fix(plan): forward LOCALAPPDATA so Node's compile cache stays outside the workspace#389
Merged
Merged
Conversation
Member
Author
This stack of pull requests is managed by Graphite. Learn more about stacking. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9818095e25
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
6f74c19 to
914d2e7
Compare
A small Node script turns on Node's compile cache and imports a few sibling modules so the cache directory gets some files. Two `vt run --cache build` calls expect a cache miss followed by a cache hit — that's how it should look when the cache lives outside the workspace. The test passes on Linux and macOS but is expected to fail on Windows CI. The e2e harness clears the spawned process env for snapshot determinism, so on Windows the task ends up without `LOCALAPPDATA`, Node falls back to a workspace-relative cache path, and the runner refuses to cache the run. The next commit makes both this fixture and real-world Windows tasks pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
… the workspace `DEFAULT_UNTRACKED_ENV` already forwards most Windows env vars to cached tasks, but `LOCALAPPDATA` was missing. Node's compile cache uses it to decide where to put cache files on Windows. Without it, the cache lands inside the workspace, the runner sees the same files both written and read in one run, and refuses to cache anything with `not cached because it modified its input`. Add `LOCALAPPDATA` to the list, and mirror it in the e2e harness so the previous commit's Node compile-cache fixture goes from failing on Windows to passing on every platform. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
914d2e7 to
c96a35c
Compare
fengmk2
added a commit
to voidzero-dev/vite-plus
that referenced
this pull request
May 19, 2026
Release vite-plus v0.1.22: Security Patch, Parallel Global Install & Scaffold Polish A critical Vitest browser-mode security fix, parallel `vp add -g` installs, a built-in oxlint rule to prefer `vite-plus` imports, and a new `--git` switch for `vp create`. ### Highlights - **Security**: bundled `vitest` bumped to `4.1.6` to address [GHSA-2h32-95rg-cppp](GHSA-2h32-95rg-cppp) (Critical, CVSS 9.6), an XSS to RCE chain via the `otelCarrier` query parameter in Vitest browser mode ([#1633](#1633)) - **Parallel global install**: `vp add/install/update -g` now installs packages concurrently with a progress bar and a `--concurrency` flag (default 5) ([#1597](#1597)) - **Prefer vite-plus imports**: new bundled oxlint rule rewrites `vite`/`vitest` imports to `vite-plus`, enabled by default in generated and migrated `lint` configs ([#1408](#1408)) - **Git init on scaffold**: `vp create` learns `--git`/`--no-git` (interactive prompt; auto-commits "Initial commit from Vite+") ([#1484](#1484)) ### Features - Spawn npm for global installation in parallel with a progress bar and a `--concurrency` option ([#1597](#1597)), by @liangmiQwQ - Add bundled oxlint rule to prefer `vite-plus` imports over `vite`/`vitest` ([#1408](#1408)), by @Han5991 - `vp create`: initialize a git repository and create an initial commit on scaffold ([#1484](#1484)), by @ryohidaka - `vp create`: rename underscore-prefixed files (`_gitignore`, `_npmrc`, `_yarnrc.yml`) to dotfiles for `@org/create` bundled templates ([#1574](#1574)), by @jong-kyung - Add `VP_PR_VERSION` env var to install unreleased PR builds via pkg.pr.new ([#1578](#1578)), by @fengmk2 ### Fixes & Enhancements - Skip merging standalone `.oxfmtrc`/`.oxlintrc` config when the `fmt:`/`lint:` key is already declared in `vite.config.ts` (fixes duplicate-block regression in `vp create fate`) ([#1601](#1601)), by @fengmk2 - Suppress the `VITE+ - The Unified Toolchain for the Web` banner for `vp lint --lsp`, `vp fmt --lsp`, and `vp fmt --stdin-filepath` so stdout stays a pure LSP / formatter stream ([#1619](#1619)), by @fengmk2 - `vp create`: detect output directory when running in the current directory ([#1606](#1606)), by @jong-kyung - `vp update -g`: skip installs when the recorded global package version already matches the npm-resolved version, and tolerate string/array outputs from `npm view ... version --json` ([#1596](#1596)), by @leno23 - `vp create`: preserve single-segment project path in `updateWorkspaceConfig` ([#1582](#1582)), by @jong-kyung - `vp env use`: keep the change session-scoped on Windows ([#1577](#1577)), by @fengmk2 - `vp rebuild`: accept positional package names ([#1564](#1564)), by @fengmk2 - Adopt the new vite-task error formatter; errors now print as `error: <top-level>` plus `* <source>` chain lines, with bold-red highlight