v1.3.0 — 1 GB staging leak fix + crash-safe staging

Fixes a 1 GB storage leak and a soft-reboot crash caused by deleting staging while it was still mounted.

Fixes:

• Staging leak (1 GB per install). Staged vendor directories were written as siblings of .staging/ (.staging_vendor_lib64) instead of inside it, so every cleanup path missed them. Each apply leaked a full copy of /vendor/lib64 — 1.0 GB — into /data permanently. Staged dirs now live in .staging/dir_*, and all cleanup paths also sweep the legacy .staging_* form. Upgrading frees the leaked space on the next apply.
• Staging cost cut ~25x. A payload lib with a new filename in /vendor/lib64 forces a merged copy of that whole directory. New names are now placed in /vendor/lib64/egl, which the linker's sphal namespace searches too (verified against /linkerconfig/ld.config.txt): staging drops from ~1 GB to ~43 MB with identical resolution. Files that replace a stock lib keep their exact path (per-file bind, no staging).
• Crash-safe staging deletion. Deleting a staging dir that is still the live mount source for /vendor/lib64 empties the mounted directory; vendor HALs then SIGSEGV on dlopen and the system soft-reboots. gm_stage_clean now refuses to delete while any of staging is mounted in init/zygote, driver-select no longer deletes staging synchronously, and a new stage-clean subcommand exposes the guarded cleanup manually.

Also:

• Reject saved plans from older versions (their staged dirs point at the leaked legacy paths) and rebuild automatically
• doctor reports the staged directory size; the engine warns above 256 MB
• Never rm -rf the module dir or .staging* while a driver is active — use driver-clear, action.sh unmount, or reboot first (mounts are in-memory, so a reboot always detaches them)

Verified on-device (Galaxy S23, SM-S911U1, Adreno 740): module size 1.0 GB → 120 MB, staging 43 MB, driver served in init/zygote/surfaceflinger namespaces, VulkanCapsViewer enumerates the custom driver, no restarts over repeated observation windows. Test suites: 40 + 20 + 74 assertions pass.

v1.2.0 — Staging corruption fix

Fixes a crash caused by two engine instances building staging simultaneously (boot service + async driver-select race).

Fixes:

• Force-overwrite in staging copy (cp -af) — no more partial copies from concurrent builds
• Staging validity check — corrupt staging is detected by entry-count comparison and auto-rebuilt
• Status function loop bug fixed (ACTIVE detection was broken by a missing fi)

If you experienced a system crash or soft restart after rebooting with a driver applied, this fixes it.

v1.1.0 — Settings UI, ALL_NS removal

Config toggles now in the WebUI. ALL_NS removed (niche and confusing — the default init/zygote/SF set covers all apps started after apply).

Changes:

• Settings card in the WebUI: toggle Restart surfaceflinger without editing config files
• ALL_NS option removed entirely (engine, config, WebUI)
• driver-clear now kills pending background jobs before clearing (fixes race when selecting a driver then immediately going back to stock)
• 69 backend test assertions

Quick start:

1. Install from releases
2. Download a driver from Adreno-Tools-Drivers
3. Push to /sdcard/Download, open the WebUI, tap Use
4. After each reboot: WebUI → Apply

v1.0.0 — First release

Runtime GPU driver swapping via bind mounts. Works with late-loaded KernelSU — no OverlayFS needed.

Features:

• Driver-zip management via WebUI (Adreno-Tools-Drivers format)
• Global mode (all apps) and per-app mode
• Auto-detects Vulkan loader targets (hw module + ICD manifests)
• Auto-fixes file permissions and SELinux labels
• SELinux-safe (relabels + sepolicy.rule fallback)
• Self-diagnosing doctor command
• 111 test assertions

Quick start:

1. Install the module zip from KernelSU/APatch manager
2. Download a driver from Adreno-Tools-Drivers
3. Push to /sdcard/Download, open the WebUI, tap Use
4. After each reboot: WebUI → Apply