[oneandonlydean]

What

guest_none_relocations times out intermittently on main's relinker (macos-latest, clang++, x86_64) job (the x86_64 build runs emulated on an arm runner). It is not tied to a commit: over the 10 main runs of 4a7770b to 66387a1 it took 16.3, 11.4, 20.04 (timeout), 19.99, 20.07 (timeout), 20.03 (timeout), 19.33, 19.99, 20.07 (timeout) and 15.4 s against its 20 s limit, while the native arm64 macOS job took 1.3 to 5.1 s. Locally the whole test takes 0.4 s, so the time on that job is process spawning: the test runs the relinker as a subprocess for every case, for sce False and then True.

The test now takes plain, sce or all (the default, so the tests that import its helpers and any direct run are unchanged) and CMake registers guest_none_relocations (plain) and guest_none_relocations_sce, each with the same 20 s limit and half the work. No case is removed or loosened.

Tested

• Linux workstation, current main 4f41f4c375 relinker: plain, sce and no argument each print "Guest no-op relocation tests passed"; an unknown mode raises KeyError.
• Not run: the CMake registration through ctest (the edit is two add_test lines and one set_tests_properties), the macOS jobs. Their timing after the split is the thing this PR needs CI to show.
• Not changed: the other timeout seen at 6e08dc3, select_tests_SelectArgumentTests (30.12 s on the arm job). It builds a real cmake project in setUpClass, took 3.4 to 18 s in the other runs, and printed nothing at the timeout; we treat it as runner noise and did not touch it.

Checklist

• Based on current main 4f41f4c375; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw; silent stubs are listed in technical debt (none added)
• No notes, investigation .md files or images added to the repository
• New third-party code is a submodule built from source (none added)
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none (intermittent red guest_none_relocations on main)
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[maskym]

Tested head acdbf77838 on Windows, merged alone into main 086c0686c8.

Host: Windows 11 26300, Intel i7-12700K, WinLibs GCC 15.2.0 (x86_64-ucrt-posix-seh, r6), Release, BUILD_TESTING=ON, Ninja. Reconfigured and built clean.

ctest -R '^(none_relocations|guest_none_relocations|guest_none_relocations_sce)$': 3/3 in each of 3 runs. Both new halves stay well under the 20 s limit (guest_none_relocations 4.4 s, guest_none_relocations_sce 4.5 s; none_relocations 3.3 s). Not run on Linux, macOS or a PS5.

AI-assisted: yes (Claude Code), the runs and the numbers above are from this machine.

[savvaniss]

#15751 goes after the same timeout from the other side. It converts the 28 valid modules of each SCE variant in one relinker run per output system, so the test starts the relinker 54 times instead of 216, and every check is kept. The 168 guest artifacts are byte-identical to the ones single-module conversions write.

The two changes compose: #15751 cuts the work, and your split halves what is left. They touch neighbouring lines of main(), so the second to land needs a small rebase. If this PR lands first, I will rebase #15751 onto it.