chore(types): derive the event-loop histogram type from monitorEventLoopDelay - #874
Merged
Conversation
…oopDelay @types/node 26.5.0 renames IntervalHistogram to ELDHistogram. The two versions share neither name, so importing either one pins the repo to one side of that bump and breaks Lint and Build on the other. Deriving the type with ReturnType tracks whatever the installed @types/node calls it. Unblocks #853.
RonenMars
force-pushed
the
fix/interval-histogram-type
branch
from
September 12, 2026 01:03
5461fdb to
d5a4a9e
Compare
Owner
Author
|
🎉 This PR is included in version 1.90.0 🎉 The release is available on: Your semantic-release bot 📦🚀 |
RonenMars
added a commit
that referenced
this pull request
Sep 12, 2026
…ndency notes [skip-ci] A @types/* type imported by name couples the repo to one version of that package. @types/node 26.5.0 renamed IntervalHistogram to ELDHistogram with neither version declaring both, which failed Lint and Build on TS2305 while runtime behaviour was unchanged (#853, fixed by #874). Deriving the type from the value is the fix and the general rule. vitest 5.0.0 (#855) landed green on every check, which is strong evidence because CI runs the suite under the new runner — but not proof over config resolution, reporters, worker pooling or timeouts. AGENTS.md's "Dependencies and build" section described vendor/scanner and vendor/agent-types as submodules with file: deps and claimed recursive CI checkouts. None of that is true: menubar is the only submodule, scanner and agent-types are published @threadbase-sh npm packages, and every build/test checkout sets submodules: false. It also understated the externals and the migration copies — the CLI externalizes better-sqlite3 as well as node-pty, and the build copies three migration directories, not two. CLAUDE.md carried the same externals error and is corrected with it. Pointers from CLAUDE.md, AGENTS.md and a new .cursor/rules/dependency-bumps.mdc so claude-code, codex and cursor all reach it.
RonenMars
added a commit
that referenced
this pull request
Sep 12, 2026
…ndency notes [skip-ci] (#877) Two lessons from the dependabot batch that landed today, plus the stale notes they exposed. ## 1. A `@types/*` type imported by name is a version coupling `@types/node` 26.5.0 renamed the `perf_hooks` histogram `IntervalHistogram` to `ELDHistogram`, and **neither version declares both names**. So #853 failed `Lint` and `Build` on `TS2305: … has no exported member` while nothing was wrong at runtime. Renaming the import moves the break to the other side of the bump and leaves the same trap for the next rename. Deriving the type from the value that produces it — `ReturnType<typeof monitorEventLoopDelay>` — tracks whatever the installed types call it (#874). Recorded with the verify-both-versions procedure, the revert-to-red check, and the sequencing constraint: the derived-type fix lands on `main` first, then the dependabot PR rebases green. ## 2. A green suite does not fully clear a major test-runner bump vitest 5.0.0 (#855) passed every check across two independent CI cycles. That is strong evidence, because CI runs the suite *under* the new runner — but it covers only what the suite exercises, not config resolution, reporters, worker pooling or default timeouts. Recorded as "suspect the runner first when tooling output turns odd". ## 3. `AGENTS.md` "Dependencies and build" was substantially wrong Found while adding the above. Every claim below was checked against the tree, not copied from `CLAUDE.md`: | Claim | Reality | |---|---| | `vendor/scanner`, `vendor/agent-types`, `vendor/menubar` are submodules | `.gitmodules` has only `vendor/menubar` | | CI checkouts "remain recursive" | every `ci.yml`/`release.yml` checkout sets `submodules: false`; only `update-menubar.yml` is recursive | | `@threadbase/scanner` are `file:` deps built by `postinstall` | `@threadbase-sh/scanner@^0.17.2` and `@threadbase-sh/agent-types@^1.0.0`, normal npm semver deps — the scope was wrong too | | "only `node-pty` is external" for the CLI | the CLI entry externalizes `node-pty` **and `better-sqlite3`**; the library entry externalizes `node-pty` and `pg` | | "both SQLite and Postgres migration copies" | the build copies **three** dirs: `migrations/`, `runtime-migrations/`, `pg-migrations/` | | preserve `prepare` / `postinstall` | correct — kept, and noted that `prepare` also sets `core.hooksPath` | `CLAUDE.md` carried the same CLI-externals error and is corrected with it. ## Reaching all three tools Canonical text lives in `docs/troubleshooting.md`. `CLAUDE.md` and `AGENTS.md` point at it; Cursor reads `AGENTS.md` natively and additionally gets `.cursor/rules/dependency-bumps.mdc`, scoped by `globs` to TS and dependency files rather than `alwaysApply`. The `.mdc` is a short form plus a pointer, not a third copy — three hand-maintained copies of one paragraph is how `AGENTS.md` drifted in the first place. Adding it falsified `docs/agents/tooling.md`'s "Cursor … read none of the files above", which is updated here. ## Verified state - `npx vitest --version` → `vitest/5.0.0 darwin-arm64 node-v24.15.0`, confirming the runner claim. - `@types/node` on `main` is `26.5.0` (#853 merged). - `../../docs/troubleshooting.md` resolves from `.cursor/rules/`; `.cursor/` is not gitignored. - Code fences balanced; `biome check` clean. Checked 2026-09-12 against `32a7977d`. Docs-only, hence `[skip-ci]`.
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.
What is wrong
@types/node26.5.0 renames theperf_hooksinterfaceIntervalHistogramtoELDHistogram.src/services/host-pressure/hostPressure.tsimports it by name, so the bump in #853 fails both Lint and Build:Why not just rename it
The two versions share neither name — 26.4.0 has only
IntervalHistogram, 26.5.0 has onlyELDHistogram.Swapping one for the other moves the breakage to the other side of the bump rather than removing it, and leaves the same trap for the next rename.
The fix
Derive the type from the function that produces it:
That is whatever the installed
@types/nodecalls it, so no future rename can break it.There was one annotation site (
HostPressureMonitorOpts.histogram); themonitorEventLoopDelaycall site needed no change.Type-only — erased at runtime, no behaviour change.
Verified state
tscagainst@types/node26.4.0 (current pin)tscagainst@types/node26.5.0 (the bump)tscagainst 26.5.0 with this change revertedTS2305abovebiome check__tests__/host-pressure.test.tsChecked 2026-09-12 against
acc7d3fa.Unblocks #853.