Repository navigation
PASS
Baseline. This re-reviews bdd16511 against the PASS on b156d936 (this PR), on the same arch 7a86ec7b. I checked these facts. since.diff equals git diff b156d936 bdd16511. pr.diff equals git diff c509b4cb bdd16511, and c509b4cb (#3028) is the base. The compare is one merge of main (#3028) and no branch commit.
The merge. git merge-tree shows one content conflict, in scripts/session-e2e.sh. The PR's own diff before the merge (3c2b2a73..b156d936) and after it (c509b4cb..bdd16511) differ only in the VM-lane selector at line 421. The resolution keeps the host_exposures_follow_the_table case from #3028 and appends host_stop_stops_boxes_then_daemon_and_vm to the same lane. In the full run, #3028's proof runs at line 31258, before the host stop proof at line 31313. The host stop proof stays after the two daemon-respawn proofs, and no proof after it depends on the boxes that it stops. No Rust file changes in the compare.
Governs: architecture.md line 540 (Providers, hosts, identity, network): host stop stops each running box as min box stop does, then the daemon and, for a minvmd-managed VM, the VM. Every record resumes later, and --force forces both. Ruling: BCLI-049 conforms, with the same status as on b156d936. No precedent departs.
Not checked: I did not compile or run the code, as the instructions require. The VM leg ./scripts/session-e2e.sh host_stop_stops_boxes_then_daemon_and_vm is deferred to the VM lane, and nothing ran it on this head. It runs in the full session run in CI, as the PASS on fdffc0f4 accepted. The failure in agent-verify (exec_run_attach_and_task_run_share_the_commands_trace) is in a test that this PR does not touch. I did not establish its cause.
Ruled at bdd1651 against gominimal/arch@7a86ec7bb3c0 by claude-opus-5-5; arch_rulings 958, job ea7e782a-7705-4f30-89ec-454af20d1315.