AtlatestRepositorysigil-doc
sigil-doc / tree / test / transcripts / sd-p1sabotage.txt
1
==============================================================2
LEG 1 (product) -- THE NAMED SABOTAGE: the root allowlist is ignored.3
Every binding, every message, every location stays in place; only the4
membership test is removed. This is the falsification for the phase.5
test/sabotage-sd-p1.sh: line 84: label: unbound variable6
--- leg1-allowlist-ignored: exit=17
9 passed, 2 failed8
✗ the root allowlist rejects a root it does not name, in both directions9
✗ @text and @expr are dispatched as author commands, not pseudo-kinds10
✗ Some tests failed11
==============================================================12
LEG 2 (product) -- multiplicity is ignored; every profile accepts any13
number of roots.14
test/sabotage-sd-p1.sh: line 84: label: unbound variable15
--- leg2-multiplicity-ignored: exit=116
10 passed, 1 failed17
✗ multiplicity is a parameter, and is enforced in both directions18
✗ Some tests failed19
==============================================================20
LEG 3 (product) -- the DEFAULT profile is silently widened, so21
sigil/doc would start accepting a folio root.22
test/sabotage-sd-p1.sh: line 84: label: unbound variable23
--- leg3-default-widened: exit=124
6 passed, 5 failed25
✗ the root allowlist rejects a root it does not name, in both directions26
✗ multiplicity is a parameter, and is enforced in both directions27
✗ the default profile is today's policy and today's diagnostics, exactly28
✗ @text and @expr are dispatched as author commands, not pseudo-kinds29
✗ a root-policy error still precedes any lowering error30
✗ Some tests failed31
==============================================================32
LEG 4 (product) -- spans are measured in SCALARS instead of UTF-8 BYTES.33
This is the exact bug the byte-vs-scalar precondition on the location arm34
exists to make detectable.35
test/sabotage-sd-p1.sh: line 84: label: unbound variable36
--- leg4-scalar-spans: exit=137
10 passed, 1 failed38
✗ a profile with a different root set parses, with exact locations39
✗ Some tests failed40
==============================================================41
LEG 5 (product) -- top-level classification reverts to the HEAD SYMBOL,42
so an author's @text{} command is mistaken for the parser's own prose43
pseudo-node again.44
test/sabotage-sd-p1.sh: line 84: label: unbound variable45
--- leg5-head-symbol-classification: exit=146
10 passed, 1 failed47
✗ @text and @expr are dispatched as author commands, not pseudo-kinds48
✗ Some tests failed49
==============================================================50
LEG 6 (product) -- trim-body-edges reverts to the HEAD SYMBOL.51
This is the site the first version of this phase MISSED, and the most52
dangerous of the family: it does not merely misread a @text{} command,53
it REBUILDS it as a five-slot prose node, so every later stage then54
behaves correctly on the wrong node. Fixing lower-block alone left the55
block-position case completely unchanged and looked like success.56
test/sabotage-sd-p1.sh: line 84: label: unbound variable57
--- leg6-trim-body-edges-head-symbol: exit=158
10 passed, 1 failed59
✗ @text and @expr are dispatched as author commands, not pseudo-kinds60
✗ Some tests failed61
==============================================================62
LEG 7 (ASSERTION layer) -- one expected byte offset in the suite is63
perturbed by one. This proves the location assertions actually execute64
rather than sitting in a branch nothing reaches.65
test/sabotage-sd-p1.sh: line 84: label: unbound variable66
--- leg7-assertion-layer: exit=167
10 passed, 1 failed68
✗ a profile with a different root set parses, with exact locations69
✗ Some tests failed70
==============================================================71
LEG 8 (instrument validity) -- CAN GATE ARM 2 FAIL AT ALL?72
Arm 2 asserts the classification change is confined: at least one73
colliding spelling moved, and every positive control is identical. Its74
INPUT is the committed BEFORE transcript. Here that input alone is75
neutered -- replaced with the CURRENT behaviour, control marker kept --76
while the gate's scoring and exit logic are left untouched. If arm 277
still passes, it was passing for free.78
--- leg8-before-transcript-neutered: exit=179
sigil 0.18.3 (sigil-lang 0.18.3-dev+0ed19fd1)80
SD_P1_GATE_01 root-profile suite 11/11, locations verified at 2 roots81
no colliding spelling changed behaviour; the classification fix is inert82
1d083
< PRE_SD_P1_CONTROL_UNBOUND (forged for leg 8 -- this transcript is NOT pre-SD-P1)84
==============================================================85
RESTORED -- every file back to its pristine hash. Re-running the suite86
green closes the loop: sabotage proves green->red, this proves red->green.87
db605aad575385989ea7a4dda851c6959ed520785e2b62e763f3f780a38cd447 src/sigil/doc/private/parser.sgl88
f7d6d8ea70931de2939a929e8f2b74fc9fe75e17aa4298042678478894740da8 src/sigil/doc/reader.sgl89
54e490b46d39ac503c8c77672a6717c745d372a9fd7da62ca19fec9e3989ce78 test/test-root-profile.sgl90
ba82579dd6be3ad26016d184d1ebba12db2bcc05dcbe3a585c55aef7040fef93 test/transcripts/sd-p1/classification-before.txt91
ok 11 passed test/test-root-profile.sgl93
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━94
✓ All tests passed!95
Files: 1 of 1 run96
Tests: 11 passed (11 total)97
Time: 3m 24.41s98
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━99
SD_P1_SABOTAGE_COMPLETE 8 legs100
SABOTAGE3_EXIT=0