[Tenith01]

Summary

Resolves the process termination occurring in video_out_flip_tests on the default lifetime mode and registers the headless execution tests in CMake.

Changes

• Wrapped AgcDriverShutdown_nid_postfix() in a try/catch within VideoOutDriver::presentLoop() to catch the shutdown error without crashing the thread, properly recording it to flipQueue->failure.
• Removed EXCLUDE_FROM_ALL on video_out_flip_tests when BUILD_TESTING is active so it's compiled with ctest.
• Registered add_test instances for lifetime, decode, and controls modes with matching headless properties.

Testing

• Modified logic validated against the original issue's propagation path descriptions to verify exception catching on thread shutdown and proper rethrow.
• Checked CMake CI definitions to assure alignment with guest_video_out_wqhd_detection.

Related Issue

Fixes #1040

Note: This PR was AI-assisted and modifies video_out_flip_tests for safe headless modes without a fixture. The reopen behavior discussed in the issue is left unchanged and can be decided by maintainers as a follow-up.

[oneandonlydean]

Have you run the three new ctest entries? The description says the change was checked against the issue text only, and video_out_flip_lifetime needs SDL and a Vulkan device (#1040 used SDL_VIDEODRIVER=offscreen and lavapipe). Please add the ctest output and fill in the PR template.

[MarcZX10]

Independent Linux validation of the current head 84be0c860ee9d0c5fd367e663815b23c8550eca6. Built video_out_flip_tests from a fresh Release build with BUILD_TESTING=ON (GCC 12.2, Debian x86_64), then ran the three registered cases under SDL offscreen with lavapipe and the CI Vulkan flags: video_out_flip_lifetime, video_out_flip_decode, and video_out_flip_controls all passed (3/3, 0.86 s total).

This validates the three Linux CTest entries; Windows was not tested. The PR description still needs the repository What / Tested / Checklist template requested in review.

[MarcZX10]

While checking the requested PR template, I found a duplicate to resolve first: open PR #1575 also wraps AgcDriverShutdown_nid_postfix() in presentLoop() and registers the same three headless flip tests with the same CTest names. Its CMake/test-registration changes overlap this PR's scope; #1575 also removes video_out_flip_tests from tools/check_test_registration.py's known-unrun list. CONTRIBUTING says to avoid a second PR implementing the same functions; please coordinate with #1575's author/maintainers on which PR should carry the fix, then update this PR's body and checks for the chosen scope.

For reference, I independently built this head (84be0c86) and got 3/3 CTest passes on Linux with SDL offscreen and lavapipe; Windows remains untested. The PR currently has no upstream check results.

[oneandonlydean]

Closing: #1225 makes the same presentLoop change and test registration for #1040, with Windows runs. If it stalls, please reopen.