[xXPBSXx]

What

The nightly sanitizer run 41573880531 (job address-undefined, head f6232fc) stopped agc_driver_recorder at its limit: Timeout 30.25 sec. The run of 2026-11-02 on 926f7dd passed it in 6.88 s.

agc_driver_recorder has TIMEOUT 30. Under the sanitizer build the test took about three times as long on d1ae628bc2 as on 926f7dd in the measurement below, which leaves it close to the limit. In a build without the sanitizer it stays at 30 s. This keeps TIMEOUT 30 and sets 120 only when ANYPS5_ENABLE_ASAN is on, the way agc_performance_overlay is set in agc_pad_status_overlay_tests.cmake L86-L88. Not bisected; which change slowed the test under the sanitizer is not known, and a higher limit in that build can also hide a later slowdown.

Overlap

All 1678 open PRs were listed on 2026-11-03 at 10:21 UTC and their changed files read until 11:16 UTC, against main 4f41f4c375; the merge tests below ran with this change applied on that main. The file of this change is core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_recorder_tests.cmake. Twenty-seven other open PRs change it; each was merged with and without this change (git merge-tree):

result	PRs
clean with and without	#14330, #13942, #13263, #13251, #13044, #12921, #10443, #10159, #9884, #9875
conflict with and without (already conflicts with main)	#15169, #15136, #14262, #14261, #13929, #13928, #13920, #13863, #13809, #13793, #10193, #9898, #9508, #9289, #7642, #7085, #6186

No PR differs between the two columns, so this change creates no conflict.

Tested

Ubuntu 24.04.5 on WSL2 (kernel 6.18.54.1), gcc 13.3.0, Release, 2026-11-03 10:41 to 11:08 UTC, on main 4f41f4c375 with this change. The effective TIMEOUT of agc_driver_recorder from ctest --show-only=json-v1:

• -DANYPS5_ENABLE_ASAN=ON (the flags of the nightly sanitizer workflow, -g1 -fsanitize=undefined -fno-sanitize=vptr): 120. agc_performance_overlay there: 120.
• without the sanitizer: 30. agc_performance_overlay there: 30.
  Runtime of ctest -R '^agc_driver_recorder$', one test at a time, three runs:
• without the sanitizer: 7.72 s, 4.34 s, 3.89 s.
• with the sanitizer: 26.87 s, 27.53 s, 70.33 s (all passed). These runs were not under a controlled load: a pull request file sweep and a title run ran on the same machine, and the 70 s run in particular is not a measurement of the test.
  Earlier sanitizer runs on the commits before (07:31 to 07:34 UTC, a pull request file sweep on the same machine): 926f7dd took 9.52 s, 6.72 s and 6.99 s; d1ae628bc2 took 23.07 s, 24.44 s and 24.09 s. The timeout of the nightly run was not reproduced, and this change was not run under a controlled load.

Checklist

• Based on current main; 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 (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 (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

AI-assisted (Claude Code).

🤖 Generated with Claude Code

[MarcZX10]

Reviewed — changes requested

The new 120-second CTest limit conflicts with the repository's explicit 30-second per-test rule, which says to split or shrink tests that approach the limit instead of raising their timeout.

What I ran

Reviewed the only changed file at head 96ece96e85f509e8aa979a909022357b99236cac against base 4f41f4c375f5abb487dbd184fbc150f31994d762; the exact-head merge preview is clean. Environment: Debian 12, Linux x86_64, Python 3.11.2, Git 2.39.5; I did not compile or build. I inspected the timeout change, the repository's timing rule, the PR's reported test timings and current hosted check state.

git diff --check 4f41f4c375f5abb487dbd184fbc150f31994d762...96ece96e85f509e8aa979a909022357b99236cac
python3 tools/pr_precheck.py --repo boykopovar/AnyPS5 --pr 15441
python3 tools/pr_test_quality.py --root /tmp/anyps5-15441-quality.HbZ5PO --json core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_recorder_tests.cmake
gh pr checks 15441 --repo boykopovar/AnyPS5

The PR description reports three sanitizer runs on base d1ae628bc2 at 23.07, 24.44 and 24.09 seconds and a prior nightly run that timed out at 30.25 seconds. Those are author-reported results, not runs I performed.

