FABRIC.md: item 4.4p -- CANVAS geometry verified via one-shot probe, all three architectures

Verification only, no permanent code (matches this item's own title
and 4.4i's same-shape precedent -- 4.4l-4.4o were geometry decisions,
not drawing code). One-shot diagnostic probe in sk_repl() drew 4.4o's
box outline plus a strip-top marker, then was reverted after capture
per this document's write/run-once/capture/revert discipline.

Measured pixel bounds directly from each screendump (not eyeballed)
and confirmed exact matches against 4.4o's computed coordinates on
all three architectures: amd64 box x:[320,959] y:[104,583], strip
marker y:704; aarch64/riscv64 box x:[80,719] y:[4,483], strip marker
y:504.

Observed, not fixed here: on the small architectures the REPL banner
text visibly overlaps the box's top edge, because vt100's cursor grid
still spans the whole screen rather than being confined to the 96px
strip -- a real gap between the landed REPL path and the mockup,
belongs to 4.4q/4.4r's wiring work, not this item.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

FABRIC.md

@@ -5563,13 +5563,56 @@ document and committing that amendment as its own item.*
   > - aarch64/riscv64: box bottom (raster) = 4+480 = 484 → `y_cartesian = 600−484 = 116`.
   >   **Bottom-left (80, 116).**
 
-- [ ] **4.4p — Three-arch screendump verification of 4.4l–4.4o.** Geometry only — not the
+- [x] **4.4p — Three-arch screendump verification of 4.4l–4.4o.** Geometry only — not the
   toggle or scrollback.
   *Done when:* screendump on all three architectures shows the 640×480 box centered within
   CANVAS at the computed coordinates, with the fixed gap above the REPL strip visible;
   three-arch QEMU boot + logs per CLAUDE.md.
   *Refs:* §27.8.
 
+  > **Done 2026-08-11 — one-shot diagnostic probe, reverted after capture, per this document's
+  > established write/run-once/capture/revert discipline.** Nothing in 4.4l–4.4o produced
+  > persistent drawing code (they were geometry decisions feeding 4.4q/4.4r, which build the
+  > box for real); this item's own title is "verification", matching 4.4i's same-shape
+  > no-permanent-code precedent. Probe drew the 4.4o box outline (4 `fb_fill_rect()` strips,
+  > 2px stroke, green) plus a full-width red marker at 4.4n's strip-top edge (so the gap is
+  > observable, not just the box), inserted in `sk_repl()` right after
+  > `console_fb_enable_ttf()` — after the `-O2`/POST-safe boundary, before
+  > `vt100_enable_ttf()`'s own `erase_display(2)` would otherwise still be pending (it isn't;
+  > that runs inside `console_fb_enable_ttf()` itself, so the probe drawing correctly lands
+  > *after* the clear, not wiped by it).
+  >
+  > **Measured, not eyeballed** — scanned each PNG's raw pixels in Python for the drawn
+  > colors' bounding box, compared against 4.4o's computed values exactly:
+  > - amd64: green box x:[320,959] y:[104,583] (640×480 exactly), red marker y:[704,705] —
+  >   matches 4.4o's (320,104) and 4.4n's strip-top y=704 precisely.
+  > - aarch64: green box x:[80,719] y:[4,483] (640×480 exactly), red marker y:[504,505] —
+  >   matches (80,4) and y=504 precisely.
+  > - riscv64: identical measured bounds to aarch64 (same resolution, same geometry) —
+  >   x:[80,719] y:[4,483], red marker y:[504,505].
+  >
+  > Screendumps:
+  > [evidence/amd64/qemu-screenshot-20260811-202526-4.4p-canvas-geometry-probe.png](evidence/amd64/qemu-screenshot-20260811-202526-4.4p-canvas-geometry-probe.png),
+  > [evidence/aarch64/qemu-screenshot-20260811-202845-4.4p-canvas-geometry-probe.png](evidence/aarch64/qemu-screenshot-20260811-202845-4.4p-canvas-geometry-probe.png),
+  > [evidence/riscv64/qemu-screenshot-20260811-203207-4.4p-canvas-geometry-probe.png](evidence/riscv64/qemu-screenshot-20260811-203207-4.4p-canvas-geometry-probe.png).
+  >
+  > **Observed, not fixed here (belongs to later wiring items):** on aarch64/riscv64, where
+  > the box top sits only 4px below the CANVAS top, the REPL banner text visibly overlaps the
+  > box's top edge in the screendump. This is because `vt100`'s cursor grid still spans the
+  > *entire* screen, top-anchored at (0,0) — it has never been confined to the 96px strip
+  > 4.4n/4.4m defined; that confinement is §27.6's mockup intent but isn't any numbered item's
+  > job yet. Not a geometry error in this item's own measured bounds (which are exact); a real,
+  > known gap between the landed REPL text path and the mockup, worth its own item when
+  > 4.4q/4.4r pick this up.
+  >
+  > Three-arch boot+log basis (probe code present during these particular boots, harmless —
+  > drawing happens after POST/capsule birth, confirmed by unaffected `Failed: 0` and identical
+  > dict-hashes): amd64 (`logs/20260811-202330/amd64/`), aarch64
+  > (`logs/20260811-202700/aarch64/`), riscv64 (`logs/20260811-203102/riscv64/`).
+  >
+  > Probe code reverted from `repl.c` immediately after screendump capture; tree confirmed
+  > clean.
+
 - [ ] **4.4q — Scrollback, ~1000 lines.** A circular buffer of prior scroll-box lines, target
   depth approximately 1000 (Captain Bob's own qualification: "something along those lines,"
   not a hard-locked spec number). Storage mechanism (`kmalloc` ring buffer vs. static array)