You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: docs/EXECUTION.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -20,7 +20,7 @@ This file owns the day-to-day implementation workflow: how to choose authority d
20
20
-[`docs/TESTING.md`](TESTING.md) owns validation policy; [`docs/info/ROM-SUITES.md`](info/ROM-SUITES.md) owns external ROM materialization, report channels, and runner commands.
21
21
-[`docs/REFERENCES.md`](REFERENCES.md) owns the current consultation order for Pan Docs, AntonioND, Gekkio, GBEmulatorShootout, DocBoy, and other active references.
22
22
-`crates/gb-test-runner/data/reports.toml` and each report-local `sources.report.toml` own report fetch metadata for `cargo rom-fetch`, `cargo rom-suite`, and `cargo rom-suite-link`; update those files rather than documenting ad hoc source lists elsewhere. Local reports such as `linked` deliberately omit `sources.report.toml` because their assets are committed under `crates/gb-test-runner/data/<report>/`.
23
-
-`/test/gb-emulator-shootout/`, `/test/docboy/`, `/test/gbmicrotest/`, and standalone exploratory report roots such as `/test/mooneye/`, `/test/little-things-gb/`, `/test/magen/`, `/test/mealybug-tearoom-tests/`, `/test/samesuite/`, `/test/wilbertpol/`, `/test/rtc3test/`, or `/test/mbc3-tester/` are generated/local evidence channels; keep before/after copies when the mandatory external-ROM regression workflow applies.
23
+
-`/test/gb-emulator-shootout/`, `/test/docboy/`, `/test/gbmicrotest/`, and standalone exploratory report roots such as `/test/mooneye/`, `/test/little-things-gb/`, `/test/turtle-tests/`, `/test/magen/`, `/test/mealybug-tearoom-tests/`, `/test/samesuite/`, `/test/wilbertpol/`, `/test/rtc3test/`, or `/test/mbc3-tester/` are generated/local evidence channels; keep before/after copies when the mandatory external-ROM regression workflow applies.
Copy file name to clipboardExpand all lines: docs/REFERENCES.md
+2-2Lines changed: 2 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -29,13 +29,13 @@ Consult primary documentation and hardware research before emulator source code.
29
29
-[gb-cycle GBEmulatorShootout fork dashboard](https://pakitovic.github.io/GBEmulatorShootout/) — project-facing summary of the current gb-cycle counted rows.
30
30
-[blargg test ROMs](https://github.com/retrio/gb-test-roms) — CPU, timing, memory, OAM, and DMG/CGB sound ROMs, currently consumed through the curated GBEmulatorShootout source path rather than a direct retrio checkout.
31
31
-[Mooneye GB](https://github.com/Gekkio/mooneye-gb) — acceptance, boot, CPU/interrupt/timer/DMA, serial, and model-specific executable tests plus documentary reasoning.
32
-
-[c-sp/game-boy-test-roms](https://github.com/c-sp/game-boy-test-roms) — release-ZIP aggregate used for archive-backed Blargg, Mooneye, GBMicrotest, Wilbertpol, standalone SameSuite, standalone rtc3test, and standalone mbc3-tester channels, including independently pinned ROM and fixture hashes.
32
+
-[c-sp/game-boy-test-roms](https://github.com/c-sp/game-boy-test-roms) — release-ZIP aggregate used for archive-backed Blargg, Mooneye, GBMicrotest, Wilbertpol, standalone SameSuite, standalone rtc3test, standalone turtle-tests, and standalone mbc3-tester channels, including independently pinned ROM and fixture hashes.
-[MagenTests](https://github.com/alloncm/MagenTests) — official source/release origin for Magen CGB priority, HDMA, KEY0, PPU-off STAT, and MBC out-of-bounds SRAM executable tests.
37
37
-[docboy-test-suite](https://github.com/Docheinstein/docboy-test-suite/) and [docboy repository](https://github.com/Docheinstein/docboy) — large high-precision DMG/CGB/linked-session ROM corpus and a structural implementation cross-check for timing-sensitive behavior.
38
-
- GBEmulatorShootout/DocBoy/c-sp/MagenTests-sourced family names actively used by manifests include `acid`, `ashiepaws`, `ax6`, `blargg`, `cpp`, `daid`, `gbmicrotest`, `little-things-gb`, `magen`, `mealybug-tearoom-tests`, `mooneye`, `samesuite`, `wilbertpol`, `mbc3-tester`, `rtc3test` report families `basic-tests`, `range-tests`, and `sub-second-writes`, `docboy-dmg`, `docboy-cgb`, `docboy-cgb-dmg`, and `docboy-cgb-dmg-ext`; pinned source revisions, archive hashes, or file-base release asset URLs plus file hashes for report-local cargo lanes live in `crates/gb-test-runner/data/reports.toml` plus each report-local `sources.report.toml`; linked-session manifests use the report-local `*.link.suite.toml` model and share the same report registry.
