PSA: Disable update library on startup in Kodi > Settings > Media > Library after setting up PKC

[jdrch]

I set up PKC on my Ugoos AM9 Pro running CoreELEC nightly last night and everything seemed OK. Then this morning CoreELEC was crashing within a minute of startup. A clean install was stable, but once I restored my backup the crashes would start again. I downloaded all the log files from /storage/.kodi/temp/ and uploaded kodi.log, kodi.old.log, and kodi_crash.log to Gemini. It suggested the above fix, which stopped the crashes. Hope this helps anyone with the same issue.

[croneter]

Kindly share your log files here as well: https://github.com/croneter/PlexKodiConnect/wiki/How-to-Report-A-Bug

[jdrch]

Here are logs:

• kodi.log
• kodi.old.log
• kodi_crash.log

Here's what Gemini Plus said about them:

Technical Diagnosis & Root Cause

Based on a detailed examination of your crash logs, the crash occurs roughly 11 seconds after the PlexKodiConnect (PKC) metadata synchronization thread initializes. The system suffers a fatal SIGSEGV (Segmentation fault) inside Kodi's main GUI thread (LWP 1755).

The crash is driven by a volatile intersection of threading conflicts, heavy database lock contention, and network timeouts operating on a bleeding-edge experimental platform:

1. Concurrent Library Scanning Conflicts

• Right after user login, Kodi's native media engine attempts to launch an automatic background startup scan for both the video and music libraries (void CApplication::UpdateLibraries()).

• Simultaneously, PlexKodiConnect launches its own highly intensive programmatic synchronization loop (PLEX.sync: Doing initial sync on Kodi startup).

• Forcing two distinct, parallel metadata ingest engines to hammer the central SQLite database at the exact same moment introduces intense thread race conditions within Kodi's core memory structures.

2. Thread and Database Contention

• The stack trace reveals that Thread 55 (LWP 2308), spawned by PKC's metadata crawler, is actively processing heavy write transactions inside the database binary (_sqlite3.so).

• Concurrently, Thread 53 (LWP 1908) is tied up in a modal window loop trying to render the CoreELEC configuration interface (XBMCAddon::xbmcgui::Window::doModal()).

• The main GUI thread ultimately encounters a memory access violation (0x000000000627bc10) while struggling to process window focus changes over conflicting background database updates.

3. Excessive Asset Cascades & Network Failures

• The logs detail a massive cascade of network timeouts and address resolution issues immediately preceding the crash.

• Kodi's native music info scanner throws continuous CCurlFile::Stat HTTP response error 22 (404 Not Found) messages while mapping remote directory paths over your Plex direct connection.

• The engine also locks up trying to mount dead local paths, such as attempting an invalid hostname lookup on standard file images (Unable to lookup host: 'folder.jpg') and missing local NFS shares.

4. Bleeding-Edge Platform Instability

• You are running an experimental 22.0-Piers_nightly_20260718 branch centered around unreleased Kodi 22.0 Alpha/Beta code combined with a nascent Python 3.14 runtime environment.

• Development nightlies are highly prone to memory management regressions within Python subinterpreters and object-destructor bindings under heavy multi-threaded write cycles.

---

Step-by-Step Recommended Fixes

To break the login crash loop, follow these troubleshooting interventions in sequence:

Step 1: Disable Native Startup Library Updates

Because PlexKodiConnect programmatically synchronizes your server media directly into the local database, Kodi's built-in startup scan is completely redundant and causes deadly lockups on Kodi 22.

• If you can catch the UI before it crashes: Instantly navigate to Settings -> Media -> Library. Under both Video Library and Music Library, toggle Update library on startup to Disabled.
• Via SSH (If the UI crashes instantly): Connect to your Ugoos AM9 Pro via an SSH client and open your configuration file:

nano /storage/.kodi/userdata/guisettings.xml

Locate the <updateonstartup> flags under the video and music sections and change their values from true to false. Save and exit (Ctrl+O, Ctrl+X).

Step 2: Clean the Local Database Cache & Sequence the Boot

The continuous stream of HTTP error 22 directory path failures indicates that your current database contains corrupted asset links. Re-indexing from a clean state prevents concurrent collisions.

1. Access your box via SSH.
2. Stop the core media service:

systemctl stop kodi

3. Remove your local video and music database cache files to clear out corrupted structural states (PKC will automatically and safely rebuild these on next boot):

rm /storage/.kodi/userdata/Database/MyVideos147.db[cite: 2]
rm /storage/.kodi/userdata/Database/MyMusic84.db[cite: 2]

4. Restart the service:

systemctl start kodi

5. Log in and do not touch any menus or open the CoreELEC settings menu. Allow PlexKodiConnect 5 to 10 minutes to run its initial caching thread in total isolation.

Step 3: Isolate or Disable Music Library Sync

The logs show that the crash heavily correlates with a dense processing loop of your music artists database layout (Original Soundtrack, Eminem, Various Artists).

1. Open the PlexKodiConnect Addon Settings.
2. Navigate to Sync Options.
3. Toggle Sync Music to Disabled, keeping only Movies and TV Shows active.
4. If the crash loop ceases, the issue is isolated to a specific malformed audio tag or corrupt asset image (such as an unreadable embedded PICTURE block in a FLAC file) inside your music collection that the new Kodi 22 scanner cannot handle.

Step 4: Downgrade to the Stable CoreELEC Release

If the crash loop persists across clean database setups, it points to a deep framework bug native to the early Kodi 22 "Piers" beta codebase or its Python 3.14 integration.

• Navigate to the CoreELEC update menu and downgrade your device to the stable CoreELEC 21.3-Omega channel. The 21.x branch utilizes a mature, highly patched database architecture and stable Python runtime environments that do not experience segmentation faults under heavy addon synchronization cycles.