Skip to content

Commit 660295f

Browse files
authored
Merge pull request #898 from AdaWorldAPI/claude/x265-x266-plans-review-h9osnl
CI: gate lance-graph-callcenter under --features query (close the blind gate)
2 parents b66c7a0 + 2222df3 commit 660295f

2 files changed

Lines changed: 34 additions & 0 deletions

File tree

‎.github/workflows/rust-test.yml‎

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -150,6 +150,24 @@ jobs:
150150
# (deepnsm / supervisor / ogar above). One scoped step arms it. Gating.
151151
- name: Run bgz-tensor tests (workspace-excluded, ndarray sibling)
152152
run: cargo test --manifest-path crates/bgz-tensor/Cargo.toml --lib
153+
# lance-graph-callcenter UNDER `--features query` — the same blind gate as
154+
# supervisor above, but one level subtler: the crate was not merely
155+
# untested, it has `default = []`, so even a bare `cargo test` on it would
156+
# compile almost nothing and report a green it had not earned.
157+
#
158+
# Measured, not hypothetical. Running this combination by hand on
159+
# 2026-08-05 found 209 passed / 2 FAILED: the `graph_table` fixture minted
160+
# rows with the V1 `CLASSID_OSINT` while `OSINT_GOTHAM.classid` resolves to
161+
# `CLASSID_OSINT_V3` under the default `guid-v3-tail`, so
162+
# `project_snapshot`'s class filter matched zero of two rows and every
163+
# assertion ran against an empty snapshot (fixed in #897; recorded as
164+
# E-A-FEATURE-CONDITIONAL-CLASSID-SILENTLY-EMPTIED-A-FIXTURE-1). The same
165+
# feature gap had already produced a clippy EXIT=0 over never-compiled
166+
# `vsa_udfs` / `transcode::ontology_table` earlier the same day.
167+
#
168+
# 211 tests, ~2 min incremental. Gating.
169+
- name: Run callcenter tests --features query (blind-gate closure)
170+
run: cargo test --manifest-path crates/lance-graph-callcenter/Cargo.toml --features query --lib
153171
# Codex P2 (PR #861): the two self-asserting probe examples validate
154172
# their fixtures via assert! inside example main() — cargo test never
155173
# executes an example's main(), so those asserts only ever ran on a

‎.github/workflows/style.yml‎

Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -102,6 +102,22 @@ jobs:
102102
- name: Clippy bgz-tensor (advisory)
103103
continue-on-error: true
104104
run: cargo clippy --manifest-path crates/bgz-tensor/Cargo.toml --all-targets -- -D warnings
105+
# Tier A (mandatory, gating): lance-graph-callcenter UNDER `--features query`.
106+
#
107+
# The feature matters more than the crate. `default = []` here, so a bare
108+
# clippy run on this crate compiles almost none of it and exits 0 while
109+
# `vsa_udfs` and `transcode::ontology_table` — both behind
110+
# `query` / `query-lite` — are never seen. That false green is measured,
111+
# not hypothetical: it happened twice in one session (2026-08-05), and the
112+
# second time it hid two genuinely failing tests as well
113+
# (E-A-FEATURE-CONDITIONAL-CLASSID-SILENTLY-EMPTIED-A-FIXTURE-1).
114+
#
115+
# A green run that never compiled the module is not coverage. Gated
116+
# mandatory rather than advisory because the crate is clean TODAY
117+
# (0 errors under this feature) — same posture as deepnsm: gate it while
118+
# it is clean so it cannot silently regress.
119+
- name: Clippy callcenter --features query (mandatory)
120+
run: cargo clippy --manifest-path crates/lance-graph-callcenter/Cargo.toml --features query --lib --tests -- -D warnings
105121

106122
format:
107123
runs-on: ubuntu-24.04

0 commit comments

Comments
 (0)