Thomas Leonard

Updating the benchmarks to use monotonic times sounds good. Getting eio_linux to use the realtime clock directly would be good too, fixing this comment: https://github.com/ocaml-multicore/eio/blob/32c26ab3133b5e51a981533cb4bb201d0eb35148/lib_eio_linux/eio_linux.ml#L1300-L1304 I don't see why that...

You don't need to (and can't) use `Uring.wait`'s timeout for real-time timeouts. You can just submit them as their own separate operations instead. I don't see what completion counts have...

Closed by #436.

`seek` and `truncate` were added in https://github.com/ocaml-multicore/eio/pull/626.

Closed by #436.

It's probably easier to make a pull request (or just edit the main feed's XML) in that case. I suspect updating it from the submitted feed will not be what...

Here's the log that results from running the reproduction commands above (I ran the build twice, so the log only has the failing part): ``` $ cat _build/log # dune...

> cc @talex5 I think you had some 'eio' optimizations too, but adding some minimal support for 'eio' first is probably best [My angstrom-eio branch](https://github.com/talex5/angstrom/tree/effects) rewrites everything to use effects,...

I rebased this. The conflict was due to #1025. If I understand the logic there correctly, this is what cohttp-lwt now does: 1. The user's handler uses e.g. `respond*` to...

It could certainly work either way. The main problem is that separate processes are harder to experiment with because you have to recompile both binaries and restart them both each...