Will

These tests could measure the latency between: 1. a filesystem event, and 2. the event being reported by these watchers The first could be a simple bash script. The second...

> It may be worth measuring also how long it takes to watch a big folder, and how much memory that takes. Definitely. I'm curious about different scenarios and profiling...

I'm comfortable with our performance tests and valgrind

Debug logging isn't clear. Is there some issue with powershell? https://github.com/e-dant/watcher/actions/runs/6426774299/job/17451612755#step:6:62

Switching to g++ seems to work for the time being: https://github.com/e-dant/watcher/actions/runs/6684463968/job/18161755411 Ideally, we would build on both clang and gcc, and we would do it with Nix.

> @e-dant I encountered the similar issue here. Maybe you could try to use clang 15 instead of clang 14. see: [snitch-org/snitch#138 (comment)](https://github.com/snitch-org/snitch/pull/138#issuecomment-1781507770) Thank you very much -- Lots of...

This could go in the "The Library" section of the readme. Can't think of a better first issue for potential contributors to get their name up on the project!

``` 74 bool wtr_watcher_close(void* watcher) 75 { 76 if (! watcher) return false; 77 if (! ((wtr::watcher::watch*)watcher)->close()) return false; 78 delete (wtr::watcher::watch*)watcher; 79 return true; 80 } ``` Either the...

I think https://github.com/e-dant/watcher/commit/343e6b5975036d43a3cadc2eaf813584c2b58f3c might fix this, but I'm not sure yet. There might be some double-close logic (for any reason at all -- implicit destructors if the api is used...

I think this is worth investigation. It's a problem if we're failing to close resources. I think a minimal reproducer would be enlightening -- Particularly, I'm curious if there's a...