Greg Methvin

ok, it's probably not worth trying to debug why that specific hack doesn't work. It's easy enough to switch to another method of getting the paths in a subtree.

@davesann try version `0.17.3` and see if that fixes your bug (note I accidentally released `0.17.2` without the fix). I believe you should get a `DELETE` followed by a `CREATE`...

Interesting... so with xed it appears from the logs that we get an `ENTRY_CREATE` for the `data_science.clj` file, but there never was a prior `ENTRY_DELETE` event for that same file....

So after looking at this a bit more, I realize there's a good reason directory-watcher suppresses the duplicate `CREATE` events. Basically, when a new directory is created, there several situations...

On Linux, directory-watcher defaults to using the default JDK WatchService, which uses inotify. As mentioned in the post you linked, inotify doesn't see certain events on mapped file systems.

One possible solution _should_ be to just switch to the `PollingWatchService` implementation but apparently it's not simple to do that: https://stackoverflow.com/questions/28732693/is-there-a-way-for-force-polling-using-watchservice I haven't looked into the specifics required to get...

Okay, that's an improvement at least. It sounds like you're looking at the inotify logs so this isn't the issue, but make sure you're on 0.17.1 since I recently fixed...

Are there specific types of edits that cause that behavior? Did you check if it's the underlying Java watch service that's duplicating the events or if they're generated by directory-watcher...

Which platforms have you tested on? Seems like it's either due to a bug in the underlying WatchService implementation or the events the OS itself is reporting.

I likely won't have time to look into this further anytime soon, but I'm happy to accept pull requests if anyone finds a viable way to handle this.