Surfaced while bisecting #652. Two smoke flows fail on origin/master:
claude_attach_detachclient_attach_on_join
Why these are different from #652
#652's client_attach_idempotent regressed inside the 1fbfa0d..7116f95 window and was traced to a deliberate message split (LLP 0174/0178), so the smoke's expectation was stale.
These two are not in that window. They fail identically at 1fbfa0d (v1.19.0) with #652's fix stashed, so they are long-standing rather than recent. Neither is listed in the CLAUDE.md release battery, which is presumably why they have gone unnoticed.
Worth deciding
For each: is the flow asserting behaviour that has since been deliberately changed (like #652, so the smoke should be updated), or is it catching a genuine defect that has simply never been run? Both are attach-path flows, and the attach preconditions have moved at least once (f17b091), so the first is plausible for at least one of them.
If they are meant to be maintained, they arguably belong in the release battery; if they are obsolete, they should be retired rather than left failing, since a permanently-red smoke trains people to ignore smoke results.
Surfaced while bisecting #652. Two smoke flows fail on
origin/master:claude_attach_detachclient_attach_on_joinWhy these are different from #652
#652's
client_attach_idempotentregressed inside the1fbfa0d..7116f95window and was traced to a deliberate message split (LLP 0174/0178), so the smoke's expectation was stale.These two are not in that window. They fail identically at
1fbfa0d(v1.19.0) with #652's fix stashed, so they are long-standing rather than recent. Neither is listed in the CLAUDE.md release battery, which is presumably why they have gone unnoticed.Worth deciding
For each: is the flow asserting behaviour that has since been deliberately changed (like #652, so the smoke should be updated), or is it catching a genuine defect that has simply never been run? Both are attach-path flows, and the attach preconditions have moved at least once (
f17b091), so the first is plausible for at least one of them.If they are meant to be maintained, they arguably belong in the release battery; if they are obsolete, they should be retired rather than left failing, since a permanently-red smoke trains people to ignore smoke results.