Results

Check	Result	Note
Diff, conventions, and merge preview	passed	No formatting issue; exact-base preview is clean.
Test-quality heuristic	passed	No heuristic findings for the CMake change.
Repository timing rule	failed	The change raises the test limit to 120 seconds despite the 30-second policy.
Hosted policy checks	passed	Conventions, intake, labels, eligibility, portability, and report passed.
Hosted Windows job	not run	In progress at the final status refresh; I did not run it.
Hosted Linux and specialized jobs	not applicable	GitHub skipped these jobs.
Local build and CTest	not run	No build or test run was needed to inspect this timeout-only edit.

Commands and unedited outputs

Precheck table:

Check	Status	Command / reason
build_and_compile	not_applicable	The precheck deliberately does not configure, build, or compile AnyPS5.
diff_check	passed	git diff --check 4f41f4c375f5abb487dbd184fbc150f31994d762...96ece96e85f509e8aa979a909022357b99236cac
conventions_and_pr_template	passed	/usr/bin/python3 tools/check_conventions.py --base 4f41f4c375f5abb487dbd184fbc150f31994d762 --head 96ece96e85f509e8aa979a909022357b99236cac --title test(agc): give the driver recorder test the same timeout as its slower siblings --body /tmp/anyps5-pr-precheck/boykopovar-AnyPS5-pr-15441-96ece96e85f5/pr-body.md
changed_export_abi_audit	not_applicable	No changed source or public header under core/libs/prx.
python_syntax	not_applicable	No changed Python implementation files.
related_python_unit_tests	not_applicable	No changed tools/ module has an existing sibling tools/tests/test_<module>.py suite.
python_test_registration	not_applicable	No changed test_*.py files.
live_dependency_and_mergeability_check	passed	/usr/bin/python3 /marc/AnyPS5/tools/pr_stack_check.py 15441 --repo boykopovar/AnyPS5 --output-dir /tmp/anyps5-pr-precheck/boykopovar-AnyPS5-pr-15441-96ece96e85f5/stack
build_dependent_cpp_tests	not_applicable	No C/C++ implementation or test files changed.
test_quality_heuristics	failed	/usr/bin/python3 /tmp/anyps5-pr-precheck/boykopovar-AnyPS5-pr-15441-96ece96e85f5/head-96ece96e85f5-dd7c0f6a/tools/pr_test_quality.py --root /tmp/anyps5-pr-precheck/boykopovar-AnyPS5-pr-15441-96ece96e85f5/head-96ece96e85f5-dd7c0f6a --json core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_recorder_tests.cmake
negative_controls	not_applicable	No changed Python test module in tools/ can be run as a no-build base-source control.

The precheck heuristic command could not start because its isolated PR snapshot lacks the workspace-only script. The direct helper run against the exact changed file returned:

{
  "findings": [],
  "negative_control_candidates": [],
  "limits": [
    "Heuristics can miss weak assertions and can flag valid tests; inspect every finding.",
    "A test that passes against base is not by itself proof that the PR test is inadequate.",
    "Negative controls are only candidates until the test actually runs against base and fails by assertion rather than import/setup error."
  ]
}

The exact-head stack report says the merge preview is clean and no dependency is declared. Hosted status at review: Windows was in progress; conventions, intake, labels, eligibility, library-portability and report passed; Linux and specialized jobs were skipped.

Negative control

Assertion	Change to code under test	Did the new test fail?
agc_driver_recorder should complete within the repository's 30-second CI budget.	No mutation or local execution; I reviewed the PR's reported runs and the timeout declaration.	Not run by me; the PR reports one nightly CI timeout at 30.25 seconds.

Findings

Blocking

The test's timeout is raised from 30 to 120 seconds at the reviewed head. Repository policy says each test finishes within 30 seconds on CI and explicitly directs contributors to split or shrink a test that gets close, rather than raising its timeout. The PR reports a 30.25-second timeout and recent runs around 24 seconds, so the reported evidence shows the test is already at or over the project's budget. Please reduce or split the test so it stays within 30 seconds under the normal CI load instead of extending its timeout.

