Commit706b69cdRecorded23 Aug 2026Repositorysigil-doc

test: correct LEG 06's expectation to the assertion it actually reaches

Message

Run from leg 1 against the seven-arm gate, legs 1-5 reddened exactly where their expectations name -- including leg 5, the required one, which changes a migrated manuscript's prose and gets UNEXPECTED minimal.sgd.

LEG 06 reddened somewhere unpredicted, and the somewhere else is better than what it predicted. Disabling the position normalization leaks the header shift into EVERY header-migrated manuscript at once, so zero lower identically and the arm dies on its identical-FLOOR before it ever reaches the unexpected check that leg 5 trips. The two legs therefore fail at two DIFFERENT assertions in the same arm, which separates them more strongly than the two counts of one assertion I had written.

Corroboration tightened to match: the wall of unexpected rows, minimal.sgd among them, AND the ROOT-ONLY rows required to survive -- which is what proves this sabotage hit the position normalization specifically rather than the comparison as a whole.

Changed
 test/sabotage-sd-p8.sh | 42 ++++++++++++++++++++++++++++--------------
 1 file changed, 28 insertions(+), 14 deletions(-)
Diff
test/sabotage-sd-p8.shmodified
@@ -262,7 +262,7 @@ restore_all || die "restore failed; the worktree is still sabotaged"
262
# ---------------------------------------------------------------- LEG 06
263
# INSTRUMENT. Remove the differential's `make-srcloc` normalization.
264
#
−265
# The normalization is what makes arm 3's equality claim possible, and an
+265
# The normalization is what makes arm 5's equality claim possible, and an
266
# instrument gets none of the scrutiny the product does. Without it, every
267
# manuscript whose header moved differs in its embedded position DATA -- so the
268
# gate must report a wall of UNEXPECTED rather than quietly passing.
@@ -271,27 +271,41 @@ plant "$DUMP" \
271
'(and (pair? value)' \
272
'(and #f (pair? value)'
273
run_gate "$work/leg06.log" && die "LEG 06: the gate passed with the position normalization disabled"
−274
leg_red 06 "$work/leg06.log" "manuscripts changed for a reason this phase did not intend"
−275
# CORROBORATION, and it is not optional here. LEG 05 and LEG 06 both redden arm
−276
# 5 with the SAME sentence, so on that string alone the two legs cannot be told
−277
# apart by their own evidence -- either one would score the other's success.
−278
# They must be distinguished by WHICH manuscripts move:
+274
# EXPECTED AT A DIFFERENT ASSERTION THAN LEG 05, and that is the distinguishing
+275
# evidence rather than a nuisance.
276
#
−280
# LEG 05 changes one manuscript's prose -> exactly minimal.sgd is UNEXPECTED
−281
# LEG 06 disables the normalization -> every HEADER-MIGRATED manuscript
−282
# is UNEXPECTED, because their
−283
# embedded make-srcloc positions are
−284
# absolute again
+277
# The first version of this leg predicted the same string LEG 05 produces
+278
# ("manuscripts changed for a reason this phase did not intend"), which would
+279
# have made the two legs indistinguishable by their own output. Run, it reddened
+280
# somewhere else entirely -- and the somewhere else is better:
281
#
−286
# So this leg requires a WALL, not a row: at least ten unexpected manuscripts,
−287
# and minimal.sgd among them for a reason opposite to leg 5's.
+282
# LEG 05 changes ONE manuscript's prose. 16 of 17 still lower identically, so
+283
# the identical-floor holds and the arm dies on `unexpected != 0`,
+284
# naming minimal.sgd.
+285
# LEG 06 disables the normalization. EVERY header-migrated manuscript's
+286
# embedded positions go absolute at once, so ZERO lower identically and
+287
# the arm dies on its FLOOR first, before the unexpected check is
+288
# reached.
+289
#
+290
# So the two legs fail at two different assertions in the same arm, which is a
+291
# stronger separation than two counts of the same one. An arm reddening
+292
# somewhere unpredicted is a finding either way
+293
# (topics/an-expectation-names-which-check-must-fail); this is that rule paying
+294
# out on the leg written to satisfy it.
+295
leg_red 06 "$work/leg06.log" "manuscripts lower identically; the migration moved more than the header"
+296
# CORROBORATION: the wall, not a row. Under this sabotage the header shift leaks
+297
# into every migrated manuscript at once.
298
unexpected_rows=$( { rg -c '^UNEXPECTED' "$work/leg06.log" || echo 0; } )
299
test "${unexpected_rows:-0}" -ge 10 \
300
|| { rg '^UNEXPECTED|^IDENTICAL' "$work/leg06.log" >&2
301
die "LEG 06: disabling the normalization produced only ${unexpected_rows:-0} unexpected manuscripts; a header-shift leak should move every migrated one, so this red is probably not the one this leg is about"; }
302
rg -qF 'UNEXPECTED minimal.sgd' "$work/leg06.log" \
303
|| die "LEG 06: the wall of unexpected manuscripts does not include minimal.sgd"
−294
printf 'SD_P8_SABOTAGE_06 CORROBORATED %s manuscripts unexpected, not the single row leg 5 produces\n' "$unexpected_rows"
+304
# ...and the ROOT-ONLY rows must SURVIVE, which is what proves the sabotage hit
+305
# the position normalization specifically and not the comparison as a whole.
+306
rg -qF 'ROOT-ONLY sd-p3-wrapped.sgd' "$work/leg06.log" \
+307
|| die "LEG 06: the root-only classification collapsed too, so this sabotage is broader than the normalization it claims to disable"
+308
printf 'SD_P8_SABOTAGE_06 CORROBORATED %s manuscripts unexpected and the root-only rows intact, at the identical-floor rather than leg 5 unexpected-check\n' "$unexpected_rows"
309
restore_all || die "restore failed; the worktree is still sabotaged"
310
311
# ---------------------------------------------------------------- LEG 07