Repository navigation
fix(update-local): Node installs on German Windows; the double-click asks, then updates - #237
Merged
Merged
Conversation
…asks, then updates
- The Node.js step never ran on the owner's PC. Windows PowerShell 5.1 failed on a
piped [void] cast ("The argument type cannot be System.Void"), and past that the
staging folder's ACL named 'Administrators' and 'SYSTEM', which a German Windows
calls otherwise ("identity references could not be translated"). New-PrivateDirectory
(tools/update-local-lib.ps1) grants by SID; a Windows test runs it in PowerShell 5.1
and checks the ACL. Both old forms turn it red. Node 26.7 -> 26.10 installed with it.
- update-local.cmd only reported, so a double-click updated nothing. The check now exits
2 when something is due, and the .cmd asks "Update these now?": Y runs -Apply.
Tested with a stand-in script: 0, 1, 2+N, 2+Y.
- pnpm install and fsaunpack's npm ci were offered on every check, so it never came
back clean. Now only when the lockfile changed after the last install
(Test-ChangedSince against the install markers).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…on, no installer left behind - #167: getJson's host guard never refused anything: `origin + pathname.startsWith(base)` concatenated a boolean, so every URL passed. isPinnedUrl now allows only nodejs.org's index.json and package documents on registry.npmjs.org; nine URL cases, six of which the old expression lets through. - #243: the Node version from nodejs.org's index becomes a URL and a file name. newestNode takes only the exact "v1.2.3" shape, and update-local.ps1 throws on anything else. An index naming "v26.11.0&calc" returned it before; now that entry is skipped. - #244: a thrown check, a declined admin prompt or a failed install left the ~30 MB installer in %ProgramData%. Invoke-InPrivateDirectory removes the folder in a finally; a Windows test shows it gone after a throw (it stayed without the finally). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nawid3333
pushed a commit
that referenced
this pull request
Oct 3, 2026
Brings in the other code-review PRs (#237, #300, #302-#313). Conflicts: - tools/sync-vendor.mjs: main's unappliedPatches check (#176) moves into this branch's main(), so importing the module for its tests runs nothing. - types/messages.d.ts: this branch's `isExt: true`, main's parentFrameId doc. - options/index.html: main's sectioned layout, with this branch's translated "Learn more" link and without the upstream feedback form. - docs/vendored-libraries.md: main's knob ref and vtt.js change list, this branch's googlevideo row removed and pointer to CHANGES in verify-vtt.mjs. - native-host/README.md: install.ps1 is about 130 lines now. - CLAUDE.md: main's sync-upstream paragraph (#163), which already says the PR waits for the owner. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JzfFWJzzaUDto87UaqoJGy
Nawid3333
pushed a commit
that referenced
this pull request
Oct 3, 2026
Brings in the other code-review PRs (#237, #300, #302-#313). Conflicts, each kept from both sides: - storage.e2e.mjs: main's per-identifier loop, plus this branch's usedOPFS check. - subtitle-search.e2e.mjs: main's OpenSubtitles DOWNLOAD_LINK, plus this branch's wait timeouts. - TabTracker.test.mjs: main's and this branch's new describe blocks. - MpvNativeHost.test.mjs: main's imports (a superset). - MpvNativeHostReadMessage.property.test.mjs: main's version of the over-size test, which matches #303's host (an over-size length now waits for the body and answers TooLarge, null only when the input ends). And one test that merged cleanly but no longer held: mpvHostSecurity's "refuses a message one byte over the limit" now expects TooLarge, the answer #303 gives once the message has been read. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JzfFWJzzaUDto87UaqoJGy
11 tasks
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.
Updating your PC with
update-localturned up three problems in the script. This fixes all three. Nothing here ships, so merging it makes no release.1. The Node.js update never worked on your PC. It failed twice over:
update-local.cmduses, can't run one of its lines (a[void]cast piped on). That was the "Der Argumenttyp kann nicht System.Void sein" error.That part now lives in
tools/update-local-lib.ps1and uses the groups' fixed IDs, which are the same in every language. A Windows test runs it in PowerShell 5.1 and checks that only you, Administrators and SYSTEM have access. Both old versions make that test fail. With the fix, Node 26.7 → 26.10 installed on your PC.2. Double-clicking
update-local.cmdupdated nothing. It only reported what was out of date. Now, when something is out of date, it asks "Update these now? [Y,N]", and Y updates everything it listed. When everything is current it doesn't ask. A Windows test drives the.cmdwith a stand-in script: nothing to do, a failed step, N, Y. The old.cmdfails the N and Y cases.3. The check never came back clean. It offered
pnpm installand fsaunpack'snpm cievery time. Now it offers each one only when its lockfile changed after the last install. A test covers this, and on your PC the check is now fully clean.Checks: lint, typecheck, and 973 unit tests pass. The new Windows tests run in CI's Windows job. The check exits with 0 on your up-to-date PC and with 2 after the lockfile was made to look changed.
4. Three more fixes in the same script, from the code-review issues:
26.10.0before it goes into a download address or file name. A test shows that an odd entry used to be taken and is now skipped.C:\ProgramData\FastStream. Now that folder is removed however the step ends, and a Windows test proves it.Fixes #167
Fixes #243
Fixes #244
🤖 Generated with Claude Code