CNA-BUG-229: docs/graphics-renderer-feature-matrix.md calls itself up to date but still says EasyGL's ReferenceStencil gap (Task 872) is open and lists its test as a known failure, contradicting its own table row

CNA snapshot 009d40f5  ·  Known Issues › Current bugs  ·  source links pinned to 009d40f5

The matrix's ReferenceStencil row says EasyGL passes (remeasured 2026-09-09), but its failure baseline and Vulkan bottom line still call the gap open, although REMED-GFX-236 fixed EasyGL on 2026-08-29.

Identifier

CNA-BUG-229

Category

Bug

Subsystem

Documentation & release tooling

Status

Narrowed (partially fixed; describes only what survives)

Verified against

CNA 009d40f5 (009d40f5dd085c4e674d3479675fac84b12b3e0a)

Severity

Low (a triage suggestion, not a project priority)

Evidence basis

Source-verified: read at TARGET, not executed

Tests touching this area

Yes: see Current tests

Affected contract

docs/graphics-renderer-feature-matrix.md ('Master, up-to-date cross-renderer feature matrix') and the documented status of GraphicsDevice::setReferenceStencilProperty on the EasyGL family

Expected behaviour

A matrix that labels itself "Master, up-to-date" should carry one status per feature and renderer. For GraphicsDevice.ReferenceStencil on EasyGL that status is fixed: REMED-GFX-236 (closed 2026-08-29 in plans/plan_postaudit.md) made EasyGLRenderer::SetReferenceStencil reissue the stencil function with the new reference, recording "1 FAIL - 2 PASS" for the test with a three-way mutation check, and rewrote the test's header accordingly.

Actual behaviour at TARGET

docs/graphics-renderer-feature-matrix.md says both things. Its GraphicsDevice.ReferenceStencil row: "shared stamp/test fixture passes on EasyGL (GraphicsDevice_ReferenceStencil, remeasured 2026-09-09)". Its "Known pre-existing test-failure baseline" still lists for EasyGL "3 - EasyGL_MRT_TwoAttachments, EasyGL_GraphicsDevice_ReferenceStencil, easy-gl-resource-smoke-tests", and its Vulkan bottom line says "the ReferenceStencil gap (Task 872) remains open only on EasyGL". The same stale state is recorded in plans/plan_graphics.md (Task 872 still open) and in the EasyGL registration comment, which still calls the test "a documented known failure".

Source locations

• docs/graphics-renderer-feature-matrix.md — header label; ReferenceStencil row; Known pre-existing test-failure baseline; Vulkan bottom line
• modules/renderers/easygl/src/EasyGLRenderer.cpp — EasyGLRenderer::SetReferenceStencil and the state ApplyDepthStencilState records for it
• modules/renderers/easygl/examples/easygl_graphicsdevice_reference_stencil_test.cpp — header: the fix history and the exit-code contract
• modules/renderers/easygl/examples/CMakeLists.txt — EasyGL_GraphicsDevice_ReferenceStencil registration comment ('documented known failure')
• plans/plan_graphics.md — Task 872 row, still open
• plans/plan_postaudit.md — REMED-GFX-236 record, closed 2026-08-29

Evidence

Checked by reading at 009d40f5; the test was not run here, so that it passes rests on CNA's own record (REMED-GFX-236 and the matrix row). Fixed since the matrix was first reported stale: the rows it then got wrong about Clear and stencil (Task 871) are corrected, FNA3D now appears (32-bit index audit), and the opening scope sentence limits the main tables to the established renderers. Whether EasyGL_MRT_TwoAttachments and easy-gl-resource-smoke-tests, the other two names in the EasyGL baseline, still fail at TARGET was not checked; the unfiltered CI job does not tolerate them.

Focused reproduction

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

Current tests

EasyGL_GraphicsDevice_ReferenceStencil (single-sided and two-sided legs), its Vulkan and Software counterparts, and GraphicsDevice_ReferenceStencil in the DirectX parity corpus. The test exists; the defect is in the records around it.

Regression test

No new test: correct the baseline and bottom-line sections, close Task 872, and update the registration comment. The CI allowlist that still tolerates this test failing is a separate entry.

Blast radius

Readers of the matrix and the graphics plan; it also keeps the stale CI allowlist entry looking justified. No behaviour is affected.

Workaround

Trust the table row and plans/plan_postaudit.md section 33 over the later sections.