Shell latency using mise on network drive (NFS)

[zaucker]

When using mise on a shared development server inside a NFS4 mounted Home directory, we noticed "latency" running commands or just pressing Ctrl-C in the shell the first time we entered the directory or after some idleness in that shell. Commands executed directly afterwards (with little wait in between) did not show that latency.
Moving $HOME/.cache, $HOME/.config, and $HOME/.local to a fast local SSD storage didn't help.

strace -f mise hook-env -s bash |& grep -P '^statx(' | wc

showed 108 filesystem accesses by mise, most of them to the NFS mounted storage.

We suspect that during the first run of mise, this takes a few seconds, afterwards filesystem buffering makes this fast, until the buffers "loose" this information through other filesystem usages and then the next call is slow again. There seems not much that can be done to help that with a network drive as we have no means to force filesystem buffering to be kept (which probably wouldn't play very well with the system as a whole).

For our situation, where we only rely on mise "providing" the proper tool versions without doing any env variable setting, we used this workaround:

   lazy_mise () {
     if [ x$LAST_PWD != x$(pwd) ]; then
        _mise_hook
        LAST_PWD=$(pwd)
     fi
   }
   PROMPT_COMMAND="lazy_mise;${PROMPT_COMMAND}"

in our $HOME/.bashrc files. This makes mise only do its thing when changing directories.

NOTE: this also "looses" the automatic/immediate mise run when, e.g., editing the .tool-versions file.

Just wanted to share this in case anybody else has a similar setup and annoying latency.