Skip to content

docs(adr-0122): D6 点明同构 pin 的编译期证明落在 check:test-typecheck 一步#6263

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
os-zhuang merged 1 commit into from
Aug 7, 2026
Merged

docs(adr-0122): D6 点明同构 pin 的编译期证明落在 check:test-typecheck 一步 #6263

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 17 additions & 4 deletions docs/adr/0122-schema-type-alias-naming-convention.md
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -133,10 +133,23 @@ Two alternatives were measured and rejected:
about types, and isomorphism rots: add a `.default()` three levels down and an alias
silently joins the covered set with no signal. So every exempt schema carries a
compile-time assertion that `z.input` and `z.infer` really are the same type, in
`packages/spec/src/type-alias-convention.pin.test.ts`. tsc proves the exemption on the
same run that type-checks the package, and the file goes red — naming the alias — the
day one stops being true. An exemption nobody can state falsely is the only kind worth
having; this is the same instinct as `check:durability-log-level`'s empty baseline.
`packages/spec/src/type-alias-convention.pin.test.ts`. tsc proves the exemption on every
`pnpm typecheck` run, and the file goes red — naming the alias — the day one stops being
true.

Which of that command's two **steps** proves it is worth stating precisely, because it is
not the obvious one. `packages/spec`'s `typecheck` script is
`tsc --noEmit && pnpm check:test-typecheck`, and the bare `tsc` runs the BUILD config,
whose `**/*.test.ts` exclusion keeps this `*.test.ts` pin file out of its program
entirely. The proof therefore lands in the **second** step, which puts the test layer back
in front of tsc over `tsconfig.test.json` (#5286). Measured, by giving a pinned schema a
`.default()`: the bare `tsc --noEmit` stays at exit 0, and `check:test-typecheck` is what
turns red, naming the file — `1 type error(s) in a file the ledger does not cover`. That
last phrase is the other half of the guarantee: the pin file carries no
`test-typecheck-debt.json` entry, so its baseline on that surface is **zero** errors and a
single new one fails the gate. An
exemption nobody can state falsely is the only kind worth having; this is the same instinct
as `check:durability-log-level`'s empty baseline.

**D7 — The backflow gate.** `pnpm check:spec-parsed-alias`
(`scripts/check-spec-parsed-alias.mjs`, in lint.yml's `lint` job) requires every bare
Expand Down
Loading