Non-blocking

None.

Sources

• Repository's 30-second test budget and direction to split or shrink slow tests
• The 120-second timeout at the reviewed head

Not verified

I did not run the test under ASan, reproduce the nightly timeout, build, or run CTest locally. The three recent timings and the nightly timeout are reported by the PR author.

To get this to a good state

Reduce or split the test to meet the repository's 30-second budget, then rerun the focused sanitizer test under the relevant CI conditions. If the head changes, I will rerun the diff, conventions, precheck and merge preview on the new head.

[XairRrR]

Reviewed — changes requested

The timeout only happens under the sanitizer, but the new 120 s limit applies to every build, including normal CI, where the test takes about 6 s. The repository already has a way to raise a limit only in ASan builds, and this change should use it.

What I ran

Head 96ece96e85f509e8aa979a909022357b99236cac, main 4f41f4c375f5abb487dbd184fbc150f31994d762, merged tree 2d43fd7829323171be33a3fa74929222f0a23b00. Machine: Debian 13 VM with 32 vCPU (Xeon Gold 6136), GCC 14.2.0, Release, lavapipe (Mesa 26.1.6). No sanitizer build.

git diff --check 4f41f4c375f5abb487dbd184fbc150f31994d762...96ece96e85f509e8aa979a909022357b99236cac
python3 tools/check_conventions.py --base 4f41f4c375f5abb487dbd184fbc150f31994d762
ctest --test-dir build -R '^agc_driver_recorder$' --output-on-failure   # three times, one after the other

The script also built the merged tree and ran the selected tests and the Khronos validation comparison.

Results

Check	Result
Merge into main	clean
git diff --check	passed
check_conventions.py	passed
Build and libs	passed
Selected tests (1403)	1377 passed, 24 expected skips, 2 failures that also fail on main (unwrap_self_..._destination_directory_group, vector_constant_factor_tininess); no new failures
agc_driver_recorder, Release, 3 runs	passed in 5.63 s, 8.05 s and 5.58 s
Validation layer	loaded in 209 of 210 tests and missing in none (the script's list of tests without the layer is empty). Its one "new" VUID (VUID-vkCmdDispatch-groupCountX-00386) comes from agc_driver_native_depth_uniform_planes, which did not run on the baseline. This PR only changes a CMake timeout in another fragment, so the VUID is not from this PR.

ctest output

1/1 Test #1971: agc_driver_recorder ..............   Passed    5.63 sec
1/1 Test #1971: agc_driver_recorder ..............   Passed    8.05 sec
1/1 Test #1971: agc_driver_recorder ..............   Passed    5.58 sec

Negative control

None. The PR changes only a test timeout and no production code or test logic, so there is nothing to revert.

Findings

Blocking

The limit is raised for every build, but the timeout only happens under the sanitizer (core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_recorder_tests.cmake:380 in the merged tree; L354 at head).

• Evidence that the problem is sanitizer-only: the PR's own timings and the failing nightly run all come from the ASan+UBSan build. In a Release build here the test takes 5.6 to 8 s, well below 30 s.
• Why this conflicts with the rules: CONTRIBUTING.md L19 says to split or shrink a test that gets close to 30 s instead of raising its timeout. With an unconditional 120 s limit, the normal Linux and Windows CI jobs would no longer stop this test if it slowed down about fifteenfold.
• Where I differ from MarcZX10's review: I agree the change breaks that rule. I don't think the test needs splitting, though, because it is nowhere near 30 s outside the sanitizer.
• Existing precedent: the repository already solves this exact case in agc_pad_status_overlay_tests.cmake L86-L88: agc_performance_overlay keeps TIMEOUT 30 and gets 120 s only inside if(ANYPS5_ENABLE_ASAN). That test also has an entry in .github/sanitizer-known-failures.txt, line 4.
• Suggested fix: keep TIMEOUT 30 on the existing line and add the same ASan-only override:

if(ANYPS5_ENABLE_ASAN)
    set_tests_properties(agc_driver_recorder PROPERTIES TIMEOUT 120)
endif()

Non-blocking

• The slowdown is not explained. By the description's own numbers, the sanitizer runtime roughly tripled between 926f7dd492 (about 7 s) and d1ae628bc2 (about 24 s), and it was not bisected. A higher limit hides that regression rather than explaining it. Please bisect it, or open an issue for it, so the slowdown is followed up.
• The title says "siblings" (plural), but only one sibling has 120 s. In this fragment, only agc_storage_write_back uses 120 s; the other larger limits are 60 s. This is minor: please reword the title or description.

