base red: cpu_class's missed-deadline test starts its chat on an agent's connection (#28963)

Merged
LA
5d ago
+13-1

Summary

Base red on feat (macOS): acpmux::cpu_class cpu_class_a_start_that_misses_its_deadline_says_what_its_harness_was_doing hits the 120 s nextest timeout. It fails at feat a30fe2b6 (job fae79b08e3411c76448859df, aws-m4pro-7) and in #30057's merge-head run 5dd8ef18.

Cause. #28963 (3214c6efb32, "the person's new chat never fails on a slow host") made the person's agent start wait with no deadline (hub/lifecycle.rs person_start). This test's client proves it is the person (_acpmux/person_prove), so the start it gates never fails, and the test waits until nextest kills it. #28963 moved its own deadline test in lifecycle_fixes.rs to an agent's connection, but tests/cpu_class.rs (macOS only) was missed.

Fix (test only). The chat is started and prompted on a connection that is not the person's (agent_client, as lifecycle_fixes.rs does), whose start keeps the deadline. The product does not change.

Testing

runrefresult
redfeat a30fe2b6macOS fae79b08 (m4pro-7): TIMEOUT 120 s, compiled acpmux
green cpu_classhead bbb0f0a773aemacOS 5a0e676c (m4pro-3): 9/9, compiled acpmux
clippyheadLinux 3bc234e0 (run 41271636272): pass

Observed / Not run

Observed: the runs above. Not run: Linux (the file is #![cfg(target_os = "macos")]).

Changelog

none

Proof

No capture: a test-only change.


View with [code]smith Autofix with [code]smith
Need help on this PR? Tag @codesmith-bot with what you need.


Summary by cubic

Fixes the macOS-only cpu_class missed-deadline test so it no longer hits the 120-second nextest timeout.

Since #28963 gave the person's agent start no deadline, the test's client—which proves it is the person—never saw its start fail. The chat now starts on an agent's connection instead, whose start keeps the deadline, matching how lifecycle_fixes.rs handles it. Test-only change; no product behavior changes.