bd03a5c7

fix(test): a wedged daemon fails leakcheck loudly; the sweep pins its own input

a73x 2026-08-14 08:57

Commit message

fix(test): a wedged daemon fails leakcheck loudly; the sweep pins its own input

leakcheck's poll could expire with the pid still alive and then grep a log
whose verdict was never written — reading nothing, finding nothing, and
returning 0. That swallowed two failures at once: the unreadable verdict,
and the `muxd stop` that had not killed anything. An expired poll now says
so through the file's own `why` channel.

The e2e sweep's `grep "$OUT".*` silently reads nothing if the captures ever
stop surviving to the end, so the gate no longer rests on that unstated
invariant: d1.d, written at daemon-one spawn and kept until the trap, is
asserted present before the sweep greps.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

test/agent.sh

Old		New
@@ -210,6 +210,11 @@ leakcheck() {
210	sleep 0.05	210	sleep 0.05
211	_i=$((_i + 1))	211	_i=$((_i + 1))
212	done	212	done
		213	# An expired poll is not a clean answer: the verdict is written on the way
		214	# out, so a daemon still alive here has written nothing and the grep below
		215	# would read a stale log and call it healthy — swallowing both the
		216	# unreadable verdict AND the `muxd stop` that failed to kill anything.
		217	kill -0 "$1" 2>/dev/null && { why "daemon still alive ${_i}x50ms after stop — leak verdict unreadable"; return 1; }
213	_dl="$XDG_STATE_HOME/mux/muxd.log"	218	_dl="$XDG_STATE_HOME/mux/muxd.log"
214	[ -f "$_dl" ] \|\| return 0	219	[ -f "$_dl" ] \|\| return 0
215	grep -q "LEAK:" "$_dl" \|\| return 0	220	grep -q "LEAK:" "$_dl" \|\| return 0

test/e2e.sh

Old		New
@@ -3090,6 +3090,10 @@ ok "hub narrates the tear; the replica re-attaches across an epoch"
3090	kill "$DPID" 2>/dev/null \|\| true	3090	kill "$DPID" 2>/dev/null \|\| true
3091	wait "$DPID" 2>/dev/null \|\| true	3091	wait "$DPID" 2>/dev/null \|\| true
3092	DPID=""	3092	DPID=""
		3093	# The sweep must have something to read: d1.d is created at daemon-one
		3094	# spawn and survives to the trap, so its absence means the capture
		3095	# convention broke and this gate is reading nothing.
		3096	[ -e "$OUT.d1.d" ] \|\| { echo "e2e FAIL: leak sweep found no captures to read"; exit 1; }
3093	# Every capture this suite wrote — daemon stderr AND client output — is a	3097	# Every capture this suite wrote — daemon stderr AND client output — is a
3094	# lifecycle log now: any binary that leaked printed a grep-able marker.	3098	# lifecycle log now: any binary that leaked printed a grep-able marker.
3095	if grep -q "LEAK:" "$OUT".* 2>/dev/null; then	3099	if grep -q "LEAK:" "$OUT".* 2>/dev/null; then