Jon Gjengset

TcpStream::shutdown on macOS on localhost yields error

5

**Version** ``` tokio 1.18.1 tokio-macros 1.7.0 ``` **Platform** macOS GitHub Runner 11.6.5 **Description** Calling `TcpStream::shutdown` on a localhost TCP socket on macOS yields ``` Os { code: 57, kind: NotConnected,...

Heavy contention on blocking

9

Code that invokes `spawn_blocking`, or `block_in_place` (which internally calls `spawn_blocking`), all end up taking a shared lock here: https://github.com/tokio-rs/tokio/blob/221f421464e6672e9cf25629998477946a8b8e1f/tokio/src/runtime/blocking/pool.rs#L153 This causes a lot of unnecessary contention between unrelated tasks if...

It can be hard to see the trend in the plots in the coz plotter as the y axis is quite squashed. It would be better if each of the...

Make tabular dependency optional

1

The `README` currently states that you must add the `tabular` plugin before `vim-markdown`. In the interest of keeping my vim install small, I'd prefer to not also install an additional...

Undoing a Sink::poll_ready

4

Before we had `Sink::poll_ready`, senders always had to have an item ready to send before interacting with the `Sink`. This sometimes led to unnecessary buffering. The most trivial example of...

Weird placement in multi-monitor setup

6

**Steps to reproduce:** 1. Have at least two monitors connected. In my case, the primary monitor is rotated 90°, so the layout is: ``` DP-1 connected primary 1200x1920+1920+0 left (normal...

We should adopt the change from https://github.com/HdrHistogram/hdrhistogram-go/pull/48 (in C: https://github.com/HdrHistogram/HdrHistogram_c/pull/107). Chances are it can give us some performance improvements too! cc @marshallpierce because I know you like this stuff :)

Deprecate non-deflate serialization

2

As far as I can tell, the V2 log format in the Java implementation does not support _not_ compressing the histograms. Should we remove support for that? Should we at...

Support deserializing into existing histogram

1

The current [deserialization example](https://docs.rs/hdrsample/6.0.0/hdrsample/serialization/index.html#examples) is a little sad, in that it needs to allocate a new histogram for each of the intermediate deserialized histograms. It would be awesome if there...

Test with loom

6

@carllerche recently published [`loom`](https://docs.rs/loom/) which does correctness checking for concurrent algorithms. It'd be neat to try to get that working for `evmap`'s tests!