CNA-VGAP-025: WEBGPU and SDL_GPU run in CI only through subsystem suites and a macOS smoke set; their renderer suites run only on CNA's own runner

CNA snapshot 3b43bd86  ·  Known Issues › Verification gaps  ·  source links pinned to 3b43bd86

CI now selects both identities, but only for the glTF-conformance, input and platform suites (on lavapipe) and, for SDL_GPU, a focused macOS smoke set; the renderers' own SdlGpu_* and WebGPU_* suites run only in CNA's recorded local runs.

Identifier

CNA-VGAP-025

Category

Verification gap

Subsystem

Testing & evidence

Status

Narrowed (partially fixed; describes only what survives)

Verified against

CNA 3b43bd86 (3b43bd8687ed46523d873f9bec4b32bd8481dcb5)

Evidence basis

Source-verified: read at TARGET, not executed

Tests touching this area

Yes: see Current tests

Affected contract

WEBGPU and SDL_GPU renderer identities and their example test suites

Expected behaviour

Published renderer identities should have their own test suites exercised by at least one automated lane on a representative host.

Actual behaviour at TARGET

At 3b43bd86 gltf-renderer-stride-ci.yml configures SDL_GPU and WEBGPU on lavapipe and runs ctest -L gltf-conformance; input-ci.yml configures SDL_GPU and runs -L input; platform-ci.yml configures both and runs the CnaPlatform* suites; apple-ci.yml builds SDL_GPU on macOS and runs the focused smoke set of run-sdlgpu-smoke.sh. None of these runs the renderers' own example suites (the SdlGpu_* and WebGPU_* CTests, including the parity fixtures registered through cna_register_parity_fixtures). Those suites run in CNA's recorded full-tree runs on its own runner (handoff.md: cmake-build-sdlgpu 11,284 tests and cmake-build-webgpu 10,964 tests on 2026-10-17, RADV), and WEBGPU's browser runs in scripted headless Chrome. The browser renderers (CNA-VGAP-023), the Windows identities (CNA-VGAP-026) and the recorded glTF golden sets (CNA-VGAP-030) have entries of their own; whether SDL_GPU is qualified on the hosts where it is the default is a separate entry.

Source locations

• .github/workflows/gltf-renderer-stride-ci.yml — SDL_GPU and WEBGPU lavapipe cells: ctest -L gltf-conformance only
• .github/workflows/input-ci.yml — SDL_GPU cell: ctest -L input
• .github/workflows/platform-ci.yml — SDL_GPU and WEBGPU cells: CnaPlatform* suites
• .github/workflows/apple-ci.yml — macos-sdlgpu: the focused smoke set
• scripts/ci/run-sdlgpu-smoke.sh — the smoke set: device, 2D drawing and a few conformance programs
• modules/renderers/sdl-gpu/examples/CMakeLists.txt — SDL_GPU registrations
• modules/renderers/webgpu/examples/CMakeLists.txt — WEBGPU registrations
• handoff.md — Tests: the 2026-10-17 full-tree runs on the private runner

Evidence

Checked by reading every workflow under .github/workflows, the smoke script and handoff.md at 3b43bd86; nothing was built or executed for this entry.

Narrowed at c6dec91f: the renderer retirement moved the lavapipe cells of the glTF-conformance and input workflows to SDL_GPU (fe5e25994, 2026-10-13), d97eff100 (RRC-051, 2026-10-14) added the gating macOS SDL_GPU smoke set, and f2f5a04aa (2026-10-17) added WEBGPU to the glTF-conformance and platform cells, so the entry no longer says that no workflow selects either identity. What survives is that their own renderer suites are not part of any workflow.

Independently reported as CNA-VGAP-048 (folded into this entry by the adversarial audit): Only the OPENGLES3 whole-corpus L7 rung runs in CI; the parity fixtures that the WEBGPU and SDL_GPU example trees register are run by no workflow, and FNA3D and DIRECTX9 are configured by no automatic workflow, so their registered parity fixtures are local evidence.

Focused reproduction

No focused reproduction is known. Nothing has been invented here; the evidence above is what exists.

Current tests

Large local suites exist for both renderers (SdlGpu_*, WebGPU_*); CI runs the glTF-conformance, input and platform labels on them and the SDL_GPU macOS smoke set.

Regression test

A Linux CI job per identity that runs its example suite (lavapipe already serves both in the existing cells).

Blast radius

Renderer-specific WEBGPU and SDL_GPU behaviour outside the glTF-conformance, input and platform suites; regressions there surface only in CNA's own full-tree runs.

Workaround

No workaround is known.