WSL + Windows shared working tree on /mnt causes CodeGraph disk I/O error unless CODEGRAPH_DIR is split manually

[Jek-Mena]

Body

Summary

When the same repo is used from both Windows and WSL, and the repo lives
under /mnt/c/..., CodeGraph can fail in WSL with disk I/O error unless
CODEGRAPH_DIR is manually split (for example .codegraph on Windows and
.codegraph-wsl on WSL).

It looks like CodeGraph already knows this scenario is unsafe, but the
default behavior still allows WSL to hit the Windows-owned .codegraph
first.

Environment

• Windows 11
• WSL
• repo located at /mnt/c/...
• same repo accessed from both Windows and WSL
• CodeGraph version: 1.0.1

Actual behavior

From WSL, CodeGraph failed with:

• Failed to get status: disk I/O error
• Explore failed: disk I/O error

Daemon log also showed socket issues on /mnt/c/...:

• Failed to start server: listen ENOTSUP: operation not supported on socket /mnt/c/.../.codegraph/daemon.sock

There was also fallback to a Windows named pipe, but the DB path still caused
failure.

Expected behavior

On WSL for repos under /mnt/<drive>/..., CodeGraph should either:

1. automatically use a separate per-project data dir such as .codegraph- wsl, or
2. fail with a clear actionable message explaining that Windows and WSL must
   not share the same .codegraph and recommending CODEGRAPH_DIR=.codegraph- wsl.

Workaround

I fixed it locally by making WSL use:

CODEGRAPH_DIR=.codegraph-wsl

After that:

- WSL used .codegraph-wsl
- Windows kept using .codegraph
- codegraph status and codegraph explore worked normally from WSL

## Notes

The current source/docs already seem aware of this issue:

- shared Windows/WSL index risks corruption
- CODEGRAPH_DIR is the intended mitigation

So this seems less like missing capability and more like missing automatic
handling / missing diagnostic UX.