spicedb 1.54.0

[BrewTestBot]

Created by brew bump

---

Created with brew bump-formula-pr.

Details

release notes

### Security
- Prevent cache poisoning. The dispatch Check cache key now incorporates check hints. See https://github.com/authzed/spicedb/security/advisories/GHSA-4vrg-r928-h5vv
Added

Query Planner: fast serialize/deserialize for query plans (feat: introduce a hand-written serialize/deserialize for iterators authzed/spicedb#3122)

Changed

Cache: switch to otter as the primary cache implementation (chore: use otter as the primary cache implementation and get rid of alternative implementations authzed/spicedb#3112)
Server handles: GRPCDialContext as a handle on the server used deprecated gRPC methods. We modernized it and renamed it to NewClient (chore: rework handles for gRPC tests authzed/spicedb#3147)

Fixed

The watching schema cache (--enable-experimental-watchable-schema-cache) no longer enters permanent fallback on transient watch errors. A new supervisor restarts the watch cycle with bounded exponential backoff and only treats caller-driven cancellation or unsupported-watch as terminal (fix(schemacache): supervisor recovers from transient watch errors authzed/spicedb#3134)
Watch consumers that request WatchCheckpoints now eventually observe every revision returned by WriteRelationships as a checkpoint. MemDB regressed this in perf: have the Changes functions return an iter authzed/spicedb#2578 for no-op writes and MySQL never emitted checkpoints at all prior to now. Both now emit a checkpoint at the new revision. (fix: Watch checkpointing in memdb and mysql authzed/spicedb#3114)
When Query Planner evaluates a union, short-circuit if one of the branches yields a positive un-caveated result (feat: metrics module authzed/spicedb#3120)
DispatchQueryPlan previously did not try to use the singleflight middleware for check calls. (fix: route query plan dispatch checks through singleflight authzed/spicedb#3119)
Fixed regression introduced in 1.53.0. Postgres HeadRevision no longer allocates a new transaction ID on every call (fix: postgres implementation of HeadRevision authzed/spicedb#3127)
Fixed regression introduced in 1.53.0 for MySQL migration scripts (fix: mysql migrations authzed/spicedb#3129)
Query Planner: LookupSubjects no longer returns a subject excluded from a wildcard (e.g. viewer:* - banned) when the exclusion feeds an intersection (experimental --experimental-query-plan ls) (fix(query): respect wildcard ExcludedSubjects in LookupSubjects intersection authzed/spicedb#3136)
Tracing: When server is shutting down, flush traces. Also, elide the need for setting OTEL_EXPORTER_OTLP_ENDPOINT. (otel: replace cobraotel with native lifecycle management authzed/spicedb#3108)
Fixed a LookupSubjects issue in the query planner around the handling of wildcards in compound permissions (fix: three correctness bugs surfaced by GRPC consistency cross-check authzed/spicedb#3140)
MySQL: identifiers (object/subject IDs and relationship counter names) are now stored with a case-sensitive (binary) collation, matching the Postgres, CockroachDB, and Spanner datastores. Previously, identifiers differing only in letter case (e.g. Foo and foo) incorrectly collided in unique indexes and lookups. ⚠️ The migration rebuilds the relation_tuple table in place via ALTER TABLE, which can hold a metadata/table lock for a long time on large datasets — run the upgrade in a low-traffic window, or apply it with an online schema-change tool (e.g. gh-ost). (fix(mysql): store identifiers with a case-sensitive (binary) collation authzed/spicedb#3161)
server.NewConfigWithOptionsAndDefaults now populates Config and its embedded structs with the same defaults as the CLI flags, fixing zero-value behavior when embedding SpiceDB as a library. (fix: set library-friendly defaults on server Config authzed/spicedb#3156)

View the full release notes at https://github.com/authzed/spicedb/releases/tag/v1.54.0.

---

[github-actions]

🤖 An automated task has requested bottles to be published to this PR.

Caution

Please do not push to this PR branch before the bottle commits have been pushed, as this results in a state that is difficult to recover from. If you need to resolve a merge conflict, please use a merge commit. Do not force-push to this PR branch.