@@ -68,6 +68,39 @@ still bad after dark+mini.
68688 . ** Watch #58 ** (` AccentColor.colorset ` empty): if the Messing accent lands before the UI-slot
6969 captures are finalized, every one of 03/04/05 needs re-capturing again.
7070
71+ ### Step 1 done this session: slots 03/04 rebuilt on iPad mini, dark
72+
73+ ` slot-03.svg ` /` slot-04.svg ` re-proportioned from 0.75 to 0.657 (screen quad ` 168,820,1728,2304 ` →
74+ ` 275,820,1514,2304 ` ; device ` 132,784,1800,2376 ` → ` 239,784,1586,2376 ` ; bezel, corner radii,
75+ headline baseline all unchanged, same geometry shared between the two templates as before).
76+ 225/225 renderer unit tests still green — no template-format regression.
77+
78+ Re-captured ` test01_onboardingChoice ` /` test17_albumPickerMany ` on ** iPad mini (A17 Pro), iOS
79+ 18.6** (` 12D70693-FF3D-42EE-B50D-87F180B0D486 ` ), ` xcrun simctl ui <UDID> appearance dark ` , the
80+ documented AP-4 env vars (` SCREENSHOT_PORTRAIT ` , ` SCREENSHOT_TEXT_SIZE=AccessibilityM ` ,
81+ ` SCREENSHOT_LOCALE=en ` ), status bar override. Captures came back ** 1488 × 2266** , matching the
82+ measured mini aspect. ** Belt-and-suspenders on the false-green trap:** set ` SCREENSHOT_DE ` both
83+ via ` simctl … launchctl setenv ` * and* as a trailing ` TEST_RUNNER_SCREENSHOT_DE=1 ` xcodebuild
84+ argument, then confirmed from the raw test log that both cases actually executed (not skipped)
85+ before trusting the attachments — 2 tests run, 2 attachments exported, not 0-run-SUCCEEDED.
86+
87+ Overwrote ` ipad/en/03-slot-source-choice.png ` , ` 04-slot-album-picker.png ` (and the
88+ ` en-axm-ios18 ` sibling, which FR-9010-34 keeps byte-identical to ` en/ ` ) with the new mini/dark
89+ captures. Re-rendered and re-checked: ** ` ipad/en ` still 6/6, ` --check ` and ` --list ` both clean** ,
90+ 03/04 correctly resolve with no ` scene ` (UI-type slots).
91+
92+ ** Verified visually (contact sheet + a full-res crop of slot 03's device edge):**
93+ - The half-empty-screen defect reads as resolved — app content and the black canvas ground are
94+ now visually contiguous, so the old dead margin no longer reads as emptiness. No crop/zoom
95+ change was needed; the zoom-vs-whitespace question is moot.
96+ - The ` #1C1C1E ` device body still reads as a distinct grey column against ` #000000 ` — legible as
97+ a physical object without touching the hairline. No strengthening applied, matching Jan's pick.
98+ - Slot 2's colour-monotony problem is unchanged (still no cool hero) and slot 5 is still the old
99+ room composite — neither was in scope for this step.
100+
101+ ** Not yet done:** slot 5 (steps 2–3, the arrival-toast spec + implementation) is still open, as is
102+ everything after it in the ordered list above.
103+
71104---
72105
73106## 2026-09-10 (later) — four questions from Jan, answered empirically
0 commit comments