Not verified

• I did not build with -DANYPS5_ENABLE_ASAN=ON or UBSan, so I did not reproduce the ~24 s sanitizer timings or the 30.25 s nightly timeout. Those are the author's numbers.
• I did not check whether the Release runtime also grew between 926f7dd492 and main.
• I did not check the overlap table against other PRs.

To get this to a good state

1. Restore TIMEOUT 30 and raise the limit only inside if(ANYPS5_ENABLE_ASAN), as agc_performance_overlay does.
2. Retitle the PR, or reword the description, to say the change is for sanitizer builds.
3. Optionally, bisect the sanitizer slowdown between 926f7dd492 and d1ae628bc2, or record it somewhere it will be followed up.

[TechLevelZero]

✅ AMD GPU test: passed

PR head 96ece96 merged onto main 4f41f4c · tested 2026-11-03 10:29 UTC · 13 min

GPU	AMD Radeon RX 6300 (RADV NAVI24), driver Mesa 25.2.8-0ubuntu0.24.04.4, Vulkan 1.4.318
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-42-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	✅ 1,709 passed · 7 failed (0 new, 5 also fail on main, 2 unstable on main) · 7 skipped (ANYPS5_REQUIRE_VULKAN=1)
Test setup	CPU-only tests without a GPU; GPU tests (and any CPU-only test that failed or skipped without one) once on the GPU with VK_LAYER_KHRONOS_validation and ANYPS5_NO_SHADER_CACHE=1; main ran the same way
Test selection	1,723 of 3,811 tests run: the ones this change can reach, chosen by tools/select_tests.py from main's build graph
Vulkan validation	✅ no new validation IDs vs main (523 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.2 s, ⚠️ public_headers 81.2 s, ⚠️ agc_driver_image_sample_unnormalized 36.3 s

Unstable on main (2, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_pipeline_journal: failed in 7 of the last 9 main runs
agc_driver_store_over_untracked: failed in 1 of the last 9 main runs

Failing on main too (5, not counted against this PR)

agc_driver_image_table
agc_driver_native_depth_bda_clear
agc_driver_native_multisample_depth
guest_mmap_executable
guest_socket_options_multicast

Validation IDs already present on main (4 IDs in 446 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (446 tests): agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, agc_driver_bindless_images, and 438 more
VUID-vkCmdCopyImage-srcImage-01548 (8 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback, agc_driver_stencil_plane_memory
VUID-vkCmdCopyImage-srcImage-01551 (8 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback, agc_driver_stencil_plane_memory
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_srgb8_color_target

Skipped (7)

agc_driver_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_linear_target_direct_write_completion: skipped, the device does not keep host imports write-watched, so no completion stamp over the target is tracked
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible

Reproduce

git fetch origin 96ece96e85f509e8aa979a909022357b99236cac 4f41f4c375f5abb487dbd184fbc150f31994d762
git checkout 4f41f4c375f5abb487dbd184fbc150f31994d762 && git merge --no-ff 96ece96e85f509e8aa979a909022357b99236cac
git submodule update --init --recursive
cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON -DANYPS5_ENABLE_SPIRV_TOOLS=ON
cmake --build build && cmake --build build --target libs
VK_DRIVER_FILES=/nonexistent/icd.json ctest --test-dir build -E '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 12 --output-on-failure
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 16 --output-on-failure
# failures are then re-run alone: --timeout 360 --parallel 2; new validation IDs: the test alone, twice, --parallel 1

_{Automated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}

[xXPBSXx]

Thanks to both reviewers. The new head 95f01f3 keeps TIMEOUT 30 and raises the limit to 120 only inside if(ANYPS5_ENABLE_ASAN), as agc_performance_overlay does; the title and description say so. The effective timeout was read from ctest --show-only=json-v1 in a build with -DANYPS5_ENABLE_ASAN=ON (120) and one without (30). The slowdown under the sanitizer is still not bisected.

