[repeat98]

VECTOR 0.2.6-experimental: VECTOR's SRC SETUP page 2 (SEED, SPAN, OFST, ROT, RPT, DIR) can open on VECTOR's own machine row wherever it sits. This is the VECTOR half of the card-build fix for #1119 (also #842/#843). The POLY8 half is #1142.

Related to #1119

What was wrong

On a card build with POLY8, a double tap on SRC on a VECTOR track shows the stock Flex/Static setup (LOOP … TSTR AUTO, TSNS) instead of VECTOR's page 2. Beside POLY8, VECTOR is machine row 6 or 7, and two things stop its page there:

• POLY8's shared hooks replace VECTOR's and never ask it. That part is fixed in POLY8 0.3.3.
• st_setup_page also wants SRC SETUP's cursor (0x460d5c30) on row 5. That part is fixed here.

Over USB the second part was solved in 0.2.5 (#972), where the adapter rewrites that test out of the package's source.

What changes

• native.c: st_setup_page drops the row-5 test. Every caller already asks on VECTOR's row: hooks.s on row 5 when VECTOR is alone, the base's mw_setup_page over USB, and POLY8 0.3.3's mr_setup_page. control.s is regenerated in the pinned toolchain; the diff is 3 instructions removed.
• vector_runtime.py no longer rewrites that seam; it only pins the new control.s. The package's control unit source is byte-identical to 0.2.5's (checked), so the 0.2.5 USB package stays published (lagsBehind with a reason).
• Pins updated: machine-ports.json, machine-recipes.json.
• Updated for the new version: the quick-test guide, the version pin in module-source.test.mjs, the catalogue, release notes and TESTING.md.
• Native comparison re-recorded: 54 images match, 108 refusals match, 0 mismatches, of 162 selections.

How I checked it

Emulator only (ot_emu from upstream octabam, MKII panel, DSP on, empty scratch card). Card images were composed by the site's builder from packages built in the isolated toolchain, then driven with scripts/capture-module-ui.py. Details are in TESTING.md.

• main (VECTOR 0.2.5 + POLY8 0.3.2 + FM Synth) reproduces the report: SRC▸VECTOR, but SRC SETUP is the stock page with TSTR AUTO.
• VECTOR 0.2.6 + POLY8 0.3.3 + FM Synth: SRC SETUP shows SEED 1, SPAN 12, OFST 0, ROT 0, RPT OFF, DIR FWD. Encoder A gives SEED 4, encoder E gives RPT 2, and the cursor on POLY8 shows the stock page.
• VECTOR alone, 0.2.5 vs 0.2.6: all eight screens of the same plan are pixel-identical.
• verify_native.py passes in the pinned toolchain (sanitizer tests; control.s reproduces).
• Not tested: a unit, MKI, save/reload, power cycle, listening.

Done since opening

• Analog BD proofs re-exported for the new module source tree: all 212 proofs are byte-identical, and only moduleSourceTreeSha256 moves.
• Tape Echo and Euclid standalone recipes re-pinned to the four imported asset files, as Analog BD 0.1.6 did. test_published_module_packages_rebuild_byte_for_byte passes again: every published package rebuilds byte-for-byte. An earlier version of this text called that failure local-only. It wasn't; this branch caused it.
• Merged current main.
• Refreshed 6 Nov 2026: merged main again (about 500 commits). Packages rebuilt in the isolated toolchain and imported; Analog BD proofs re-exported (all 212 unchanged, only the source-tree hash moves); Tape Echo and Euclid recipes re-pinned; the machine recipes name the merged machine-ports.json (images unchanged). npm run pr:ready: Ready.

Owner decision (6 Nov 2026): 0.2.4's card evidence stands

src/catalog/vector.test.ts now binds media/capture.json to 0.2.4's native source by name (1255e991…). It requires exactly native.c and control.s (the row-5 test) to differ from the capture's build inputs, and the manifest's qualification to name the current source. The 0.2.4 evidence was not re-run.

Qualification record

The qualification record moves to 0.2.6 the way 0.2.5's did in #1017. Its hardware report is 0.2.5's USB run on your MKII, and 0.2.6 keeps publishing that exact USB package. I added a limitation saying the 0.2.6 card change ran in the emulator only. Approved with the rest on 6 Nov.

The changelog tells players the page shows once POLY8 0.3.3 is out too. Both PRs can merge in either order: each alone changes nothing a player sees in POLY8 sets.

What the guides lacked: docs/ADD_A_MODULE.md didn't say that every package import needs Analog BD's proofs re-exported and Tape Echo's and Euclid's standalone recipes re-pinned. Fixed in #1144, kept out of this release. Nothing new in the 6 Nov refresh.

🤖 Generated with Claude Code

[repeat98]

Closing: VECTOR 0.3.0 (#1339) and 0.3.1 (#1342) carry this fix, so VECTOR's second page opens beside POLY8 on card builds (#1119). The POLY8 half stays in #1142.