GLM-5.3 · Balanced

implement-access-token-context-encoding-framework-keycloak

Keycloak · Java · L · 300–799 lines · 6 files

The normal review, with one generalist agent. Judge: Claude Haiku 5.5.

Gold bugs found
0 of 0
Comments published
3
Comments matching a bug
0
Comments not matching
3
Cost of this PR
$1.565

Published comments (3)

mediumbugtestsuite/integration-arquillian/tests/base/src/test/java/org/keycloak/testsuite/AssertEvents.java:476-492

isAccessTokenId slices wrong chars and inverts grant-shortcut comparison — The grant check in `matchesSafely` never fires, so `expectRefresh` and sibling matchers degrade to a bare UUID check that accepts any grant, and short prefixes throw `StringIndexOutOfBoundsException`. Slice the grant segment with `substring(4, 6)` per `DefaultTokenContextEncoderProvider` and return false only when it differs from `expectedGrantShortcut`.

mediumbugservices/src/main/java/org/keycloak/protocol/oidc/TokenManager.java:1052-1054

Grant-type token ID encoding always emits "na" for client-credentials, UMA and token-exchange tokens — The registered "cc" and "pg" shortcuts stay dead because client-credentials, UMA, and token-exchange issuance (check `ClientCredentialsGrantType`, `AuthorizationTokenService`, `AbstractTokenExchangeProvider`) never set `Constants.GRANT_TYPE` on `clientSessionCtx`, defeating the introspection and admin optimizations. Set the attribute in those paths or derive the grant centrally in `initToken`.

mediumbugservices/src/main/java/org/keycloak/protocol/oidc/encode/AccessTokenContext.java:69-73

Constructor re-checks grantType instead of rawTokenId, leaving null rawTokenId unchecked — A null `rawTokenId` slips through, and `encodeTokenId` would later concatenate it into a token ID ending in ":null". Pass `rawTokenId` to the final `Objects.requireNonNull` call.

Gold bugs found (0)

None.

Gold bugs missed (0)

None.