AI-assisted (Claude Code).

[XairRrR]

Re-checked — approved

The new commit 95f01f3601 keeps TIMEOUT 30 for agc_driver_recorder and raises it to 120 s only when ANYPS5_ENABLE_ASAN is on, as I suggested. That resolves my blocking finding.

What changed since my last review

• New commit 95f01f3601 ("raise the driver recorder test limit only in ASan builds"). The first commit, 96ece96e85, is unchanged. Together the PR now leaves the existing TIMEOUT 30 line alone and adds an if(ANYPS5_ENABLE_ASAN) block that sets TIMEOUT 120.
• The title and description were rewritten to say the change is for sanitizer builds. The author's 12:21 comment reports the effective timeout from ctest --show-only=json-v1 with ASan (120) and without it (30), and says the slowdown is still not bisected.

What I ran

Head 95f01f3601506c913ddb9ea37229a237faf0a111, main 4f41f4c375f5abb487dbd184fbc150f31994d762, merged tree 29b759b7f3f11487608ff73660459afb5b4ab983. Machine: Debian 13 VM, 32 vCPU (Xeon Gold 6136), GCC 14.2.0, Release, lavapipe (Mesa 26.1.6). This build has ANYPS5_ENABLE_ASAN=OFF.

git diff --check 4f41f4c375f5abb487dbd184fbc150f31994d762...95f01f3601506c913ddb9ea37229a237faf0a111
python3 tools/check_conventions.py --base 4f41f4c375f5abb487dbd184fbc150f31994d762
ctest --test-dir build -R '^agc_driver_recorder$' --show-only=json-v1   # TIMEOUT property
grep -n ANYPS5_ENABLE_ASAN build/CMakeCache.txt
grep -rn 'ANYPS5_ENABLE_ASAN\|fsanitize' .github/workflows/*.yml
git status --porcelain

The script also built the merged tree, ran the selected tests and ran the Khronos validation comparison.

Check	Result
Merge into main	clean
git diff --check	passed
check_conventions.py	passed ("All conventions checks passed.")
Build and libs	passed
Selected tests (1382)	1357 passed and 23 were expected skips. Two tests failed, and both also fail on main (unwrap_self_..._destination_directory_group, vector_constant_factor_tininess). No new failures and no timeouts.
Validation layer	loaded in 207 of 208 tests and missing in none; 0 tests with new VUIDs
Effective TIMEOUT of agc_driver_recorder, non-ASan build	30.0
Sanitizer workflow sets the guard	yes: .github/workflows/sanitizers.yml:20 passes -DANYPS5_ENABLE_ASAN=ON

Previous findings

• Blocking: the limit was raised for every build. Resolved. At core/libs/prx/libSceAgcDriver/tests/cmake/agc_driver_recorder_tests.cmake:380-383 in the merged tree, the line TIMEOUT 30 is unchanged and the override to 120 sits only inside if(ANYPS5_ENABLE_ASAN). This is the same pattern as agc_performance_overlay in agc_pad_status_overlay_tests.cmake L86-L88. In this non-ASan build, ctest reports TIMEOUT 30.0. The value of 120 under ASan is the author's report; I did not configure an ASan build.
• Non-blocking: the sanitizer slowdown is not explained. Still open. The description and the author's comment say plainly that it was not bisected. That is acceptable for a timeout change, but it would still be good to follow up on it.
• Non-blocking: the title said "siblings" (plural). Resolved. The PR title and the new commit now describe an ASan-only limit. The first commit's subject still says "siblings", which only matters if the PR is merged without squashing.

Negative control

None. The PR changes only a CTest property and no production code or test logic, so there is nothing to revert.

New findings

Blocking

None.

