Exit interview #2

[makepaddev]

As i think winit is useful to the rust community here is my exit interview #2. Windows.

My last issue was just on mac/web, but since i'm redoing the platform layer on windows heres some feedback:.

WHY IS EVERYTHING THREADED. sorry but really. I just reduced thousands of lines of winit complexity crap to like 5 platform calls and a windows message loop. See, this is what you would expect.
https://github.com/makepad/makepad/blob/master/render/src/cx_windows.rs
This is about as complex as the windows platform layer should be. I've also reduced Cocoa to
this
https://github.com/makepad/makepad/blob/master/render/src/cx_cocoa.rs

Sorry but winit is just architected wrong, done wrong, everything is wrong. I can't use it, i think its bad for the Rust community to exist in its current form. Please reconsider everything. I emplore you, with live working example code on how i would rewrite winit from scratch for all platforms. Windows, mac, web and linux is coming up.

So. I'll make a more detailed post later, but i'm fixing it regardless. Please. Its bad. think about how you are doing things, its bad for Rust.

Goodluck.

[makepaddev]

Also you have a C++ dependency that doesn't build on GNU windows. Come the F* on. Really.

[makepaddev]

Reason i'm upset about a (needless) C++ dep on windows is it destroys the trivial-compile-anywhere-easily idea of Rust which it truly has from the compiler/stack even the winapi level. People put real hard effort in making this work as smoothly as it does. Only to be undone by a careless C++ dep here. So there.

[goddessfreya]

A) Could you be a bit less aggressive?

Please be kind and courteous. There’s no need to be mean or rude.

Respect that people have differences of opinion and that every design or implementation choice carries a trade-off and numerous costs. There is seldom a right answer.

Please keep unstructured critique to a minimum. If you have solid ideas you want to experiment with, make a fork and see how it works.

-- https://www.rust-lang.org/policies/code-of-conduct

B) We try out best to support multi threading because that's one of Rust's strengths. Not everything is multi threaded however. Examples: EventLoop, macos's Window possibly (see #873)

C) Winit is as complex as it is for two reasons:
1- partially because of bad design (which we are always willing and trying to improve, see the eventsloop-v2 branch); but mostly
2- because we have to handle more than one use case. I suspect your 300 line windows implementation does not provide all the functionality that winit provides and does not cover all use cases.

Nevertheless, if those 300 lines are all you need for your use case, go for it! Maybe even write Winit-lite, or better yet, Winit-not (shortened as Winot)!

D) I wasn't, nor do I think anyone else was aware that this - so far unnamed - dep was causing compile issues. Here's how you can help:
1- Add to our CI suit some builds that use MinGW so we can spot these kind of issues in the future before merging PRs.
2- Fix the compile issue.

[makepaddev]

Sorry i've been asked to give my feedback instead of just ignoring it and moving on (via twitter).

[makepaddev]

Since apparently lots of people are having problems but not giving feedback leaving the project nothing to go on.

[makepaddev]

Also i've long since realised i can't 'fix' winit with pull requests since i'd really rewrite the whole thing from scratch. But its also 'kinda' hurting Rust as an ecosystem by being architected so poorly.

[makepaddev]

And yes, i understand the 'more than one usecase' thing, but that doesn't mean the primary Rust window-abstraction library is almost unusable for anyone standing up a real product on it, as i am, and have been running into massive unfixable issues over the last few months.

[makepaddev]

So sorry its nothing personal, i'm not angry at any one person, or you, but i am angry that this is the state of multiplatform windowing on Rust. Its garbage. Sorry, and really consider the architecture. I can't be more forceful about it than i am. Working code in links. And yes, my 300 loc windows thing is just the beginning, but it will work just fine.

[jdm]

Just because you feel strongly that winit does not serve your needs does not mean it is acceptable to say things like "it's garbage". Feedback is one thing; insults are unwarranted.

[makepaddev]

And the argument 'you dont have our usecase' yea ok, fine, but if you wanna be a fullscreen GL thing whats all the windowing for then eh :) The cocoa winit thing missed core features like 'being able to reliably start up more than 4 out of 5 times'. Clipboard, IME support. The windows thing didnt' even get the core message loop concept right. Sorry but. yea. Its harmful to the Rust ecosystem.

[makepaddev]

So yea sure it may be hurtful, it may be harsh. but you know what is harsh, releasing a software product on top of Winit and expecting to get paid for it by real users. So yeah. you get harsh feedback. Its not personal right. But winit? as a project? Yeah its total garbage. Period.

[makepaddev]

Anyway, flamewar. I'm here to help i'm here to point in the way i think things should be. Thats why i bother to respond. :)