You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 361aaef
Browse filesBrowse the repository at this point in the historyBrowse files
fix(stack-encrypt): address the review of the bytes order refusal
The engine change in the previous commit was sound. The review found
text and tests that still described the old rule, or that no longer
checked anything.
- The declaration grammar's admission table said bytes admit `ore`,
`ope` and `ore_block`. A runner written from it would admit a
declaration the corpus refuses. The table now refuses them, and says
why.
- `coverage.md` listed the order index on bytes as a combination the
corpora record. They no longer do.
- The Go CHANGELOG named `index=ore_block` as a tag that is now
refused. It is not a tag word, and stashgen refused it before this
change. Only a hand-built plan with an `ore_block` output on bytes
is newly refused.
- The rustdoc for `Reason::IndexNotAdmitted` and `FieldPlan::with_type`
lists the refused pairs. It now names an order index on bytes.
- A stashgen test checked that no notice held a text the generator no
longer writes, so it could not fail. It now checks that no notice
names one of the test's fields.
- The Go fake engine kept its own list of the kinds the orderings are
answered over, beside `indexApplies`. That is the second list the
previous commit removed from the engine. `answeredBy` now reads
`indexApplies` alone, and the unused `logical` helper is removed.
0 commit comments