7b22a8584f
The runtime probe waits on child exit with stdout piped and undrained, so a CLI printing more than the pipe buffer holds blocks mid-write and the probe times out. It imports the whole backend, which is exactly what emits stray stdout, so a healthy install gets reported broken and sent to repair after a 10s stall on every launch. Drain concurrently instead, here and in the desktop-capabilities probe next to it. The install marker was cleared only on success, so a failed update, or a cancelled elevation prompt, left it behind and pinned a working install into repair it may not be able to finish offline. Clear it on every terminal outcome and treat marker bookkeeping as best effort rather than failing the install over it. Skip pip flag lines in studio.txt: an unparseable line reported the install broken, and repair reinstalls the same file, so that never cleared. Allow prereleases against a floor for the same reason. |
||
|---|---|---|
| .. | ||
| capabilities | ||
| icons | ||
| linux | ||
| src | ||
| windows | ||
| build.rs | ||
| Cargo.lock | ||
| Cargo.toml | ||
| Entitlements.plist | ||
| Info.plist | ||
| tauri.conf.json | ||
| tauri.macos.conf.json | ||
| tauri.windows.conf.json | ||