diff options
| author | Lukasz Kasprzak <lukas@labunix.xyz> | 2026-08-12 22:46:36 +0200 |
|---|---|---|
| committer | Lukasz Kasprzak <lukas@labunix.xyz> | 2026-08-12 22:46:36 +0200 |
| commit | 0f6454c123f1514c1cf255faed8d9b8c752c6347 (patch) | |
| tree | a14a1e43e95d5bb3b03cd9832776461ea9440605 /lib/rites/rite_ef | |
| parent | a4ffb0234d198ef295554e3202db15403346fbed (diff) | |
| download | colitur-0f6454c123f1514c1cf255faed8d9b8c752c6347.tar.gz colitur-0f6454c123f1514c1cf255faed8d9b8c752c6347.zip | |
test(differential): two Layer C predicates that counted rows, not identities
Both are the "cardinality where identity was required" flavour, and both
were demonstrated passing a real defect by the fixture-refresh review
rather than argued.
C6 accepted `subset diffs [ Slug_f; Rank ]` while its own note said the
shape had narrowed to slug-only after lectio's Christmas-octave rank was
corrected. That did not merely over-permit: every one of the 138 rows
already carries a Slug_f diff, so an added Rank diff changed neither the
diff-set membership nor the count, and the entry was blind. Dropping the
RG 91 entry 17 elevation in temporal_ef (Class2 -> Class4 on octave days
5-7) left both differential tests green with C6 still reporting 138.
Narrowed to `diffs = [ Slug_f ]`; the same mutation now reddens both.
C14 gated on lectio's date alone, with no colitur-slug guard, though fix
round 1's finding 1 had established that a predicate must pin which
celebration its citation is about -- C1, C6 and C8 all carry that guard.
With compare_deferred switched to RG 98 origin order the observed
celebration flips to the Annunciation on all three dates, and C14 still
matched exactly 3 rows with its count pin green; the suite reddened only
because the sibling dates fell outside every predicate. Guard added.
No behaviour change: 277 tests green, and the allow-list counts are
unmoved (C6 138, C14 3).
Diffstat (limited to 'lib/rites/rite_ef')
0 files changed, 0 insertions, 0 deletions
