Tags: dormonbear/codegraph
Tags
test: raise colbymchenry#662 daemon-recovery waitFor timeouts for CI … …load The shared-daemon recovery test flaked twice in the v0.5.3 publish run — daemon cold-start + first response exceeded the 10s wait under runner contention (logs showed 'too many open files'). Bump those two waits to 20s and the test budget to 75s; passes in <1s locally in the full suite.
chore: release v0.5.2 — large-org retrieval-hang fix + upstream sync (f… …34f606) - Bound the MCP catch-up gate (shipped in the prior commit) so the first query on a large Salesforce org serves the slightly-stale graph instead of blocking on a tens-of-thousands-of-files catch-up until the 60s watchdog kills it. - Drop the sync change-scan event-loop yield from that prior commit: it let a tool call racing the in-flight catch-up touch the SQLite connection mid-sync and hang (reproduced via the e2e harness, delay<300ms). The bounded gate alone fixes the reported hang; the yield was only a latency refinement. - Sync upstream @colbymchenry/codegraph f34f606 (same-file value-reference edges, 15 languages, colbymchenry#897); forkedFrom.upstreamCommit -> f34f606. EXTRACTION_VERSION unchanged (upstream didn't bump; avoids forcing a full re-index on large orgs — the new edges surface on the next index).