Description

JosunLP

opened on Sep 18, 2026

Issue body actions

Summary

Nothing verifies the exports map, type resolution under each module-resolution mode, or the contents of the published tarball. The exports map has 24 entries and is edited by hand on every new module.

Proposal

Add publint and @arethetypeswrong/cli to the release job, and to the quality job from #212, running against a fresh bun run build. Both are single-command tools.

Why

Both would have surfaced #219 automatically — subpath entries resolve for ESM but not for require, and types is ordered after import in every subpath entry.

Notes

Effort: under a day · Risk: low.

Roadmap item C7; see the audit in #209.

Activity

1. JosunLP

   added

   buildChanges to the build and meta filesChanges to the build and meta files

   on Sep 18, 2026
2. JosunLP

   added 5 commits that reference this issue on Sep 20, 2026

   build: drop the attw ignore and close three gaps in the declaration r…

   f4360d3

   build: close the silent-skip paths in the bundle-size check

   de97328

   Merge branch 'dev' into feat/216-bundle-budgets

   6074a54

   build: validate the published package with publint and are-the-types-…

   8e24663

   Merge origin/dev into feat/216-bundle-budgets

   9b8bd92
3. JosunLP

   closed this as completedon Sep 21, 2026
4. JosunLP commented on Sep 21, 2026

   

   JosunLP

   on Sep 21, 2026 – with Claude

   CollaboratorAuthor

   More actions

   Done in #236, merged to dev as 8e24663.

   Your prediction held: publint reported all 23 types-ordering errors from #219 verbatim. It also found a third problem the audit had missed, which #236 fixed as well — 315 × InternalResolutionError: every emitted .d.ts used extensionless relative imports, a hard TS2834 for any consumer with skipLibCheck: false.

   Measured against a real consumer (moduleResolution: nodenext, skipLibCheck: false):

   before: dist/index.d.ts(10,15): error TS2834 ... (×12 shown, tsc exit 2)
   after:  tsc exit 0
   

   Fixed by post-processing the declarations (scripts/postbuild-types.mjs) rather than adding .js extensions across src/ — only the declarations need it, and the rewrite only appends an extension when the target actually exists in dist/.

   A correction I owe this thread. An earlier revision of #236 claimed a .d.cts twin produced green tooling over broken types, and left false-esm ignored on that basis. That conclusion came from testing a single copied declaration, which does push TS1479 one level down. A full .d.cts tree with every specifier resolved to .cjs type-checks cleanly — 0 errors with and without skipLibCheck. That tree shipped in #234, so --ignore-rules false-esm is gone and publint now reports "All good!".

   check:publish runs in the PR gate, the publish workflow and prepublishOnly.

   Closing on merge to dev; it ships with the next release to main.

   ---

   Generated by Claude Code
5. JosunLP

   added 3 commits that reference this issue on Sep 21, 2026

   build: validate the published package with publint and are-the-types-…

   This issue will close once commit 656e795 is merged into the 'main' branch.

   656e795

   release: 1.17.0

   This issue will close once commit f4e7a86 is merged into the 'main' branch.

   f4e7a86

   release: 1.17.0 (#245)

   This issue will close once commit 71a7d16 is merged into the 'main' branch.

   71a7d16

Issue actions