fswatch: Event queue overflow with pyright lsp in git repo about neovim

[deathbeam]

Yea looks like it isnt working, I will make a PR with a fix

[deathbeam]

Looking at fswatch documentation this error should probably not be handled as actual error, idk:

• The inotify monitor, available on Linux since kernel 2.6.13, may suffer
  a queue overflow if events are generated faster than they are read from the
  queue. In any case, the application is guaranteed to receive an overflow
  notification which can be handled to gracefully recover. fswatch
  currently throws an exception if a queue overflow occurs. Future versions
  will handle the overflow by emitting proper notifications.

EDIT:

Also can be adjusted and fixed with something like this: https://stackoverflow.com/a/62501929/4915340

Maybe including the config in the error message is worthwhile or having like :h fswatch include it and error reference it?

[lewis6991]

This is just a limitation of Linux and pyright asking to watch too many files. It's not really a problem we can feasibly solve.

[deathbeam]

@lewis6991 Sure but at least adding some documentation or improving the error message might help. For example intellij tells you to increase inotify limits if it detects they are too low. The current behaviour where it just throws error message twice for something that is supposed to be just notification per fswatch docs is not very good.

[lewis6991]

Ok. Will review a PR.

[deathbeam]

Oh, also another thing, I was looking at the PR that implemented this feature and there is this: https://github.com/neovim/neovim/pull/27347/files#diff-bfeb92628ce760a1e10360bddb9393ad66b5396a26c0cb7fcdcd39bc60eaa194R283

That is supposed to ignore the .git directory I assume but I cannot reproduce this issue when I remove the .git directory so maybe similar check is missing somewhere else or its not working?

[deathbeam]

Hmm, found this and it looks like --exclude actually doesnt do anything with --recursive specified:

emcrisostomo/fswatch#151

So the PR isnt as easy as i thought. thought about adjusting the regex, as that one is also slightly wrong but I think as per emcrisostomo/fswatch#151 the change would need to provide list of files beforehand, but that would also break watching for new files I assume.

[deathbeam]

Actually I think the real issue was that pyright changed from single file mode to workspace mode when I added .git dir and then it started watching python lib so even if my .git dir was empty same issue, so probs just the documentation is needed, closing again I guess