prisma-risk / tsoracle / 26416149253

95%

push

github

refactor(client): split retry.rs god-module into retry/attempt/leader_hint (#433)

* refactor(client): collapse StaleLeaderHint/HintRejected into HintUnusable { reason }

The two AttemptOutcome arms had converged onto near-identical bookkeeping;
fold them into one variant carrying a HintUnusableReason that selects the
loop's reaction (StaleEpoch rides out the election + bumps the stale metric,
Rejected fails fast). No behavior change; the existing ride-out and
fail-fast tests pin the split.

* refactor(client): lift shared retry test helpers into test_support

enable_tracing/short_policy/make_status_with_hint are used by the retry,
attempt, and leader-hint unit tests; co-locate them in a cfg(test) module so
the upcoming module split does not duplicate them.

* refactor(client): extract attempt() and AttemptOutcome into attempt module

Move the single-attempt connect/rpc/evict path and its outcome enum out of
the retry loop module. All seven counter emissions move verbatim with their
code paths. Doc links to crate::retry::attempt updated to crate::attempt::attempt.

* refactor(client): extract NOT_LEADER classification into leader_hint module

classify_not_leader_hint and rejects_plaintext_hint now live with the cache
write they gate (compare_and_set_leader), co-locating the decision and its
side effect. attempt() and the retry loop both route NOT_LEADER through here.
Doc link in transport.rs updated.

* refactor(client): retarget module docs after the retry split

Point retry.rs's module doc at the new attempt/leader_hint modules, fix a now
out-of-scope intra-doc link, align transport.rs prose with the moved plaintext
guard, drop an internal doc path from a test comment, and remove a redundant
test import. No code change.

454 of 482 new or added lines in 4 files covered. (94.19%)

10598 of 10994 relevant lines covered (96.4%)

547136.95 hits per line