# Validation on the server and physical display

These checks require your Unraid server. Only the user-confirmed checks listed in [TEST-REPORT.md](TEST-REPORT.md) have hardware confirmation. Never run manual TRCC alongside the plugin.

1. Record the environment before installation:

   ```bash
   cat /etc/unraid-version
   uname -m
   lsusb -d 0416:8001
   docker version
   docker ps -a --filter name=thermalright-screen
   docker images thermalright-screen
   test ! -d /mnt/user/appdata/thermalright-screen || ls -ld /mnt/user/appdata/thermalright-screen
   ```

2. Install following the README. Check the Plugins entry, logo and **Settings → THERMALRIGHT DIGITAL** page. Start with autostart disabled and choose an identifiable CPU sensor. Compare its reading to the host:

   ```bash
   for h in /sys/class/hwmon/hwmon*; do
     echo "$h $(cat "$h/name" 2>/dev/null)"
     for t in "$h"/temp*_input; do
       test -f "$t" && echo "$t $(cat "$t")"
     done
   done
   php /usr/local/emhttp/plugins/thermalright-digital/control.php status
   ```

3. Click **Start**, wait for the initial image load, then inspect the real Docker contract:

   ```bash
   docker inspect TR-DIGITAL --format '{{json .HostConfig.Devices}}'
   docker inspect TR-DIGITAL --format '{{json .Mounts}}'
   docker inspect TR-DIGITAL --format '{{json .Config.Labels}}'
   docker inspect TR-DIGITAL --format '{{.HostConfig.Privileged}} {{.HostConfig.NetworkMode}} {{.HostConfig.ReadonlyRootfs}} {{json .HostConfig.CapDrop}}'
   cat /var/run/thermalright-digital/driver/status.json
   tail -n 60 /var/run/thermalright-digital/driver/driver.log
   ```

   Expect one `/dev/bus/usb/BBB/DDD` node, two private binds, no boot flash/socket/appdata mounts, and false / none / true / ALL. Check the uppercase name, dedicated Docker icon and WebUI link. After upgrade, no older owned lowercase or long uppercase backend should remain. The log should identify the handshake PM and model. Export diagnostics for unsupported style or USB permission errors; do not bypass isolation with privileged mode.

4. **Observe the physical display.** Temperature should match the selected sensor and load should change with host activity. “Frames sent” confirms transfers, not physical rendering. Uninstrumented GPU areas should stay blank. Stop should release USB. Restart may reuse this boot's handshake cache without assuming unlimited firmware handshake replies.

   Under **Display colors**, try red/green/blue fixed colors, brightness at 25/100%, category colors, breathing and rainbow. Check digits/decorations targeting against reported capabilities. In temperature mode, click **Temperature only**: °C digits should follow the palette while W/MHz/% and decorations retain fixed colors. Then test multiple categories and the global scope filter. Compare colors with thresholds and the displayed sensor; smoothing must affect color only.

   Click **Test colors and segments**: expect an 8-second red/green/blue/segment sequence, an active-test message and automatic return. An unsaved settings draft must remain unsaved. Page readings must remain real.

   Inspect MHz/watt sources read-only:

   ```bash
   for f in /sys/devices/system/cpu/cpufreq/policy*/{affected_cpus,cpuinfo_cur_freq,scaling_cur_freq}; do
     [ -f "$f" ] && printf '%s = %s\n' "$f" "$(cat "$f")"
   done
   for f in /sys/class/hwmon/hwmon*/{name,power*_average,power*_input,power*_label} /sys/class/powercap/*/{name,energy_uj,max_energy_range_uj} /sys/class/powercap/*/*/{name,energy_uj,max_energy_range_uj}; do
     [ -f "$f" ] && printf '%s = %s\n' "$f" "$(cat "$f")"
   done
   cat /var/run/thermalright-digital/feed/metrics.json
   cat /var/run/thermalright-digital/driver/status.json
   ```

   cpufreq uses kHz, hwmon power uses µW and powercap provides energy counters in µJ. The first energy sample is unavailable. RAPL `enabled=0` controls power limiting and does not invalidate a readable advancing counter. `scaling_cur_freq` is labeled requested frequency. Without a supported CPU power source, **Unavailable** and blank W digits are expected. An individually selected rail is not a whole-package measurement. Check `capabilities.power` / `frequency`: host readings may exist without corresponding physical display groups.

5. Test **Scheduled display off** using a window starting in the next few minutes. Keep the backend running. Check the server clock/timezone in Settings, choose today's weekday, save and wait for the boundary. Every digit and decoration must switch off, the status must report scheduled off, and CPU readings must continue updating. Verify the global switch and wakeup on the real screen, not only in JSON. The container must remain the same instance across both boundaries. During off, a visual test must be refused; disabling the schedule and saving must restore output within the next host sample. Test an overnight interval with its starting day selected, entire-day mode, restart inside an off period and changed settings. Check the compact Docker name beside its icon. Avoid changing a production server clock merely to exercise DST; those transitions have deterministic automated tests.

6. Enable autostart and reboot during a maintenance window. After the array and Docker start, verify one backend and one supervisor. Stop/restart Docker through Unraid and check again. Test with the array stopped: bounded waiting and no premature backend.

7. For USB absence/reconnection, use a connection designed for safe hot plugging; do not manipulate an internal motherboard connector live. Disconnecting before startup already tests absence. Record old/new USB addresses and inspect delegated devices after reconnecting. Allow 30 seconds between retries. After 180 seconds of absence, click Start once the device returns.

8. Reinstall the same bundle with Force Install and check settings retention. For a future different backend, test updates, explicit rollback with `force` and automatic rollback after backend failure. Automatic rollback is simulated locally but still needs a real differing backend test before unattended reliance.

9. Export diagnostics before removal and keep `audit-residuals.sh` under `/tmp` or on your workstation. Remove the plugin with Docker available. After 10 seconds, run the audit: no private files or exclusively owned container/image should remain. Verify any unrelated legacy container/image/appdata survived. Native UI caches and system logs are managed separately by Unraid.

10. To test removal with Docker unavailable, use a maintenance window and native UI controls. Stop Docker, request plugin removal, confirm failure and retained inventory, then enable Docker and retry. Do not delete inventory to force success.

Local tests already simulate interrupted installations. Do not cut server power to reproduce them; boot flash and array failure behavior extend beyond this plugin's scope.

## Language checks

Switch Unraid to French and reload the plugin: confirm settings, help, buttons, schedule weekdays, status and errors are translated. Switch back to English and confirm it returns to English even if a previously saved locale was French. Test another installed language, including Arabic for panel direction if available. The rest of Unraid must retain its native layout.

Confirm saved RGB colors, thermal category selection and schedule remain unchanged after language changes. The server clock must retain the Unraid timezone, with the weekday translated. Technical source names and diagnostic logs remain English. Native-speaker review and physical layout checks for long translations remain welcome; automated PHP/FPM/UI checks do not replace them.