0e9ae602

refactor(test): one owner for the agent suite's daemon bring-up

a73x 2026-08-13 21:17

Commit message

refactor(test): one owner for the agent suite's daemon bring-up

Three scenarios spelled the same four lines: start_daemon, check for a
pid, wait_ready, and a `why` string for each half. start_ready folds them
into one call.

The pid variable is set the instant the pid is known, BEFORE the readiness
wait and whatever that wait decides — a daemon that came up and then never
answered is still a daemon this run started, and a version that assigned
only on success would leave it running past cleanup. That ordering is why
the helper takes the variable's name rather than returning the pid on
stdout: a command substitution would have to choose between carrying the
pid and carrying the failure reason.

The TUI and QUIC-tear scenarios keep their own bring-up: they want a
failure to become a stashed message rather than an immediate return.

Suite passes 9/9, run twice.

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

test/agent.sh

Old		New
@@ -222,6 +222,32 @@ wait_ready() {
222	return 1	222	return 1
223	}	223	}
224		224
		225	# start_ready PIDVAR LOG SOCK ARGS... — bring a daemon up and wait for it to
		226	# answer, storing its pid in the named variable and calling `why` with the
		227	# whole story if either half fails. The bring-up is identical in every scenario
		228	# that needs a plain daemon, and a scenario that got only PART of it right — a
		229	# pid but no readiness wait — would fail later, somewhere else, as a flake.
		230	#
		231	# PIDVAR is set the INSTANT the pid is known, before the readiness wait and
		232	# whatever that wait decides. That ordering is the load-bearing part: a daemon
		233	# that came up and then never answered is still a daemon this run started, and
		234	# a version that only assigned on success would leave it running past cleanup.
		235	#
		236	# The two scenarios that do NOT use this (the TUI and the QUIC tear) want a
		237	# daemon that failed to come up to be a stashed message rather than an
		238	# immediate return, and folding that in would cost more than it saves.
		239	start_ready() {
		240	_var="$1"
		241	_log="$2"
		242	_sock="$3"
		243	shift 3
		244	eval "$_var=\$(start_daemon \"\$_log\" \"\$@\")"
		245	eval "_pid=\$$_var"
		246	[ -n "$_pid" ] \|\| why "daemon never printed an up-line [$(cat "$_log")]" \|\| return 1
		247	wait_ready "$_sock" \|\| why "daemon never answered on $_sock" \|\| return 1
		248	return 0
		249	}
		250
225	now_ms() { python3 -c 'import time; print(int(time.time() * 1000))'; }	251	now_ms() { python3 -c 'import time; print(int(time.time() * 1000))'; }
226		252
227	echo "agent: ports ${PORT_TEAR}/${PORT_RELAY}/${PORT_QUIET}/${PORT_SINK}, tmp $TMP"	253	echo "agent: ports ${PORT_TEAR}/${PORT_RELAY}/${PORT_QUIET}/${PORT_SINK}, tmp $TMP"
@@ -232,9 +258,7 @@ echo "agent: ports ${PORT_TEAR}/${PORT_RELAY}/${PORT_QUIET}/${PORT_SINK}, tmp $T
232	# in `exit_code` the command's rather than a guess.	258	# in `exit_code` the command's rather than a guess.
233	scen_marks() {	259	scen_marks() {
234	[ -x /bin/bash ] \|\| { WHY="no /bin/bash to inject marks into"; return 2; }	260	[ -x /bin/bash ] \|\| { WHY="no /bin/bash to inject marks into"; return 2; }
235	D_MARKS=$(start_daemon "$TMP/marks.log" --sock "$SOCK_MARKS" --shell /bin/bash)	261	start_ready D_MARKS "$TMP/marks.log" "$SOCK_MARKS" --sock "$SOCK_MARKS" --shell /bin/bash \|\| return 1
236	[ -n "$D_MARKS" ] \|\| why "daemon never printed an up-line [$(cat "$TMP/marks.log")]" \|\| return 1
237	wait_ready "$SOCK_MARKS" \|\| why "daemon never answered on $SOCK_MARKS" \|\| return 1
238		262
239	timeout 20 "$MUXA" run --sock "$SOCK_MARKS" --timeout 8000 'true' >"$TMP/m1" 2>&1	263	timeout 20 "$MUXA" run --sock "$SOCK_MARKS" --timeout 8000 'true' >"$TMP/m1" 2>&1
240	_rc=$?	264	_rc=$?
@@ -399,9 +423,7 @@ run_scenario "ephemeral TUI: alt_screen seen, quit driven, death reported as JSO
399	# question a markless session can answer two different ways, and "I don't	423	# question a markless session can answer two different ways, and "I don't
400	# know" is not one of them.	424	# know" is not one of them.
401	scen_settle() {	425	scen_settle() {
402	D_SETTLE=$(start_daemon "$TMP/settle.log" --sock "$SOCK_SETTLE" --shell /bin/sh)	426	start_ready D_SETTLE "$TMP/settle.log" "$SOCK_SETTLE" --sock "$SOCK_SETTLE" --shell /bin/sh \|\| return 1
403	[ -n "$D_SETTLE" ] \|\| why "daemon never printed an up-line [$(cat "$TMP/settle.log")]" \|\| return 1
404	wait_ready "$SOCK_SETTLE" \|\| why "daemon never answered on $SOCK_SETTLE" \|\| return 1
405		427
406	timeout 20 "$MUXA" run --sock "$SOCK_SETTLE" --settle 300 --timeout 10000 'sleep 1' >"$TMP/s1" 2>&1	428	timeout 20 "$MUXA" run --sock "$SOCK_SETTLE" --settle 300 --timeout 10000 'sleep 1' >"$TMP/s1" 2>&1
407	_rc=$?	429	_rc=$?
@@ -622,10 +644,8 @@ run_scenario "quic: a tear with no path back fails with the whole story" scen_te
622	# happened, and the honest answer (still running) would have been one field.	644	# happened, and the honest answer (still running) would have been one field.
623	scen_keepalive() {	645	scen_keepalive() {
624	quic_gate \|\| return $?	646	quic_gate \|\| return $?
625	D_QUIET=$(start_daemon "$TMP/quiet.log" --sock "$SOCK_QUIET" --shell /bin/bash \	647	start_ready D_QUIET "$TMP/quiet.log" "$SOCK_QUIET" --sock "$SOCK_QUIET" --shell /bin/bash \
626	--quic "127.0.0.1:$PORT_QUIET" --key "$KEY")	648	--quic "127.0.0.1:$PORT_QUIET" --key "$KEY" \|\| return 1
627	[ -n "$D_QUIET" ] \|\| why "daemon never printed an up-line [$(cat "$TMP/quiet.log")]" \|\| return 1
628	wait_ready "$SOCK_QUIET" \|\| why "daemon never answered on $SOCK_QUIET" \|\| return 1
629		649
630	timeout 40 "$MUXA" await --quic "127.0.0.1:$PORT_QUIET" --key "$KEY" --timeout 20000 >"$TMP/q3" 2>&1	650	timeout 40 "$MUXA" await --quic "127.0.0.1:$PORT_QUIET" --key "$KEY" --timeout 20000 >"$TMP/q3" 2>&1
631	_rc=$?	651	_rc=$?