fix: avoid locked exe on forced upgrade

[fengmk2]

Summary

• Install forced reinstalls of the active version into a unique semver build-metadata directory
• Repoint current after the install completes so Windows does not overwrite the running vp.exe
• Add focused unit coverage for forced reinstall directory selection

Directory structure

Before this change, vp upgrade --force reused the active version directory when the target version matched the current version:

~/.vite-plus/
├── current -> 0.1.23
└── 0.1.23/
    └── bin/
        └── vp.exe   # currently running on Windows

That meant the forced reinstall extracted the downloaded platform package back into 0.1.23/bin/vp.exe, which fails on Windows because the running executable is locked.

After this change, a forced reinstall of the active version writes into a new semver-compatible build-metadata directory, then repoints current only after installation succeeds:

~/.vite-plus/
├── current -> 0.1.23+force.<pid>.<timestamp>
├── 0.1.23/
│   └── bin/
│       └── vp.exe   # previous active install remains untouched
└── 0.1.23+force.<pid>.<timestamp>/
    └── bin/
        └── vp.exe   # freshly installed binary

Normal upgrades to a different version still use the plain target version directory, for example 0.1.24/.

Verification

• Reproduced the original Windows vp upgrade --force failure with os error 32
• cargo build -p vite_global_cli
• Temp VP_HOME end-to-end locked-executable test: upgrade --force succeeds
• cargo test -p vite_global_cli commands::upgrade::tests
• cargo fmt --check

---

Note

Medium Risk
Touches the global CLI upgrade/install swap path and directory cleanup; wrong naming or cleanup could break upgrades or rollbacks, but behavior is scoped to forced same-version reinstalls with new unit and Windows E2E coverage.

Overview
Fixes Windows vp upgrade --force when the target version is already active: the upgrade no longer extracts into the running install folder (which locks vp.exe and caused os error 32).

Upgrade flow now picks an install directory name via target_install_dir_name. For a forced reinstall of the active version, it uses a unique semver build-metadata path like {version}+force.{pid}.{nanos}; otherwise it still uses the plain version folder. current, rollback metadata, and old-version cleanup follow that directory name (not only the semver string).

CI adds a Windows-only E2E step that simulates a locked same-version install under a temp VP_HOME and asserts current moves to a +force. dir, the original version tree stays, and .previous-version is correct.

Tests cover forced-dir selection in upgrade and that cleanup_old_versions treats +force dirs like other semver installs (removable when old, kept when protected).

^{Reviewed by Cursor Bugbot for commit cc5393e. Configure here.}

[netlify]

✅ Deploy Preview for viteplus-preview canceled.

Name	Link
🔨 Latest commit	8d23be7
🔍 Latest deploy log	https://app.netlify.com/projects/viteplus-preview/deploys/6a1ce12d5c178d0008b87f4e

[fengmk2]

@cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit cc5393e. Configure here.}