38
+
- GBEmulatorShootout/DocBoy/c-sp/MagenTests-sourced family names actively used by manifests include `acid`, `ashiepaws`, `ax6`, `blargg`, `cpp`, `daid`, `gbmicrotest`, `little-things-gb`, `magen`, `mealybug-tearoom-tests`, `mooneye`, `samesuite`, `turtle-tests`, `wilbertpol`, `mbc3-tester`, `rtc3test` report families `basic-tests`, `range-tests`, and `sub-second-writes`, `docboy-dmg`, `docboy-cgb`, `docboy-cgb-dmg`, and `docboy-cgb-dmg-ext`; pinned source revisions, archive hashes, or file-base release asset URLs plus file hashes for report-local cargo lanes live in `crates/gb-test-runner/data/reports.toml` plus each report-local `sources.report.toml`; linked-session manifests use the report-local `*.link.suite.toml` model and share the same report registry.
Copy file name to clipboardExpand all lines: docs/TODO.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -21,7 +21,7 @@ PPU hardening toward a hardware-true engine (post PR #245): the sprite Mode 3 pe
21
21
-[PPU][MODE3-OBSERVATION-TABLES] Mid-line register-write landing on sprite lines is still resolved through per-sprite-`x`/`scx` phase-fitted tables (`PpuMode3ObservedLcdc3PhaseTable`, `Lcdc4PhaseTable`, `Lcdc2ObjSizePhaseTable`, `ScyObjPhaseTable` in `helpers/mode3_policies.rs`; the window-activation tilemap masks in `mode3/window.rs`; the `SCX` coarse-bit refetch gating) which PR #245 refit to the corrected schedule rather than deriving from the fetcher. Evidence: mealybug DMG + CGB-compat suites are green. Superseded direction: do not extend these tables for new cases. Highest-value next step: once [MODE3-FETCHER-LEAD] is fixed, resolve tilemap / tile-data / `SCY` / tile-select at the actual fetch dots and delete the tables.
22
22
-[PPU][MODE0-PUBLISH-HALT-GRID] The steady-frame published-`STAT` Mode 0 boundary override gated on `scx == 0 || Mode 0 IRQ enabled`, plus the Mode 0 halt-wake deferral table (CGB-family all `SCX` low bits, DMG-family `SCX&7` in `{1,2,5,6}`), compensate a measured 1-4 T-cycle offset between the halt-dispatch and nop-dispatch grids — wilbertpol `intr_2_mode0_timing_sprites_nops` measures on post-LCD-enable frames, mooneye halt variants on steady frames. Superseded direction: never refit sprite penalties (PR #245 froze them hardware-true). Highest-value next step: a uniform halt-wake dispatch convention plus recalibrated Mode 0 / Mode 2 pretrigger apertures, validated against SameBoy/DocBoy.
23
23
-[PPU][MODE3-SCY-OBJ-PHASE-POLICY]`m3_scy_change.gb` is green through `PpuMode3ScyObjPhasePolicy` plus `PpuMode3ObservedScyObjPhaseTable`; if a stronger oracle resolves the exact BG/OBJ handoff phase, replace the observed-table ranges with direct shared BG/OBJ fetcher arbitration rather than extending the table.
24
-
-[PPU][WINDOW-GLITCH-ORACLE] The active window ROM block is green (`m3_window_timing`, `m3_window_timing_wx_0`, `m3_lcdc_win_en_change_multiple*`, `m3_wx_*`), but stricter hardware/oracle evidence is still missing for `WX = 0`, `WX = 166`, `WX`/`WY`, and `LCDC.5` mid-frame glitches, especially `WX = 0 && (SCX & 7) > 0`.
24
+
-[PPU][WINDOW-GLITCH-ORACLE] The existing active window ROM block is green (`m3_window_timing`, `m3_window_timing_wx_0`, `m3_lcdc_win_en_change_multiple*`, `m3_wx_*`), but standalone c-sp v7 `turtle-tests/window_y_trigger_wx_offscreen.gb` now fails against its upstream framebuffer fixture while `turtle-tests/window_y_trigger.gb` passes. Stricter hardware/oracle evidence is still missing for offscreen `WX`,`WX = 0`, `WX = 166`, `WX`/`WY`, and `LCDC.5` mid-frame glitches, especially `WX = 0 && (SCX & 7) > 0`.
25
25
-[PPU][OAM-CORRUPTION-ORACLE] Deterministic unit/integration coverage exists for Mode `2` OAM access, `FEA0-FEFF` reads, `inc rr`, `[hli]` / `[hld]`, stack/interrupt paths, DMG variants, and the CGB negative path; the promoted GB Emulator Shootout `oam_bug` subset is green, and the standalone `blargg` report now keeps `oam_bug.gb` plus `7-timing_effect.gb` green with pinned framebuffer fixtures, but independent oracle comparison is still missing.
26
26
-[PPU][FF44-HBLANK-SEAM] The exact DMG `FF44` advance point inside late HBlank remains hypothesis-only; a direct later-dot retune regressed `mooneye acceptance/ppu/hblank_ly_scx_timing-GS`, so re-entry must start from a narrow trace or oracle comparison around the late-HBlank `LY` / `SCX` polling seam while keeping that ROM green.
27
27
-[PPU][LCDC2-8X16-ARTIFACTS] Core `8x16` rules and the mid-frame `LCDC.2` shrink crash are fixed through a narrow model; re-check finer DMG-visible artifacts only with a targeted ROM or oracle, not by broadening the current line-start latch / queued-FIFO / repaint policy speculatively.
0 commit comments