Cetra6d ago

I'm pretty surprised that a major bug slipped through the latest rust version: https://github.com/rust-lang/rust/issues/153486

& this is in tokio as well, one of the most heavily used crates!

#rust

Breaking change in `std::os::windows::fs::OpenOptionsExt` affecting 1.94.0 · Issue #153486 · rust-lang/rust

This is a regression from stable to stable that affects Tokio's test suite: tokio-rs/tokio#7955 Code I tried this code: use std::os::windows::fs::OpenOptionsExt; pub struct MyCustomOptions { // Pla...

GitHub
Show threadzyxhere💭6d ago
@cetra3 Its fixed in 1.94.1 which released yesterday
Show threadCetra6d ago
@zyx Yeah, understood, but why was the version released without catching it?
Show threadJacob Pratt6d ago
@cetra3 @zyx because tests don't cover literally every scenario? Mistakes happen.
Show threadCetra

@jhpratt @zyx yeah definitely, but would be good to know if there's any plans to extend the rust release process to catch it if it happens in the future.

Like maybe something like cargo semver could be ran to check whether there are any unknown changes etc..

I'm assuming this is already happening behind the scenes though. It's super rare that version updates break things like this

Mar 27 at 6:50amMastodon for Android
Show threadJacob Pratt6d ago
@cetra3 What you you mean by "extend the Rust release process"? Adding in more checks? Unless things have changed, there is a crater run before every release. That checks every crate on crates.io and on GitHub (with some exceptions due to flakiness) to ensure the lack of regressions. However, a human has to review the output, and that's where things can be missed or dismissed as acceptable breakage.
Show threadCetra6d ago
@jhpratt would the crater run have caught this? From what I understand the issue is on Windows platforms, which I don't think goes through crater
Show threadJacob Pratt6d ago
@cetra3 I've not looked at this issue specifically beyond knowing it's bevy. But yes, crater is Linux-only.