You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For Linux users, ucm will fail to react to changes to their scratch files, if their system has run out of inotify watches. I've hit that due to running a backup utility. Linux seems to have a low default limit here that it's easy to run into.
This ticket covers getting unison to emit a warning saying "you've hit this error case, here's what to do".
It's not fundamentally impossible to emit a message: tail -f emits the following when it hits this case as it tries to add an inotify watch.
tail: inotify cannot be used, reverting to polling: Too many open files
I've raised a PR on the docs site to add an FAQ entry for this. That's not worth much though: people are much more likely to give up than to find that FAQ, if they hit this issue.
Hopefully the fact that this hasn't come up on the slack means it can't be very prevalent. Maybe it's just my backup application (Tresorit).
For Linux users,
ucm
will fail to react to changes to their scratch files, if their system has run out of inotify watches. I've hit that due to running a backup utility. Linux seems to have a low default limit here that it's easy to run into.This ticket covers getting unison to emit a warning saying "you've hit this error case, here's what to do".
I fear it may be unfixable without changes to hfsnotify - this code https://github.com/haskell-fswatch/hfsnotify/blob/master/src/System/FSNotify/Linux.hs#L99 looks a bit devoid of error handling for adding a recursive listen. And haskell-fswatch/hfsnotify#85 suggests others may have found the same. (The relevant unison code is at https://github.com/unisonweb/unison/blob/master/parser-typechecker/src/Unison/Codebase/Watch.hs#L64 )
It's not fundamentally impossible to emit a message:
tail -f
emits the following when it hits this case as it tries to add an inotify watch.I've raised a PR on the docs site to add an FAQ entry for this. That's not worth much though: people are much more likely to give up than to find that FAQ, if they hit this issue.
Hopefully the fact that this hasn't come up on the slack means it can't be very prevalent. Maybe it's just my backup application (Tresorit).
Previously mentioned under #812
The text was updated successfully, but these errors were encountered: