the nightly runs the big C++ tests - ctest -L big after the AOT sweep on every Release cell; the two style_lint fixture tests that were red on master lint their fixtures again and run per PR as small

[borisbat]

Why. Every per-PR and nightly ctest in CI is -L small, so the big-labelled C++ tests - a generated standalone context linked and run, the nano context, concurrent init, the C API split init - are built by the default target on every lane and executed on none. v0.6.4-RC2 shipped a -ctx whose generated code did not link and, once linked, never ran its initializers (#3967); tests-cpp/big/standalone_ctx shows both, and no lane ran it.

What changes.

• The nightly "Slow Release Tests" step runs ctest -L big -E memory_model_4gb after the AOT sweep, on every Release cell the step already covers - Linux, macOS and Windows, sanitizer cells included. No extra build: the binaries are already in the default target. memory_model_4gb allocates a real 4 GB chunk and stays local-only, as its CMakeLists says.
• Running the set showed the two style_lint fixture tests red on master: the lint CLI's content skip for an expect file fired before the --lint-fixtures override the comment beside it promised, so both fixtures were skipped and neither STYLE014 nor STYLE015 was ever printed. lint_file_skip_reason takes an expect_skips flag (default true, so the MCP lint tool is unchanged) that the CLI clears under --lint-fixtures; a // lint-skip-file directive keeps winning whatever the caller, which tests/lint/test_stale_nolint.das pins.
• Those two tests are relabelled small: each is one sub-second lint of one fixture, so every per-PR ctest runs them and the lint fix has a per-PR row.
• tests-cpp/big/REVIEW.md states the contract as it now is: per-PR CI runs -L small, the nightly runs -L big after the merge.

Observable behavior.

• A generated standalone context that does not link, or does not initialize, goes unnoticed until a release -> fails the next nightly on every platform.
• daslang utils/lint/main.das -- --lint-fixtures true <expect fixture>: skipped -> linted, findings printed. Without the flag an expect file is still skipped.
• Per-PR CI: two sub-second lint tests added to the small set; nothing else changes.

Where to look. .github/workflows/build.yml, the "Slow Release Tests" step - the ctest line sits after the case, so every arm, the 32-bit Windows one included, runs it; utils/lint/main.das set_skip_reason, one condition.

Validation, claims, ledger

Validation

• The exact nightly command, locally (macOS arm64, Release): ctest --test-dir build --build-config Release -L big -E memory_model_4gb --output-on-failure - 6 of 6 pass in 21 s (standalone_ctx, nano_ctx, concurrent_init, concurrent_init_dyn, capi_split_init, cpp_module_pull). Before the lint fix and relabel the same command was 6 of 8, the two style_lint tests failing on a skipped fixture.
• ctest -L small -R style_lint: 2 of 2 after the relabel; tests/lint/test_stale_nolint.das 7 of 7, its lint-skip-file arm being the one a first cut of this fix broke on the modules lane.
• Negative control on the lint fix: the fixture without --lint-fixtures still reports SKIP; with it, two STYLE014 findings.
• python3 ci/test_ci_matrix.py: 18 tests OK; the workflow parses.
• Preflight fast tier on the tip: 10 gates pass, 0 fail, 3 skip as not reached (format, lint on both rails, hash-refs, review-md, md-ascii, ast-verify, ci-das, ci-matrix, compile-sweep over 760 roots).
• The codex round ran over the branch at its tip, read the workflow file, and returned no findings.
• The dragon read tests-cpp/big/REVIEW.md twice: the first round corrected the rationale (a manual full-workflow run can target a PR branch), the fresh cold read after it is all-OK, and it found the sentence in skills/internal/writing_cpp_tests.md that this change made false, fixed here.
• The comment harvest and style-hygiene rows were skipped by standing instruction.

Lint candidate

• A PR gate that maps each changed path to its CMakeLists.txt LABELS and fails a big hit whose PR body names no ctest -L big run; tests-cpp/REVIEW.das already parses those labels. Not written here.

Claims - stated, not tested

• The Windows and sanitizer cells were not run here; the step's case arms and env exports are unchanged, and the added line uses the same ctest --test-dir build --build-config shape the RelWithDebInfo nightly already uses. A big test that fails under a sanitizer shows up on the first nightly, which is the point.
• The 32-bit Windows arm skips the AOT sweep but runs the big set: its binaries come from the same Build step that already generates and compiles the watchdog's -ctx context on that lane.

Not done

• lookibed's proposed tests-cpp/big/standalone_module_global (a required module owning initialized globals, linked and run) is a real gap beside master's emit-only test and is not added here.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The changes are cohesive and low-risk, and the updated skip/label logic matches the stated CI/test intent without introducing obvious behavioral regressions.

Pull request overview

This PR closes a CI coverage gap by actually executing the big-labelled C++ tests in nightly Release runs (after the full AOT sweep), and fixes utils/lint so the --lint-fixtures override truly forces linting of expectation/fixture files. It also reclassifies the two style_lint fixture tests as small so they run on every PR.

Changes:

• Run ctest -L big -E memory_model_4gb in the nightly “Slow Release Tests” step for Release matrix cells.
• Fix utils/lint skipping logic so --lint-fixtures bypasses content-based skip reasons (including expect-style files).
• Relabel the two style_lint fixture tests from big to small, and update the related checklist/docs accordingly.

File summaries

File	Description
.github/workflows/build.yml	Adds nightly execution of big ctest label after the AOT sweep in Release cells.
utils/lint/main.das	Makes --lint-fixtures override content-based skip reasons so fixtures are actually linted.
tests-cpp/big/style_lint/CMakeLists.txt	Relabels fixture lint tests as small and wires them into the test-small aggregate target.
tests-cpp/big/REVIEW.md	Updates the “big tests” checklist language to reflect nightly/manual workflow coverage.
skills/internal/writing_cpp_tests.md	Updates guidance to reflect that big tests run on nightly/manual full-workflow runs, not per-PR.

Review details

• Files reviewed: 5/5 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🟢 Approval recommended

The changes are coherent and consistent across workflow, lint behavior, and documentation, and I didn’t find any correctness or integration issues in the updated call sites/targets.

Review details

• Files reviewed: 6/6 changed files
• Comments generated: 0 new
• Review effort level: Lite