[JoeMatt]

Closes #1010

Draft: stacked on #1006, which is stacked on #1000. The branch is cut from fix/app-disk-control-images. The new commits are the top two: fix(app): follow a disc swap in the content identity... and docs(app): .... Merge #1000, then #1006, then rebase this one. It targets develop because app CI only runs on PRs into develop.

Problem

After a disc swap (File > Disc, or Add disc image...), the core reboots on the new disc. The app still named the first disc as the content, so:

• Save states went to the first disc's slot. A state saved on disc B landed in disc A's slot. The core refuses that state whenever A is mounted (savestate v14 records which disc it was taken on).
• Everything else named disc A too: achievements, cheats, the window title, Discord, and the TV menu title.

What follows what

Follows the disc in the tray (contentPath())	Stays with the opened disc (loadedPath())
Save-state slots and thumbnails, including the auto slot. The core only accepts a state on the disc it was taken on.	The .srm file, and the fallback copy a second app instance reads.
Achievements, through rc_client_begin_identify_and_change_media. A disc of the same game keeps the session and hardcore. rcheevos decides what an unknown disc means.	Restart (reloadContent).
Cheats re-key, because the core re-reads the per-title row on insert (titledb_probe_disc).	Recents and the library's recently played.
Window title, Discord, TV menu title. These read contentPath every frame and needed no change.	Per-game control overrides (input.setContent). These are picked once per open; left unchanged to keep scope.

Why the .srm stays. The core's save RAM is the EEPROM pair plus the Memory Track. It survives an insert and is never reloaded (disk_set_eject_state only re-installs the NVM BIOS module). So the buffer still holds what was read from A's .srm, plus anything B wrote.

• Re-keying the file to B would write A's Memory Track into B.srm.
• It would also split one Memory Track across two files.

Keeping the save with the opened content is what RetroArch does. It is also what an .m3u session would do.

Recents follow the same reasoning. Reopening from recents is a fresh load, and it reads the .srm of the entry it opens. Disc A's entry is the one that brings the session's saves back.

Change

CoreHost

• contentPath() / contentName() now name the disc in the tray. followMountedDisc moves them only when an insert worked and left the tray closed.
  • The image index alone is not enough: the core moves disk_index before it tries the insert, and leaves it moved when the insert fails (it puts the previous disc back and leaves the tray open).
  • Off the owner thread, these answer from EmuSnapshot, because discSetEject / discSetIndex may run as queued commands.
• New loadedPath() / loadedName() hold what was opened; saveRamPath() uses them. New discSwaps() counts swaps.

App::followDiscSwap()

• Runs once per UI iteration, after addDisc, and in the headless check, so it catches every disc menu: ImGui, native and TV.
• On a swap it drops a queued state load and the resume offer that belonged to the previous disc.
• In one call(), it hands the swap to achievements and re-keys cheats, then drops the cached state thumbnails.

Achievements — onDiscChanged()

• If the first disc was never identified (.chd, .cdi, or signed out), it starts a fresh identify.
• If rcheevos disables hardcore for an unidentified disc, the app leaves it off. Calling applyHardcore() would have turned it straight back on.

Tests

--disc-swap (discSwapHeadlessCheck) now checks, after each step and as the UI reads it (the pinned snapshot when threaded):

• After the add: contentPath, contentName and statePath(slot) name disc B. loadedPath and saveRamPath are unchanged. A state is saved, and synth_jagcd_multi.slot0.state exists.

• With the tray open: the content is still disc B.

• Back on disc A: the content is A again, and exactly one swap was counted.

• New line: it prints DISC content synth_jagcd -> synth_jagcd_multi -> synth_jagcd, save RAM synth_jagcd.

• Data dirs: smoke_disc_swap, smoke_disc_swap_threaded and smoke_disc_swap_bios_threaded each get their own --data-dir in the build tree, so the saved state never reaches a user folder.

• Negative check: with followMountedDisc disabled, smoke_disc_swap fails with after the add, the content is '.../synth_jagcd.chd' and the state slot is .../synth_jagcd.slot0.state.

• Full app suite: 117/117 pass on macOS arm64 after the restack, including all 21 emu_thread tests.

• TSan, local (macOS): the 11 emu_thread tests that don't run the app binary pass. That includes core_host_thread, which queues discSetEject / discSetIndex (the snapshot write path). macOS refuses to load the sanitizer into the app binary ("Sanitizer load violates platform policy"), so the threaded disc-swap tests get TSan only from the Linux tsan job.

• CI from the base: before fix(app): register the loaded CD with disk control so the disc menu works (#1003) #1006 was rebased, its ThreadSanitizer job was red on a lock-order inversion in smoke_yarc_threaded, a cart run with no disc. Its macOS job also failed emu_thread_unit once. Both come from feat(app): emulation on its own thread, phase 1 (#967) #1000, and its 4d7668f targets them. This branch now sits on fix(app): register the loaded CD with disk control so the disc menu works (#1003) #1006 at 08b340f, on top of feat(app): emulation on its own thread, phase 1 (#967) #1000's tip c7de134.

