You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Root-caused and fixed everything CI flagged red after D4 (7672f52) landed:
- indexer/parser.rs: walk_calls fed an immediately-invoked closure's whole
multi-line body to split_receiver_callee/selected_callee_byte_span as if it
were a simple dotted callee expression. The naive last-`.`/`::` scan latched
onto an unrelated nested call and derived a byte span that collided with
that call's own correctly-walked RawCall, tripping D4's new UNIQUE
call_sites identity index and crashing fitness-check on a real repo. Now
guarded to only run that logic on single-line fn_node text.
- scip/ingest.rs, scip/parse.rs, lsp/overlay.rs, lsp/provider.rs,
tools/recover.rs: clippy fixes (type_complexity via a CallEdgeRow alias,
needless_borrow, needless_update, map_flatten, too_many_arguments,
assertions_on_constants, collapsible_if) across every feature-combo CI
actually lints with, including one only visible under lsp-overlay/
all-languages and one masked by the fmt-check step short-circuiting first.
- embedding.rs: pre-existing (non-D4) cargo fmt violations.
Verified: all 6 CI clippy invocations + fmt --check clean locally, fresh
index of a real repo no longer hits the UNIQUE constraint, and two clean
full `cargo test --workspace --all-features` runs both pass (937 tests, 0
failed) with no competing background processes -- including
watcher_reindexes_add_and_delete and
coordinator_refuses_phase_three_commit_after_a_newer_baseline_is_queued,
each of which flaked once earlier under this session's own resource
contention (stacked calm serve daemons + concurrent builds) but passed
cleanly in isolation and in the final clean run.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
0 commit comments