Entry 94

The staff roll: the ending's credits as ccThStaffRoll runs them, checked frame by frame

2026-09-25uiaudiotest

Filescrates/piney-desktop/src/staffroll.rscrates/piney-desktop/src/lib.rscrates/piney-desktop/src/view.rscrates/piney-desktop/tests/staffroll.rscrates/piney-desktop/tests/staffroll_fixture.txtcrates/piney-game/src/desktop.rscrates/piney-game/src/mode.rscrates/piney-game/src/main.rstools/test_staffroll_rs.pydocs/engine/desktop.md

With story starts to the ending (93), story:31 reached the instruction staff_roll. The desktop host had nothing for it: its wait ended at once, and the credits never showed. The desktop page listed ccThStaffRoll as unknown.

What the game does

The details are on the desktop page.

• The task. ccThStaffRoll (desktop.prg 0x004125c0) loads stfroll1 and builds a ccThStaffRollCtrl of 19 lines. It plays BGM.BIN track 1 and runs Main once a frame until the controller is done.
• The pages. There are 28 (g_srDataGrp): credits and a picture, or a picture alone.
• The four phases.
  1. _Random. The view zooms in from 5 to 1. Every line shows 44 random characters (ccRand) and settles one more each frame into the credits (StopChar, the C library's rand, seeded by the frame count).
  2. _Fix. The random characters fade while the picture fades in behind the credits.
  3. _End. The page holds for 110 frames, then shrinks and fades.
  4. _BGOnly. A page with a picture alone fades in, holds and fades out.
• Names. A credit's #0 and #1 are the save's two names (ccTransCode2Name).

The port

• The controller. piney-desktop's staffroll.rs keeps the fields under their offsets' names and records each frame's draws. StaffRoll::render draws them with the port's ccKanji (the LargeFixed font) and the page picture's mask.
• The task. Desktop::start_staff_roll: two breaths, then Main each frame, and ccBgmPlay(1) / ccBgmStop as new requests. They reach the audio as Event::BgmStream and Event::BgmStreamStop (Audio::bgm_stream, bgm_stream_stop).
• The desktop host. begin(StaffRoll) starts the task, busy(StaffRoll) waits for it, and end does nothing.
• The layer centre. The layer's view needed a centre (LayerView::frame_centred: SetFrame then SetLayerCenter).
• Ending music. bgm_control came in 4d7a889.

Two mistakes the check caught

• Except's column. v * 0x2aaaaaab >> 32 is v / 6, and the sra 1 after it makes it v / 12, the column. The first port used v / 6.
• The picture's MakePacket flag. _Fix and _End never set $a2 before the call; it still holds the 1 stored in wi just before. The first port passed 0.

Checked

• The whole roll against the game. tools/test_staffroll_rs.py runs the game's controller natively in eemu over all of it (7,231 frames, seed 12345, names Kite and Tester, ccRand as at boot). Only the drawing calls are hooked, and they are recorded.
• What each frame keeps. The controller's words, every line's bytes, and the draws (hashed). The fixture is 650 KB; detail N prints a frame in full.
• The replay. crates/piney-desktop/tests/staffroll.rs replays every frame with 0 mismatches.
• The view. The harness also runs the game's ccView: SetFrame at one scale, SetLayerCenter, then SetFrame at another (as _Random and _End do each frame). All 36 points through ApplyLayerScreenMatrix match LayerView::frame_centred. So a later SetFrame keeps the centre.
• The rest. The workspace's tests, clippy, fmt and the docs check pass.
• Shots. From story:31: "Story, Kazunori Ito" over Kite's picture, and the CAST page (the English voice actors in two columns).