1b993c93

test: a styled specimen, so the byte-exact leg has something to compare

a73x 2026-08-09 20:49

Commit message

test: a styled specimen, so the byte-exact leg has something to compare

Campaign row 4 broke the delta paint's leading SGR reset and both suites
passed. The reading at the time was that the styled (--vt) comparison is
the right instrument handed no specimen: nothing in the corpus had ever
emitted a colour, so a suppressed reset reset nothing.

This is the specimen — two bold-red rows from the session's own shell,
followed by a plain prompt row that must come back unstyled. It does not
catch mutation 4 (that resolution is written up in the campaign doc; the
premise was wrong, and the row is banked). It does catch a mutation that
strips every SGR run from delta rows: plain diff empty, styled leg
alone, at this scenario before any other. Until now the only source of
styled bytes anywhere in the suite was the prediction overlay's own
underline, so daemon-side style emission was never under the check.

Pins move to 10 scenario checkpoints and 22 convergence points.

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

test/e2e.sh

Old		New
@@ -216,7 +216,7 @@ cleanup() {
216	"$OUT.p1.err" "$OUT.pb.err" "$OUT.pw.err" "$OUT.rw.err" "$OUT.pr.err" \	216	"$OUT.p1.err" "$OUT.pb.err" "$OUT.pw.err" "$OUT.rw.err" "$OUT.pr.err" \
217	"$OUT.p1.d" "$OUT.pw.d" "$OUT.rw.d" "$OUT.pr.d" \	217	"$OUT.p1.d" "$OUT.pw.d" "$OUT.rw.d" "$OUT.pr.d" \
218	"$SOCK" "$SOCK2" "$SOCK3" "$SOCK4" "$SOCK4.second" "$QKEY" "$QKEY.bad" \	218	"$SOCK" "$SOCK2" "$SOCK3" "$SOCK4" "$SOCK4.second" "$QKEY" "$QKEY.bad" \
219	"$OUT" "$OUT.kill" "$OUT.re" "$OUT.a" \	219	"$OUT" "$OUT.kill" "$OUT.re" "$OUT.a" "$OUT.st" \
220	"$OUT.b" "$OUT.via" "$OUT.dead" "$OUT.abort" "$OUT.m7" "$OUT.m7b" \	220	"$OUT.b" "$OUT.via" "$OUT.dead" "$OUT.abort" "$OUT.m7" "$OUT.m7b" \
221	"$OUT.q" "$OUT.qc" "$OUT.qr" "$OUT.qa" "$OUT.qk" "$QKEY.wrong" \	221	"$OUT.q" "$OUT.qc" "$OUT.qr" "$OUT.qa" "$OUT.qk" "$QKEY.wrong" \
222	"$OUT.nokey" "$SOCK4.nokey" \	222	"$OUT.nokey" "$SOCK4.nokey" \
@@ -298,6 +298,29 @@ rm -f "$OUT.doctored" "$OUT.doctored.render" "$OUT.doctored.dump" \
298	"$OUT.doctored.rvt" "$OUT.doctored.dvt"	298	"$OUT.doctored.rvt" "$OUT.doctored.dvt"
299	ok "convergence control fires on a doctored stream"	299	ok "convergence control fires on a doctored stream"
300		300
		301	# --- M11: a styled specimen, so the byte-exact leg has something to compare.
		302	# The campaign's row 4 broke the delta paint's leading SGR reset and BOTH
		303	# suites passed — not because the styled comparison is weak but because no
		304	# scenario in the corpus had ever emitted a colour. A suppressed reset resets
		305	# nothing when nothing upstream is styled, so the instrument was handed no
		306	# specimen. This scenario is the specimen: two delta rows carrying a live
		307	# attribute, followed by a plain prompt row that must come back unstyled.
		308	#
		309	# The escapes are doubled because they are written by the SESSION's shell,
		310	# not by this one: what goes down the pipe is the literal text
		311	# `printf "\033[1;31mstyled-%s\033[0m\n" red bold`.
		312	{ printf 'printf "\\033[1;31mstyled-%%s\\033[0m\\n" red bold\n'; sleep 2; printf '\034'; } \| \
		313	timeout 30 "$MUX" --sock "$SOCK" > "$OUT.st"
		314	grep -q "styled-red" "$OUT.st" \|\| {
		315	echo "e2e FAIL: styled output never reached the client"; cat "$OUT.st"; exit 1; }
		316	grep -q "styled-bold" "$OUT.st" \|\| {
		317	echo "e2e FAIL: the second styled row never reached the client"; cat "$OUT.st"; exit 1; }
		318	# The plain leg would pass on a bled attribute — same glyphs, wrong colours.
		319	# The --vt leg inside assert_converged is the one that speaks here.
		320	assert_converged "$OUT.st" "$SOCK" "styled content"
		321	rm -f "$OUT.st"
		322	ok "styled content survives the paint path"
		323
301	# --- M3: kill a client mid-run; daemon survives; reattach lands correctly.	324	# --- M3: kill a client mid-run; daemon survives; reattach lands correctly.
302	# No convergence on $OUT.kill: kill -9 truncates the stream mid-paint, possibly	325	# No convergence on $OUT.kill: kill -9 truncates the stream mid-paint, possibly
303	# mid-escape-sequence, so a half-drawn capture is the expected shape here.	326	# mid-escape-sequence, so a half-drawn capture is the expected shape here.
@@ -1476,15 +1499,15 @@ D5PID=""
1476		1499
1477	# The pins. Literals, not variables set from counting something else —	1500	# The pins. Literals, not variables set from counting something else —
1478	# "assert the literal, never the constant the code under test reads"	1501	# "assert the literal, never the constant the code under test reads"
1479	# (decisions.md, M10). 9 scenario checkpoints; 21 convergence points.	1502	# (decisions.md, M10). 10 scenario checkpoints; 22 convergence points.
1480	# Anyone adding a scenario updates these by hand, on purpose.	1503	# Anyone adding a scenario updates these by hand, on purpose.
1481	[ "$OK_COUNT" = "9" ] \|\| {	1504	[ "$OK_COUNT" = "10" ] \|\| {
1482	echo "e2e FAIL: $OK_COUNT scenario checkpoints ran, the pin says 9 —"	1505	echo "e2e FAIL: $OK_COUNT scenario checkpoints ran, the pin says 10 —"
1483	echo " a scenario was added (update the pin) or silently lost"	1506	echo " a scenario was added (update the pin) or silently lost"
1484	exit 1	1507	exit 1
1485	}	1508	}
1486	[ "$CONV_COUNT" = "21" ] \|\| {	1509	[ "$CONV_COUNT" = "22" ] \|\| {
1487	echo "e2e FAIL: $CONV_COUNT convergence points ran, the pin says 21"	1510	echo "e2e FAIL: $CONV_COUNT convergence points ran, the pin says 22"
1488	exit 1	1511	exit 1
1489	}	1512	}
1490	echo "e2e OK (9 scenarios, 21 convergence points)"	1513	echo "e2e OK (10 scenarios, 22 convergence points)"