[gHashTag]

Refs #13296 #13297 #13298 #13214

Style lane, stacked on #13315 (renderers). Merge order: #13254, then #13315, then this PR, then the uikit lane.

What changed

The renderers now draw the UI Kit frame that already exists. They carry no palette or glyph table of their own. No new style spec was added: theme, tokens, type and painter were already specs, so this PR wires them in.

Source	What the renderers take from it
vf_theme.t27 (TH_T27)	Ground 0x000000, text 0xFFFFFF, accent 0x00FF88, outline 0x333333, control fill 0x141414
vf_card_tok.t27	vct_render: the token style (26 Bold heading, 16 Regular body, padding, border) in one call
vf_type.t27	Outfit glyphs
vf_card.t27	The painter rule, now public: vc_box / vc_text return what vc_node paints, and vc_fill / vc_off_m are pub

• web: the draw list (FILL + TEXT ops, 6 words each) is built from vc_box / vc_text. Replaying it equals the raster byte for byte.
• native: t27os_blit gets the vct_render raster, and only on VC_OK.
• fpga: the AXI4-Lite write plan is FB_W, FB_H, then 8 box fills (6 writes each, 50 writes so far). Text follows as 1-px-high fill spans. No glyph ROM is needed. The fill-engine model replays the plan to the vct_render frame in every byte.
• main: VF_TARGET=web|native|fpga ./m > out.ppm.

Evidence

Tests (t27b):

Spec	Result
vf_render_web	7/7
vf_render_native	2/2
vf_render_fpga	4/4, with an n-1 negative control and a refused out-of-window address
vf_render_main	2/2: three targets give one frame, with a one-byte negative control
vf_card_tok	11/17, including the new vct_render_is_the_token_style (heading-size negative control)
vf_card	10/21

Artifact:

• web.ppm, native.ppm and fpga.ppm are byte-identical: 121515 bytes each, sha256 36d330f03a1c50462d0bb047c87be030475f9fa74cce0c51e3898e1337769963.
• Pixel (0,0) = 0x333333 (outline); pixel (17,53) = 0x00FF88 (accent).

Gates: tools/dupe_scan.py reports no new or grown duplicate body. The lefthook own-language and secret-gate checks pass.

Known gap, not fixed here

Six tests in vf_card_tok and eleven in vf_card fail. They fail at the same asserts on the HEAD copy of vf_card.t27, before this PR. The cause is layout drift: card heights, a title top of 20, and a min-target check that follow vf_layout text wrap, last touched on the #13254 lane. These belong to #13254 or a follow-up, not to the style lane.

Not done: no seal (none was asked for), no window has shown the native frame yet, and no board has run the FPGA plan.

A session merges only a PR it opened itself, and only when its brief or the owner says to. It never merges, approves, closes, retargets or force-pushes a PR or branch that another session, bee or person opened, even when that PR is the one holding its own gate.

🤖 Generated with Claude Code

---

Note

Medium Risk
Large cross-cutting renderer refactor with new public painter APIs and FPGA span encoding; risk is drift or performance in the text-span plan, mitigated by byte-for-byte tests across targets.

Overview
Unifies web, native, and FPGA on one UI Kit frame: theme TH_T27, token metrics via vct_render, and Outfit text—no per-target palettes or 5×7 glyph ROM.

vf_card now exposes a single placement rule: public vc_box, vc_text, VcBox, VcText, and vc_fill. vc_node only replays those records into the raster so vector paths cannot drift.

vf_card_tok adds vct_render / vct_render_style() as the shared raster entry for all three renderers.

vf_menu takes padding and gaps from vct_metrics(), height 108 (was 84), and Browser alone is PRIMARY; Desktop/FPGA are SECONDARY.

Web: vf_draw_ops is built from vc_box / vc_text; vf_ops_replay must match vct_render byte-for-byte. Canvas text uses Q8 pen, baseline, px, and weight.

Native: blits the vct_render raster; returns VC_* instead of VF_R_*.

FPGA: drops GLYPH commands; boxes from the draw list, text as 1px-high FILL spans after diffing vct_render vs a base replay. The engine model uses vc_fill; tests pin equality to the full raster (121500 bytes).

vf_render_main: all VF_TARGET values emit the same 375×108 PPM from that unified pipeline.

^{Reviewed by Cursor Bugbot for commit 75f8206. Bugbot is set up for automated code reviews on this repo. Configure here.}