docs: v3.11.0 release validation

Co-Authored-By: Claude Opus 5.5 <[email protected]>
This commit is contained in:
Joseph MattielloandClaude Opus 5.5 committed 2026-11-01 20:14:20 -05:00
1 parent 5a976df863
commit 8ad0d2b644
1 file changed
+32 -1
+32 -1
View File
@@ -150,4 +150,35 @@ portable and installer packages; Linux has AppImage and Debian packages.
## Validation
To be filled in from the release gates on the Mac Studio.
Candidate `5a976df8` passed a clean build on the Mac Studio. The library reports `v3.11.0`, and the
Makefile, `.info` and fallback versions agree. The full core suite (`make TEST_EXPORTS=1 test`) passed
with the private corpus, including:
- the audio presence and clipping tests;
- the TOM readback and MMULT tests;
- the GDB stub tests;
- the idle-skip A/B tests, which confirm that skip-on and skip-off output is bit-identical;
- the Doom pace gate.
Six checks were skipped because their fixtures or tools were missing on that host.
Against v3.10.0:
- **Cartridge matrix (155 titles, HLE and BIOS):** zero backward moves with either blitter. Gorf 2000
improves in both modes, from a trap loop to game code. The three known HLE/BIOS asymmetries are
unchanged.
- **CD matrix (28 title/mode rows, full disc corpus):** zero backward moves.
- **Corpus A/B (160 cartridges, 300 frames with audio):**
- 157 are bit-identical, with the DSP idle skip off and on.
- Atari Karts, Gorf 2000 and White Men Can't Jump differ, as expected. All three read TOM's horizontal
counter, which #1094 changed. None of them differs in audio.
On the Mac Studio, in both a quiet and a loaded state, the app ran Cybermorph and yarc with the thread
off:
- Repeated and dropped images fell from 19-53 a minute to 0-1 repeats.
- Skips stay near the 3 a minute that a 60.05 Hz core has to make on a 60 Hz display. A loaded host
reached up to 8 a minute.
- PAL is unchanged.
GitHub CI on the release pull request is the remaining gate.