get connection closed constantly

[kingofthebongo2008]

code-index - index_repository (MCP)(repo_path: "H:/code/", mode: "full")
⎿  Error: MCP error -32000: Connection closed

[kingofthebongo2008]

from claude

[DeusData]

@kingofthebongo2008 Could you share more details?

• Which version are you running? (codebase-memory-mcp --version)
• Which OS and editor/agent? (Windows + Claude Code? OpenCode?)
• Does the error happen on every index_repository call or intermittently?
• Is H:/code/ a network drive or local?
• Any output in the MCP server logs?

The "Connection closed" error typically means the server process crashed or was killed during indexing. A path like H:/code/ suggests Windows — there are known Windows path issues we're actively fixing.

[DeusData]

This may be caused by the server crashing during reindexing. Two fixes landed that address likely root causes:

1. c66f688 — Eliminated rename() during reindex. The old pattern (rename(.db.tmp, .db)) could corrupt the database when the MCP server had the file open, potentially causing a crash or connection drop.

2. 08d7c15 — Auto-detects corrupt databases and deletes them with an ERROR log instead of crashing or returning confusing errors.

If you're still seeing this on the next release, please share the MCP server stderr logs — they should now show clear error messages instead of silent crashes.

[DeusData]

Closing as likely fixed by the following changes in the upcoming release:

• 6d4c6b1 — All threads now get 8MB stack (was 512KB on macOS ARM64, causing SIGBUS crashes during autoindex — Stack overflow crash in autoindex_thread → cbm_pipeline_pass_configlink (v0.5.6, macOS ARM64) #139)
• c66f688 — Eliminated rename() during reindex (failed silently on Windows when MCP held the .db file open — index_repository fails silently on Windows: rename() cannot replace open .db file #140)
• 08d7c15 — Auto-detects and cleans corrupt databases instead of crashing

"Connection closed" means the server process crashed. All known crash paths are now fixed. If this still occurs after the next release, please reopen with --version output and any stderr logs.

[DeusData]

Thanks @kingofthebongo2008 for reporting. If the issue persists after the next release, please reopen — we'd love to get it fully resolved.