[MarcZX10]

What

Adds an execution regression for #5693's repeated unwatched storage-image case. It stores an image, samples the live image into a separate surface, marks the original image dirty again without re-uploading it, then changes guest bytes while the repeated result is pending. The next guest read must preserve the CPU-written bytes.

This case failed at the existing "guest bytes not saved" guard before #5679. That fix is now on main; the case passes there and guards the behavior. It covers the sample-then-redirty path, distinct from #5760's direct second-store kernel case. The overlap bot also flags #5761, an explicitly test-only combined integration build that is not intended for merging; this PR contains only this regression case.

Tested

• Rebased on current main (145cb82f3b54d8360e728c3cd7ed82a6fa2eee1a), including fix(agc): save the guest bytes of units without a write generation when results land outside the write watch #5679 and fix(agc): keep the saved guest bytes of units a fill or image copy does not cover #5707.
• Debian 12, GCC 12.2, Ninja; Mesa llvmpipe (LLVM 15.0.6), Mesa Vulkan driver 22.3.6: the focused CTest passes (1/1). This host lacks the PS5-required LZCNT/BMI1/BMI2/MOVBE/AVX2/FMA CPU features, so I temporarily bypassed only the emulator startup CPU gate in the local test build; that source edit was restored and is not in this PR.
• tools/check_conventions.py --base upstream/main --head HEAD and git diff --check pass.
• No title run; this is a test-only follow-up.

Checklist

• Based on current main; no other open PR adds this sample-then-redirty regression case
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository
• New third-party code is a submodule built from source
• Generic behavior, not specific to one title
• Depends on: test(agc): keep untracked store surfaces alive through queued work #5790 (shared fixture lifetime fix)
• Unblocks: main: PPSA01746, PPSA01530 and PPSA03001 still stop at the storage-image write-back throw after #5657 #5693
• GPU tested: Mesa software Vulkan, llvmpipe (LLVM 15.0.6), Mesa 22.3.6, Debian 12
• AI-assisted: yes

[MarcZX10]

The overlap is with #5761, whose description explicitly marks it as a combined test build, not for merging. This PR adds only the sample-then-redirty regression; it is distinct from the direct second-store case in #5760. I updated the body to record that distinction and the dependency on #5790 for safe fixture cleanup.

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 5761	core/libs/prx/libSceAgcDriver/tests/execution/StoreOverUntracked.cpp
PR 6188	core/libs/prx/libSceAgcDriver/tests/execution/StoreOverUntracked.cpp

[adriadam10]

Checked the new UntrackedRepeat case: it builds and passes on Linux (agc_driver_store_over_untracked, repeated runs).

Added one commit on your branch: output was a Surface local to the if block, so its memory was freed while ReadImage could still be queued. It is now an std::optional<Surface> declared before waitForWork, so it outlives the queued dispatch (same lifetime issue as #5790).

Merging once the full Build is green.

[adriadam10]

Rebased onto current main: the conflict was with the new NeighbourLeftWatch case in StoreOverUntracked.cpp; both cases are kept and the cases array is now sized 10. agc_driver_store_over_untracked passes on Linux (repeated runs). The earlier commit that keeps the sampled output surface alive until the queued read finishes stays.

Merging once the full Build is green.