Sylpheed/docs/re/structures/ui-keyframe-rotation.md

✅ The keyframe block's +12 is a screen-plane ROTATION, in degrees

Status: ✅ DECODED — field, disc-wide census, and confirmed against the framebuffer (a GPU draw capture), not against our own renderer. 🟡 +4 and +8 remain unexplained.

This closes the ⚠️ that stood on the Keyframe type: "they do not explain the title screen — every element of GP_TITLE build 4 has all three at zero, while the game demonstrably submits rotated quads there." That note was wrong about reach, not about the bytes it read.

The block

+0   u32  ARGB fade colour
+4   i32  🟡 unexplained
+8   i32  🟡 unexplained
+12  i32  ✅ rotation, DEGREES, clockwise-positive in screen space (Y down)
+16  u32  scale X, percent
+20  u32  scale Y, percent
+24  u32  tint
+28  i32  X   ← signed
+32  i32  Y   ← signed
+36  u32  time  (the last block of a group does not own this word)

How it was confirmed: two quads, two different angles, measured

The live title submits one draw of two rotated quads that our renderer does not reproduce (title-draw-capture-vertex-colours.log, draw 2). Converting its NDC vertices to screen space — X = (x+1)·640, Y = (1−y)·360 — and taking the quads' edge lengths rather than their bounding boxes:

quad	size	rotation	centre
A	400.1 × 1076.3	+30.26°	(992.0, 359.1)
B	400.2 × 1444.5	−45.28°	(467.2, 360.0)

Both are true rectangles (adjacent edges perpendicular to < 0.6 %).

The two elements they belong to are the title's nested leaf records ptloop01.rat → pteff03.t32 and ptloop02.rat → pteff03a.t32. Their keyframe blocks, at 0xbb5966+0x68 and 0xbb5a82+0x68 in the decompressed build 4:

	ptloop01	ptloop02
sprite	pteff03.t32 399×180	pteff03a.t32 399×180
scale	100 %, 600 % → 399 × 1080	100 %, 800 % → 399 × 1440
+12	30	−45
measured	400.1 × 1076.3 at +30.26°	400.2 × 1444.5 at −45.28°

Magnitude and sign, on two different values, from the GPU.

Why this is an identification and not a coincidence

The same capture was first attributed to a different element entirely (REFUTED.md), so the identification is carried by five independent agreements, not by elimination:

1. The draw has exactly two quads; the screen declares exactly two ptloop elements.
2. Both widths measure 400 against a declared sprite width of 399.
3. The heights are different numbers that both land: 1076 ≈ 1080 (600 %) and 1444 ≈ 1440 (800 %). A coincidence would have to hit both.
4. Between the capture's two frames quad A moves +19 px right and quad B moves left — matching each record's own decoded sweep direction (ptloop01 x: −639 → 1521, ptloop02 x: 1721 → −839).
5. The drawn vertex alphas 0xC3 / 0xB6 both fall inside the declared 0x80 → 0xff alpha ramps.

A second corroboration, by ramp shape rather than by angle

GP_BUNK entry 117ca14f at 0x7395f3 holds a two-keyframe group in which +12 ramps 0 → 360 while position, scale and alpha all stay constant. A full-turn ramp that changes nothing else is a spin in place; no non-rotation reading of the field explains it. Disc-wide, 360 occurs 1 173 times and −360 180 times.

✅ The strongest evidence is GP_DIALOG, and it is visible in the CLI

sylpheed-cli screen info --build 0 --geometry dat/GP_DIALOG.pak. These are top-level elements, so no hex dump is needed to see them:

element	+12 across the group	what else moves
pceff03.t32	90 → 30 → 10 → 3 → 0	alpha 0 → 128 → 192 → 224 → 255; position slides (630,348) → (470,198)
pceff04.t32	90 → 30 → 10 → 3 → 0	the same ramp, mirrored path (413,77) → (553,247)
pzeff02.t32 (build 6)	43 → 61 → 75 → 90	scale 112 % → 200 %, alpha 255 → 0

The first two are a swing-in that decays to zero exactly as the element settles; the third is a burst that spins as it expands and fades. A monotone decay to 0 synchronised with a fade-in, and a spin-out synchronised with an expanding fade-out, are what a rotation field does and are not what any other reading of the word would produce.

GP_DEBRIEFING_PILOTLOG build 10 holds the static case: pjeff24a.t32, one keyframe, r=90 at 210 % scale.

These four rows are committed verbatim at data/rotation-top-level-examples.txt. ⚠️ The sweep that found them — every build of every GP_*.pak, grepped for a non-zero r= — had not finished at the time of writing; it was in GP_DIALOG build 37 with two archives hit so far. Treat the count as incomplete and the existence as settled.

