[lathiat]

It’s due to what is effectively a network file system. You have the same issue with nfs, 9pfs, parallels prl_fs, etc. It has to stat every file in the repo and that gets expensive on ang RPC protocol. Even local ones apparently.

This is made worse because git uses the os-specific file statistics including inode number to track file changes. These change over most such file systems types which triggers it to rebuild the entire file cache every time you switch OS running git commands. Which if you have shell integration is constantly.

I share my git repos between MacOS and a Linux volume and it hurts on any repo with 100s of thousands of files or more. Ceph, Linux kernel, etc.

[nsteel]

Turns out you guys are right. I tried a bigger repo and it was slow. I tried our large monorepo and it was utterly unusable (literally minutes for git status).

I've found this thorough explanation https://github.com/microsoft/WSL/issues/4197#issuecomment-60...