Context
From the first full mutation-testing run (ADR-007 workflow, 2026-07-04, partial). Two client-side survivors indicate plausible behavioural gaps; both need a confirming test (or a finding of existing coverage) before fixes.
Surviving mutants
src/client/hooks.rs:111 — replace <impl Clone for LifecycleHooks<C>>::clone -> Self with Default::default(). Default yields all-None hooks, so cloning on the pooled-reconnect path (src/client/connect_parts.rs:81) would silently drop configured callbacks. No test sets a hook, reconnects, and asserts the hook still fires.src/client/send_streaming.rs:96 — replace SendStreamingConfig::timeout -> Option<Duration> with None. Disabling the timeout entirely passes the suite: existing timing tests assert success paths only, not that a too-short timeout aborts a slow stream. (The neighbouring chunk_size mutants are already caught by frame-count assertions.)
Suggested tests
- Configure
on_connect (or another lifecycle hook), force a pool recycle/reconnect, and assert the callback fires on the new connection.
- Configure a very short
send_streaming timeout against a deliberately slow body reader and assert the call returns TimedOut and stops emitting frames.
Context
From the first full mutation-testing run (ADR-007 workflow, 2026-07-04, partial). Two client-side survivors indicate plausible behavioural gaps; both need a confirming test (or a finding of existing coverage) before fixes.
Surviving mutants
src/client/hooks.rs:111—replace <impl Clone for LifecycleHooks<C>>::clone -> Self with Default::default().Defaultyields all-Nonehooks, so cloning on the pooled-reconnect path (src/client/connect_parts.rs:81) would silently drop configured callbacks. No test sets a hook, reconnects, and asserts the hook still fires.src/client/send_streaming.rs:96—replace SendStreamingConfig::timeout -> Option<Duration> with None. Disabling the timeout entirely passes the suite: existing timing tests assert success paths only, not that a too-short timeout aborts a slow stream. (The neighbouringchunk_sizemutants are already caught by frame-count assertions.)Suggested tests
on_connect(or another lifecycle hook), force a pool recycle/reconnect, and assert the callback fires on the new connection.send_streamingtimeout against a deliberately slow body reader and assert the call returnsTimedOutand stops emitting frames.