build(scripts): fold the root package.json scripts into runners and derived lists - #688
Merged
Merged
Conversation
…erived lists 135 scripts became 103. The 23 `perf:*` scripts, which differed only in the harness file behind a shared 90-character Node prefix, are now one runner (`pnpm perf <name>`, `pnpm perf` lists them); `perf:smoke` stays as the sync gate's entry point. The pure aliases (`test:browser:*:chromium`, the five `lint:*:fix` variants, `verify:ci`, `site:build:api`) are gone: scripts forward their arguments, so `pnpm lint --fix` is the variant. `lint` and `lint:packages` are one eslint invocation. `bench` runs the cross-library benchmarks at the root as it does inside the package; `perf:baseline` was its duplicate. The stale `release` script and `scripts/release.ts` (a superseded flow calling a script that no longer exists) are removed. The hand-written package lists are derived: `typecheck:packages` and `build:packages` select by workspace directory, `test` selects vitest projects by the `exojs*` wildcard, and the new `verify:publish` packs and publints every package in `lockstep-packages.ts` - which showed that aseprite, ldtk, react and create-exo-app had never been publinted. CI's extension build step uses `build:packages` instead of its own list. The parity tests follow: the vitest project guard applies vitest's wildcard matching, and a new test keeps the perf runner's table honest.
Exoridus
enabled auto-merge (squash)
September 6, 2026 07:31
Bundle ReportChanges will increase total bundle size by 18.88MB (59.43%) ⬆️
Affected Assets, Files, and Routes:view changes for bundle: site-server-esmAssets Changed:
App Routes Affected:
|
Exoridus
added a commit
that referenced
this pull request
Sep 6, 2026
…f heap (#698) ## Summary Folding the package lint into the engine's invocation (#688) put one type-aware program for the engine, its tests, the examples, the scripts and every package into a single heap. That reached V8's 4 GB ceiling and failed the push gate with an out-of-memory crash rather than a lint error - intermittently, because whether it tipped depended on what the machine had just done. A `build:all` beforehand was enough. `scripts/lint.ts` runs the same globs as two passes, one process each, and forwards its arguments to both, so `pnpm lint --fix` and `lint:fast`'s `--cache` still behave as one invocation. Nothing is linted twice, and each pass's memory is returned when its process exits. ## Why not raise the ceiling Node caps the default old space at about 4.3 GB on 64-bit regardless of installed RAM, so a 64 GB workstation and a 16 GB CI runner hit the same wall. Passing `--max-old-space-size=8192` would clear it on both, and would move the failure onto whichever machine has the least memory: past physical RAM, V8 no longer fails with a clean heap error, the operating system swaps or kills the process. For a gate that has to run on other people's machines, less work under the ceiling beats a higher ceiling - and the ceiling is also what reported the regression in the first place. ## Verification - `pnpm lint:all` green immediately after `pnpm build:all`, which is the sequence that crashed - `test/ci/` green (269 tests), `typecheck:scripts` clean - Pre-push ran `verify:quick` plus every lane the range selects Co-authored-by: Exoridus <github@codexo.de>
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.
Summary
The root
package.jsonhad 135 scripts across 293 lines. This folds them into 103 scripts, with runners and derived lists where the duplication was:pnpm perf <name>replaces the 23perf:*scripts, which differed only in the harness file behind a shared 90-character Node prefix.pnpm perflists the runs; arguments are forwarded (pnpm perf renderers:alloc:cell --id "mesh/1000").perf:smokestays as thesyncgate's entry point and ispnpm perf smoke.test:browser:*:chromium, the fivelint:*:fixvariants,verify:ci,site:build:api,perf:baseline. Scripts forward their arguments, sopnpm lint --fixandpnpm test --coverageare the variants (CONTRIBUTING says so now).lintandlint:packagesare one eslint invocation.typecheck:packagesandbuild:packagesselect by workspace directory (./packages/**);testselects vitest projects by theexojs*wildcard; the newverify:publishpacks and publints every package fromscripts/release/lockstep-packages.ts. That last one showed that aseprite, ldtk, react and create-exo-app had never been publinted (all pass). CI's extension build step now usesbuild:packagesinstead of its own copy of the list.benchforwards to the cross-library harness inpackages/exojs-bench, so the reproduce instructions work from either directory; the engine's own micro-benchmarks remainbench:micro.releasescript andscripts/release.ts(a superseded flow calling averify:releasethat no longer exists) are removed.The parity tests follow: the vitest project guard now applies vitest's own wildcard matching (several
!patterns are OR-ed by vitest, which is why the selection is a positive wildcard), andtest/ci/perf-runs.test.tskeeps the runner's table honest.Verification
test/ci/*(13 files) green;typecheck:scripts, fulllint,format:checkgreenpnpm perf:smoke,pnpm verify:publish(14 packages),pnpm typecheck:packages,pnpm perf renderers:alloc:cell --id "mesh/1000"run through the new scriptsvitest listconfirms--project="exojs*" --project=rendering-perfselects exactly the 16 projects the explicit list namedpnpm bootstrap:devend to end, thenpnpm doctorreports the clone ready and the CI parity tests pass therehttps://claude.ai/code/session_01CyAQuoGJMBAZuDiWyPeGfW