Non-blocking

• agc_performance_overlay has an entry in .github/sanitizer-known-failures.txt that explains its ASan timeout. agc_driver_recorder has none, and with the 120 s limit it does not need one. No action is needed.

Not verified

• I did not configure or run an ASan/UBSan build, so I checked neither the effective 120 s limit under ASan nor the sanitizer runtimes of about 24 s. Both are author-reported.
• I did not reproduce the 30.25 s timeout of the nightly run, and I did not bisect the slowdown.
• I did not check the overlap table.

[TechLevelZero]

✅ NVIDIA GPU test: passed

PR head 95f01f3 merged onto main 4f41f4c · tested 2026-11-03 12:27 UTC · 16 min

GPU	NVIDIA RTX 2000 Ada Generation, driver 595.99.02, Vulkan 1.4.329
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-38-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	✅ 1,715 passed · 4 failed (0 new, 3 also fail on main, 1 unstable on main) · 4 skipped (ANYPS5_REQUIRE_VULKAN=1)
Test setup	CPU-only tests without a GPU; GPU tests (and any CPU-only test that failed or skipped without one) once on the GPU with VK_LAYER_KHRONOS_validation and ANYPS5_NO_SHADER_CACHE=1; main ran the same way
Test selection	1,723 of 3,811 tests run: the ones this change can reach, chosen by tools/select_tests.py from main's build graph
Vulkan validation	✅ no new validation IDs vs main (523 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.2 s, ⚠️ agc_driver_lookup_log_fits_build 30.6 s

Unstable on main (1, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_pipeline_journal: failed in 4 of the last 10 main runs

Failing on main too (3, not counted against this PR)

agc_driver_image_table
guest_mmap_executable
guest_socket_options_multicast

Validation IDs already present on main (4 IDs in 446 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (446 tests): agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, agc_driver_bindless_images, and 438 more
VUID-vkCmdCopyImage-srcImage-01548 (9 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback, and 1 more
VUID-vkCmdCopyImage-srcImage-01551 (9 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_plane_sampling, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback, and 1 more
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_srgb8_color_target

Skipped (4)

agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_stencil_ref_export: skipped, the device lacks VK_EXT_shader_stencil_export

Reproduce

git fetch origin 95f01f3601506c913ddb9ea37229a237faf0a111 4f41f4c375f5abb487dbd184fbc150f31994d762
git checkout 4f41f4c375f5abb487dbd184fbc150f31994d762 && git merge --no-ff 95f01f3601506c913ddb9ea37229a237faf0a111
git submodule update --init --recursive
cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON -DANYPS5_ENABLE_SPIRV_TOOLS=ON
cmake --build build && cmake --build build --target libs
VK_DRIVER_FILES=/nonexistent/icd.json ctest --test-dir build -E '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 32 --output-on-failure
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 4 --output-on-failure
# failures are then re-run alone: --timeout 360 --parallel 2; new validation IDs: the test alone, twice, --parallel 1

_{Automated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}

[TechLevelZero]

✅ AMD GPU test: passed

PR head 95f01f3 merged onto main 4f41f4c · tested 2026-11-03 12:42 UTC · 14 min

GPU	AMD Radeon RX 6300 (RADV NAVI24), driver Mesa 25.2.8-0ubuntu0.24.04.4, Vulkan 1.4.318
Host	Ubuntu 26.04.1 LTS, kernel 7.0.0-42-generic, container ubuntu:24.04
Logs	build and test logs for this run
Build	✅ passed (Release, -DANYPS5_ENABLE_SPIRV_TOOLS=ON)
Tests	✅ 1,709 passed · 7 failed (0 new, 5 also fail on main, 2 unstable on main) · 7 skipped (ANYPS5_REQUIRE_VULKAN=1)
Test setup	CPU-only tests without a GPU; GPU tests (and any CPU-only test that failed or skipped without one) once on the GPU with VK_LAYER_KHRONOS_validation and ANYPS5_NO_SHADER_CACHE=1; main ran the same way
Test selection	1,723 of 3,811 tests run: the ones this change can reach, chosen by tools/select_tests.py from main's build graph
Vulkan validation	✅ no new validation IDs vs main (523 tests run on the GPU)
Slow tests (>30 s, GPU tests timed with validation layers)	⚠️ agc_driver_pipeline_journal 120.2 s, ⚠️ public_headers 57.4 s, ⚠️ agc_driver_lazy_resolved_abi 44.9 s, ⚠️ agc_driver_image_sample_unnormalized 38.5 s