✅ Bonus, from the same two quads: the pivot-anchored scale is measured-correct

ui_layout::blit places a scaled element as ox = kf.x − pivot·(scale−100)/100, keeping the pivot fixed rather than the top-left. That was implemented and reasoned about, never measured — and until now there was nothing to measure it against, because the title's other elements are all at 100 % scale where the term vanishes. The ptloop pair scale 600 % and 800 % vertically, so the term is worth 450 and 630 px:

element	measured centre Y	pivot-anchored	top-left	position-as-centre
ptloop01 (600 %)	359.1	360.0 ✅	810.0	270.0
ptloop02 (800 %)	360.0	360.0 ✅	990.0	270.0

Two different scales, both landing on 360.0. Horizontally the same term is worth 199.5 px, and there is an independent cross-check available: solve the group's t from the quad's position, then from its vertex alpha, and see whether the two agree.

element		t from position	t from alpha	Δ
ptloop01	with the pivot term	571.98	571.65	0.33
	without	579.65	571.65	8.00
ptloop02	with the pivot term	668.92	668.27	0.65
	without	659.71	668.27	8.55

🟡 What this does not show is that interpolation is linear. Both fields were inverted through the same linear map, so a shared easing curve would cancel out and still agree. What it does show is that position and alpha are driven off one shared parameter — which is worth knowing on its own, and is why the agreement is a real check on the placement formula rather than a tautology.

Disc-wide census

tools/re-capture/kf_rotation_census.py, output committed at data/kf-angle-census.txt. Over 83 862 keyframe blocks in the GP_*.pak UI archives:

word	non-zero	dominant values
+4	3 990 (4.76 %)	180 (3 880)
+8	3 878 (4.62 %)	180 (3 102), 90 (201), −180 (156)
+12	12 164 (14.50 %)	90, −90, 360, 120, 180, and a long tail of arbitrary angles (53, −58, −125, 114, −33, …)

+12's long tail of arbitrary values is itself part of the decode: +4 and +8 are almost entirely ±180/±90, which is the signature of a flip flag expressed in degrees rather than of a free angle. 🟡 That is a hypothesis about +4/+8, not a decode — nothing observed turns on them.

⚠️ Reach and limits

• rotation_deg is decoded but NOT rendered. ui_layout's blitter draws axis-aligned quads only, so sylpheed-cli screen render still paints a rotated element upright. Rotating the blit is a real change to the compositor and was not attempted here. The title render's residual band is expected to persist until it is.
• 🔴 "Rotation appears to live only in nested .rat leaf records." Refuted by my own sweep, within the hour. It held for GP_TITLE, GP_BUNK and GP_CHALLENGE — the three archives I had checked when I wrote it — and fails on the very next one: GP_DIALOG and GP_DEBRIEFING_PILOTLOG carry rotation on top-level elements, and those are the best evidence on the disc (below). Rotation lives in both places. The half that survives is the one that matters to a consumer: a composer that reads only the declaration table will miss the title's rotations, because those particular ones are nested.
• The blocks are not 4-byte aligned. A nested RATC blob can start at an odd offset (ptloop01.rat at 0xbb5966), and its blocks inherit that. Any scanner over raw bundle bytes must not assume alignment — see METHOD.md.
• +4 and +8 are not claimed as X/Y rotation. Three adjacent signed degree-valued words invite that reading; only +12 has been observed to do anything.

✅ The top-level census, finished (2026-08-28)

The earlier attempt drove sylpheed-cli screen info --geometry in a shell loop and never finished — it decodes every texture per build. tools/re-capture/rotation_toplevel_census.py walks the top-level placement region the way ui_layout.rs does; seconds instead of hours. Output: data/rotation-toplevel-census.txt.

Both controls pass — GP_TITLE build 4 reports 0 (its rotations are the nested ptloop records) and GP_DIALOG build 0 reports the 2 expected pceff03/pceff04 elements.

top-level elements with a keyframe group	15 493
carrying a non-zero rotation	2 152 (13.89 %)

By archive: GP_READY_ROOM 1 868, GP_DIALOG 40, each of the six GP_MAIN_GAME_*2D 37, GP_GAMEOVER 10, GP_DEBRIEFING_PILOTLOG 8, GP_TITLE 4.

Two more corroborations, both free

The rotated population is tactical-map ship icons. The most-rotated names are pbb_destroyer (444), pbr_destroyer (402), pbr_fighter (276), pbb_fighter (174), pbb_aircraft (168) — map markers, in the archive that holds the tactical map. Rotating a ship icon to its heading is what the field is for, and it is the single largest use of it on the disc.

GP_TITLE entry 7's four rotated elements carry their own control. Three of them are the Japanese wordmark's pieces, and they settle to upright from alternating tilts:

ptlogo3a.t32   r = 0, -14, -4, -1, 0, 0, 0, 0
ptlogo3b.t32   r = 0, +14, +4, +1, 0, 0, 0, 0
ptlogo3c.t32   r = 0, -14, -4, -1, 0, 0, 0, 0

Same magnitudes, alternating sign, all decaying to 0 — three sibling elements tilting opposite ways and straightening. A misread field does not produce that. (The fourth is ptlogo_eff3.t32, r = 0, 0, 80, 150 — the transient bloom discussed in ui-resting-pose.)

⚠️ This is not subtractable from the byte-scan census. That one counts blocks (12 164 of 83 862 with a non-zero +12); this one counts elements. The two are different units and the difference is not "the nested population".