Not covered:

• Real disc identification for achievements. The fixtures are .chd, which rcheevos cannot hash.
• The windowed menus. Per the repo rule, no app windows are opened on this Mac.

Not in this PR

• Resume after quitting on disc B. The auto-save goes to B's slot. Reopening disc A from recents offers A's older auto-state. B's auto-state is reachable only by opening B directly, which reads B.srm. This follows from the table above. A real fix needs a multi-disc set identity, such as .m3u.
• Per-game control overrides stay with the opened disc.
• Cheats lag by one frame. On a swap from the ImGui or native menus, cheats re-key on the next loop iteration, so B's first frame still runs with A's cheats.

🤖 Generated with Claude Code

[github-actions]

No automated review was posted on this PR — the plan is out of quota for the current billing cycle. This is a billing limit, not a repo or credential problem — it resumes on its own when the quota refreshes.

_{This PR has not been machine-reviewed. Posted by the review workflow so that an unreviewed PR is not mistaken for a clean one; it updates in place rather than repeating on every push.}

[github-actions]

Regression: linux-arm64

Regression Test Results

ROM	Status	Details	Diff
jagniccc	✅ PASS	0 pixels differ	-
yarc	✅ PASS	0 pixels differ	-
jagniccc (determinism)	✅ PASS	identical across runs	-
yarc (determinism)	✅ PASS	identical across runs	-
jagniccc (frameskip)	✅ PASS	skip=0 matches skip=3	-
yarc (frameskip)	✅ PASS	skip=0 matches skip=3	-
jagniccc (save state)	✅ PASS	round-trip matches	-
yarc (save state)	✅ PASS	round-trip matches	-
jagniccc (rewind)	✅ PASS	rewind matches	-
yarc (rewind)	✅ PASS	rewind matches	-

Platform: Linux aarch64

_{Updated by CI at 2026-10-17T23:45:26.296Z}

[github-actions]

Regression: linux-x64

Regression Test Results

ROM	Status	Details	Diff
jagniccc	✅ PASS	0 pixels differ	-
yarc	✅ PASS	0 pixels differ	-
jagniccc (determinism)	✅ PASS	identical across runs	-
yarc (determinism)	✅ PASS	identical across runs	-
jagniccc (frameskip)	✅ PASS	skip=0 matches skip=3	-
yarc (frameskip)	✅ PASS	skip=0 matches skip=3	-
jagniccc (save state)	✅ PASS	round-trip matches	-
yarc (save state)	✅ PASS	round-trip matches	-
jagniccc (rewind)	✅ PASS	rewind matches	-
yarc (rewind)	✅ PASS	rewind matches	-

Platform: Linux x86_64

_{Updated by CI at 2026-10-18T00:53:17.747Z}

[github-actions]

Regression: macos-arm64

Regression Test Results

ROM	Status	Details	Diff
jagniccc	✅ PASS	0 pixels differ	-
yarc	✅ PASS	0 pixels differ	-
jagniccc (determinism)	✅ PASS	identical across runs	-
yarc (determinism)	✅ PASS	identical across runs	-
jagniccc (frameskip)	✅ PASS	skip=0 matches skip=3	-
yarc (frameskip)	✅ PASS	skip=0 matches skip=3	-
jagniccc (save state)	✅ PASS	round-trip matches	-
yarc (save state)	✅ PASS	round-trip matches	-
jagniccc (rewind)	✅ PASS	rewind matches	-
yarc (rewind)	✅ PASS	rewind matches	-

Platform: Darwin arm64

_{Updated by CI at 2026-10-18T00:44:12.147Z}

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[github-actions]

App packages

Built from 1108a36 (workflow run).

Platform	Download	Size
Linux x86_64: Flatpak bundle (.flatpak)	app-package-flatpak	19.1 MB
Linux aarch64: .AppImage + .deb	app-package-linux-aarch64	44.2 MB
Linux x86_64: .AppImage + .deb	app-package-linux-x86_64	45.6 MB
macOS (universal): .zip	app-package-macos-universal	30.0 MB
Windows ARM64: portable .zip + -setup.exe	app-package-windows-arm64	41.5 MB
Windows x64: portable .zip + -setup.exe	app-package-windows-x64	45.1 MB

Downloading an artifact needs a GitHub login. macOS: Developer ID signed and notarized (PRs from forks get
no signing secrets and stay ad-hoc signed: right-click > Open, or xattr -dr com.apple.quarantine "Virtual Jaguar.app").
Windows: not code-signed yet, SmartScreen may warn. Linux AppImage: chmod +x it.

_{Updated by CI at 2026-10-18T03:54:36.836Z}