Publish prep: the six runtime packages at 0.1.0 - #1
Merged
Merged
Conversation
The export map pointed `./hooks` at `dist/hooks/`, which tsc has no reason to emit - the hooks live in `runtime/hooks.ts` and are already re-exported from the root. Nothing in the repository imported the subpath, so it compiled clean and would have been ERR_MODULE_NOT_FOUND for the first consumer who tried it. A type checker cannot catch that, so `check-exports.mjs` does: every exports and bin target has to exist in the built tree and be covered by files[], and every publishable package needs a README and a LICENSE. It runs in CI on every pull request, beside `check-version.mjs`, which the release workflow uses to hold the tag and the manifests to the same number.
…t all `"license": "MIT"` in a manifest is a label, not the text, and npm only picks up a LICENSE sitting in the package's own directory - so nine packages declared a licence they did not carry. documents, textide and textide-git are marked private. They build in CI and they are in the repository, but their surface has not settled and an IDE is not something to tie to the runtime's release cadence. Being private is what holds them back, and deliberately the only thing that does: a filter you have to remember at publish time is a filter that gets forgotten.
`pnpm publish -r` already walks the dependency order and rewrites workspace:^, so the workflow's job is to refuse rather than to sequence: the same build, typecheck, lint and test CI runs, then the exports guard, then a check that the tag and the manifests agree. It packs every tarball before it publishes any, and signs them with provenance. workflow_dispatch runs all of that and publishes nothing, so the first real tag is not also the first rehearsal.
Someone arriving from npm wants the install line, the one idea, the packages and a picture. Building the workspace, running the playgrounds, serving the Jekyll site and cutting a release are all real, and none of them are that - they move to DEVELOPER.md, along with the erasable-syntax argument, whose count had drifted to fourteen across twelve files and whose tsconfig flag is not actually set. The README keeps the claim it can make: no dependencies. Two things that were simply wrong: the documents guide put the JSON adapter in `@textui/core/adapters`, which is a deliberately empty placeholder that says so itself - it is in `@textui/documents`. And six package READMEs linked to siblings as `../core`, which resolves in the repository and 404s on npmjs.com.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Everything needed before the first npm publish. The codebase itself needed no work — build, typecheck, lint, docs and 1692 tests were green before this branch, with zero
TODO/FIXMEinpackages/*/src. What was missing was publish mechanics.One real bug
@textui/coredeclared an./hooksexport subpath pointing atdist/hooks/, which tsc never emits — hooks live inruntime/hooks.tsand are already re-exported from the root. Nothing in the repo imported that subpath, which is exactly why every check stayed green. It would have beenERR_MODULE_NOT_FOUNDfor the first consumer who tried it.scripts/check-exports.mjsnow runs in CI on every PR so the next one fails a review instead of a release. It was verified against the bug it exists for: reintroducing./hooksmakes it exit 1 and name both targets.What ships
Six, in dependency order —
pnpm publish -rworks this out itself:@textui/documents,@textui/textideand@textui/textide-gitare markedprivate. They build in CI and stay in the repo, but an IDE should not be tied to the runtime's release cadence. Being private is the only thing holding them back — a publish-time filter you have to remember is one that gets forgotten.Also in here
"license": "MIT"is a label, not the text, and npm only picks up a LICENSE in the package's own directory.RELEASING.md,CHANGELOG.md, and a tag-triggeredrelease.ymlthat reruns the full gate, checks the tag against the manifests, packs every tarball, then publishes with provenance.workflow_dispatchruns all of it and publishes nothing, so the first real tag is not also the first rehearsal.DEVELOPER.md. That argument's count had drifted to fourteen across twelve files and its tsconfig flag is not actually set, so it now reads as a direction rather than a property.@textui/core/adapters(a deliberately empty placeholder); it is in@textui/documents. Six package READMEs linked to siblings as../core, which resolves here and 404s on npmjs.com.Still needed before a tag
The
@textuiscope is reserved, but nothing can publish yet: there is noNPM_TOKENrepository secret, and npm is not authenticated on the machine this was prepared on.RELEASING.mdcovers it — it needs an automation token (a classic token with 2FA on publish cannot be used unattended).Gate on this branch: build, typecheck, lint, docs, exports guard, version check and all 1692 tests pass.