on a TTY ([vite-task#390](voidzero-dev/vite-task#390)), by @branchseer - vite-task: forward `LOCALAPPDATA` so Node's compile cache stays outside the workspace on Windows ([vite-task#389](voidzero-dev/vite-task#389)), by @branchseer - Bump vite-task to `c945cc0` ([#1628](#1628)), by @branchseer ### Refactor - Revert `vp pm plugin` command (per discussion in #1038) ([#1623](#1623)), by @jong-kyung ### Docs - Add `vitepress-plugin-llms` to the docs site so the published docs include LLM-friendly outputs (`/llms.txt`) ([#1625](#1625)), by @jong-kyung - Refresh home stats for oxlint, vite, and vitest ([#1512](#1512)), by @nozomee - Mention `vp env doctor` in agent instructions ([#1603](#1603)), by @leno23 ### Chore - Consolidate the upstream build chain into a single `pnpm build` script (justfile recipe now just calls `pnpm build`) ([#1626](#1626)), by @fengmk2 - Fix bootstrap-cli on Windows ([#1583](#1583)), by @fengmk2 - Refresh trusted stack stats ([#1573](#1573), [#1616](#1616)), by @voidzero-guard[bot] - Update GitHub Actions ([#1611](#1611), [#1612](#1612)), by @renovate[bot] - Address zizmor findings in composite actions and the release workflow; drop unused `actions-cool/issues-helper` ([#1630](#1630)), by @Boshen - Switch plain checkouts to `taiki-e/checkout-action` ([#1620](#1620)), by @Boshen - Switch release to a version-bump PR + push trigger flow ([#1575](#1575)), by @Boshen - Gate release publish on environment approval with a Discord notice ([#1571](#1571)), by @Boshen - Enable `cargo clippy` with `-D warnings` ([#1579](#1579)), by @Boshen - Drop unused `setup-node` from the version-check job ([#1600](#1600)), by @fengmk2 - Add Void deploy workflows for the docs site ([#1590](#1590)), by @fengmk2 - Add `--help` case to config snap tests for npm10/yarn1/yarn4 ([#1585](#1585)), by @jong-kyung - Add `--help` case to publish snap tests for npm10/yarn1/yarn4 ([#1584](#1584)), by @jong-kyung - Verify `.gitignore` and `.yarnrc.yml` in the new-vite-monorepo snap ([#1576](#1576)), by @jong-kyung - vite-task: bump pnpm to `11.1.2` ([vite-task#383](voidzero-dev/vite-task#383)), by @branchseer - vite-task: update lint-staged to v17 ([vite-task#385](voidzero-dev/vite-task#385)), by @renovate[bot] ### Bundled Versions | Tool | Version | Source | | --- | --- | --- | | vite | `8.0.11` | [`66f3194`](vitejs/vite@66f3194) | | rolldown | `1.0.0` | [`ac5c710`](rolldown/rolldown@ac5c710) | | tsdown | `0.22.0` | [npm](https://npmx.dev/package/tsdown/v/0.22.0) | | vitest | `4.1.6` | [npm](https://npmx.dev/package/vitest/v/4.1.6) | | oxlint | `1.63.0` | [npm](https://npmx.dev/package/oxlint/v/1.63.0) | | oxlint-tsgolint | `0.22.1` | [npm](https://npmx.dev/package/oxlint-tsgolint/v/0.22.1) | | oxfmt | `0.48.0` | [npm](https://npmx.dev/package/oxfmt/v/0.48.0) | ### New Contributors Welcome to all new contributors! 🎉 @nozomee, @ryohidaka, @leno23 **Full Changelog**: v0.1.21...v0.1.22 --- Merging this PR will trigger the release workflow. --------- Co-authored-by: voidzero-guard[bot] <278573678+voidzero-guard[bot]@users.noreply.github.com> Co-authored-by: MK <fengmk2@gmail.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
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's wrong
On Windows, any Node ≥ 22 task that uses the built-in compile cache (which Vite, esbuild's CJS loader, Storybook, and anything that touches
node:moduleopts into) is currently uncacheable undervt run --cache. Node picks the cache directory fromLOCALAPPDATA, but the planner doesn't forward that variable to cached tasks, so Node falls back to a workspace-relative path. The cache files end up next to the source, the runner sees the same paths both written and read in one run, and refuses to cache it withnot cached because it modified its input.What this PR does
Adds
LOCALAPPDATAto the planner's default env-passthrough list. WithLOCALAPPDATApopulated, Node finds the realC:\Users\<user>\AppData\Localand the compile cache lands outside the workspace, where the runner ignores it — so cached runs work the way they're supposed to.The PR has two commits to make the bug easy to see:
First commit — Adds a small e2e fixture (
node_compile_cache_outside_workspace) that runs a Node script with the compile cache turned on. On Linux and macOS the test passes. On Windows CI it fails with the exact "modified its input" symptom. (Windows-only failure run)Second commit — Adds
LOCALAPPDATAto the planner's default list and mirrors it in the e2e harness (which clears env for snapshot determinism, so it has to be told which vars to forward). All platforms green. (Final green run)Test plan