ci(gqt): build the ordinary job's engine graph once

.github/workflows/gq-logic-tests.yml

@@ -175,19 +175,26 @@ jobs:
           save-if: ${{ github.ref == 'refs/heads/main' }}
           cache-on-failure: true
 
+      # Every cargo invocation of this job selects the same three packages,
+      # so the engine graph resolves one feature set and builds once; a
+      # narrower selection rebuilds every crate whose features differ (the
+      # cache rule in docs/dev/ci.md). The seam guard's crate declares no
+      # workspace crate, so it adds no engine build.
       - name: Test unavailable DST runtime refusal
         if: needs.classify_changes.outputs.run_gqt == 'true' && matrix.mode == 'ordinary'
         # The flagless build shape this test exists for: an env RUSTFLAGS,
         # empty included, replaces the workspace `[build] rustflags`.
         env:
           RUSTFLAGS: ""
-        run: cargo test -p omnigraph-gqt --locked --timings --lib --test runner_dispatch -- --nocapture
+        run: cargo test -p omnigraph-gqt -p omnigraph-gqt-core -p omnigraph-bench --locked --timings --test runner_dispatch -- --nocapture
 
-      - name: Test ordinary GQT core execution
+      # The benchmark units include the fixture parity test, which reads a
+      # corpus case, so it must run when a cases-only diff skips `Test Workspace`.
+      - name: Test GQT, GQT core and benchmark units
         if: needs.classify_changes.outputs.run_gqt == 'true' && matrix.mode == 'ordinary'
         env:
           RUSTFLAGS: ""
-        run: cargo test -p omnigraph-gqt-core --locked --timings --lib -- --nocapture
+        run: cargo test -p omnigraph-gqt -p omnigraph-gqt-core -p omnigraph-bench --locked --timings --lib
 
       # A case's `at:` name arms a seam in the seam guard, so a cases-only
       # diff can change its verdict while `Test Workspace` is skipped.
@@ -197,14 +204,6 @@ jobs:
           RUSTFLAGS: ""
         run: cargo test -p omnigraph-seams --locked --timings --test failpoint_names_guard
 
-      # This parity test reads a corpus case, so it must run when a
-      # cases-only diff skips `Test Workspace`.
-      - name: Test benchmark fixture parity with the GQT corpus
-        if: needs.classify_changes.outputs.run_gqt == 'true' && matrix.mode == 'ordinary'
-        env:
-          RUSTFLAGS: ""
-        run: cargo test -p omnigraph-bench --locked --timings --lib branch_merge::tests::generated_dataset_matches_the_legacy_fixture_and_reversible_history -- --exact --nocapture
-
       # The workspace .cargo/config.toml enables the seeded Tokio runtime
       # from any directory; these steps run from the repo root.
       - name: Run complete GQ logic tests with DST

docs/dev/ci.md

@@ -54,8 +54,8 @@ all under `GQ Logic Tests`, whose `dst-clippy` job compiles the crate so
 case arming a seam turns the guard red where the PR can see it. The fixture
 parity test in `crates/omnigraph-bench/src/branch_merge.rs` also reads
 `generated_branch_merge_dataset.gqt` from the corpus. `GQT (ordinary)` runs
-that exact test on `run_gqt`, including cases-only changes, while
-`Test Workspace` covers it on engine input. No crate reads a deployment
+it with the other benchmark unit tests on `run_gqt`, including cases-only
+changes, while `Test Workspace` covers it on engine input. No crate reads a deployment
 file. `scripts/check-change-classes.py` keeps the
 literal-spelling half of this true: it replays the classifier over fixture
 diffs and fails when a string literal in a Rust or TOML file under `crates/`
@@ -152,13 +152,18 @@ Current-version live schema and policy deployment remains required by `Test Work
 
 `GQ Logic Tests` (`gq-logic-tests.yml`) owns the complete `.gqt` corpus as a
 required context aggregating three qualification jobs. `GQT (ordinary)` checks
-unit tests and unavailable-DST refusal under an empty `RUSTFLAGS`, then runs
-the seam guard (`crates/omnigraph-seams/tests/failpoint_names_guard.rs`) in
-the same flagless shape; the guard is a source walk whose crate declares no
-workspace crate (its dev-dependencies are `serde_yaml`, `syn`, `tempfile` and
-`toml`), so
-it adds no second engine build. The same ordinary job runs the benchmark
-fixture parity test against the generated branch-merge corpus case.
+unavailable-DST refusal and the unit tests of `omnigraph-gqt`,
+`omnigraph-gqt-core` and `omnigraph-bench` under an empty `RUSTFLAGS`, then
+runs the seam guard (`crates/omnigraph-seams/tests/failpoint_names_guard.rs`)
+in the same flagless shape. Both of its engine invocations select those same
+three packages, so they resolve one feature graph and the engine builds once:
+each crate alone resolves a different graph (`omnigraph-gqt` enables the
+engine's `test-util` and depends on `omnigraph-dst`), and three selections
+cost three engine builds, about 20 minutes each warm. The guard is a source
+walk whose crate declares no workspace crate (its dev-dependencies are
+`serde_yaml`, `syn`, `tempfile` and `toml`), so it adds no second engine
+build. The benchmark units include the fixture parity test against the
+generated branch-merge corpus case.
 The dispatch owner also checks external-store admission and persistence.
 Automatic corpus runs supply no `--store`, so schema-less query files are
 refused there; schema-and-seed datasets with zero steps are admitted.