Unstable on main (2, these also failed in recent main runs on this machine; not counted against this PR)

agc_driver_pipeline_journal: failed in 7 of the last 9 main runs
agc_driver_store_over_untracked: failed in 1 of the last 9 main runs

Failing on main too (5, not counted against this PR)

agc_driver_image_table
agc_driver_native_depth_bda_clear
agc_driver_native_multisample_depth
guest_mmap_executable
guest_socket_options_multicast

Passed on retry (1, failed in the parallel run, passed when re-run with a 360 s timeout; not counted against this PR)

vulkan_validation_gate_attribution: Failed in the parallel run; also passed only on retry in 13 other runs in the last 7 days

Validation IDs already present on main (4 IDs in 446 tests)

VUID-VkDeviceCreateInfo-pNext-pNext (446 tests): agc_driver_address_passes, agc_driver_alias_copy_label, agc_driver_alpha_to_coverage, agc_driver_array_layer_target, agc_driver_bc_block_copy, agc_driver_bda_anchor_reads, agc_driver_bda_span_reads, agc_driver_bindless_images, and 438 more
VUID-vkCmdCopyImage-srcImage-01548 (8 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback, agc_driver_stencil_plane_memory
VUID-vkCmdCopyImage-srcImage-01551 (8 tests): agc_driver_depth_clear_hle, agc_driver_depth_plane_buffer_memory, agc_driver_depth_plane_partial_store, agc_driver_depth_storage_dispatch, agc_driver_depth_surface_from_storage, agc_driver_dispatch_move_depth, agc_driver_stencil_null_feedback, agc_driver_stencil_plane_memory
VUID-VkPipelineColorBlendStateCreateInfo-pAttachments-00605 (1 test): agc_driver_srgb8_color_target

Skipped (7)

agc_driver_copied_writer_label_order: skipped, the device reports a minStorageBufferOffsetAlignment of 4, so the dispatch binds its record at +0x1004 in place and nothing is written back; the test needs an alignment above 4
agc_driver_float16_unused_nan_payloads: skipped, the device has no native f16 conversions with RTE rounding and kept f16 denormals
agc_driver_linear_target_direct_write_completion: skipped, the device does not keep host imports write-watched, so no completion stamp over the target is tracked
agc_driver_ngg_vertex_wave_reduction: skipped, s3 describes a 32-lane wave only on 32-lane subgroups, and the device has 64 lanes
agc_driver_scattered_runs: skipped, host imports are unavailable: /dev/udmabuf is not accessible
agc_driver_sparse_dma_buf_import: skipped, host imports are unavailable: /dev/udmabuf is not accessible

Reproduce

git fetch origin 95f01f3601506c913ddb9ea37229a237faf0a111 4f41f4c375f5abb487dbd184fbc150f31994d762
git checkout 4f41f4c375f5abb487dbd184fbc150f31994d762 && git merge --no-ff 95f01f3601506c913ddb9ea37229a237faf0a111
git submodule update --init --recursive
cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=ON -DANYPS5_ENABLE_SPIRV_TOOLS=ON
cmake --build build && cmake --build build --target libs
VK_DRIVER_FILES=/nonexistent/icd.json ctest --test-dir build -E '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 12 --output-on-failure
ANYPS5_REQUIRE_VULKAN=1 ANYPS5_NO_SHADER_CACHE=1 VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation \
  ctest --test-dir build -R '^(agc|shader|recompiler|videoout|video_out|guest_video_out|vulkan)|gpu' --timeout 120 --parallel 16 --output-on-failure
# failures are then re-run alone: --timeout 360 --parallel 2; new validation IDs: the test alone, twice, --parallel 1

_{Automated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs.}