced by building three variants in a row through a shared incremental arduino-cli build directory (/tmp/memlnaut-firmware-build), which reused stale objects and under-reported by roughly 75 KB. The error surfaced when the PlatformIO migration could not reproduce them; a clean-cache rebuild of the identical commit in a throwaway worktree confirmed it. Never measure firmware size through a reused build directory.

Measured properly (clean build dir per measurement, SelfTest variant):

memllib pin	flash	RAM
`b37fc53` (old pin, pre-bump)	216520	18480
`e291192` (upstream main)	216736	18492
delta	+216	+12

The conclusion is unchanged and if anything stronger: the 31-commit upstream bump costs ~216 bytes of a 16 MB flash. The bulky new upstream code (GrainDelayI16, ReverbI16, ModFXI16, CCSelectView, NameInputView, RLView, VUMeterView, PSRAMManager) is header-only and unreferenced, so the linker drops all of it; the delta is the AudioDriver/DisplayDriver changes.

Absolute sizes for every variant, post-bump, are in the Phase 4 migration commit — all 16 build, and none exceeds 2% of flash.

Exactly one compile error had to be fixed: MEMLNaut-NISPS.ino:169 used kSampleRate in a constexpr, and upstream 1997699 "mode sample rate" made it a runtime extern size_t so a mode can choose its own rate. constexpr → const; it is a once-per-second diagnostic print.

Where this leaves the vendoring

Vendor from current upstream, not from the old pin. Take the five linked subdirs + PicoDefs.hpp; drop examples/ (~1.8 MB → the vendored surface). Record the exact upstream commit so a future re-sync is a documented diff rather than an archaeology exercise.

The submodule now points at upstream rather than the fork: with the pin moved to an upstream commit, the fork has nothing the firmware compiles. The fork's feat/nisps-core-swap branch remains pushed, so nothing is destroyed.

This unblocks the PlatformIO migration (plan §5).

7 KiB Raw Blame History