Depend on accesskit with tokio feature instead of defaults#1372
Depend on accesskit with tokio feature instead of defaults#1372DJMcNab merged 1 commit intolinebender:mainfrom
accesskit with tokio feature instead of defaults#1372Conversation
DJMcNab
left a comment
There was a problem hiding this comment.
This makes sense as the default for Xilem as things stand, but I don't think it makes sense for Masonry Winit.
As such, can we instead do async-io/tokio feature flags on Masonry Winit, which then get passed through to accesskit_winit, and then enable the tokio feature on masonry_winit in Xilem?
|
This definitely makes sense from the Xilem perspective. However, Longer term this needs help from Cargo. It's an ancient issue and we have a Pre-RFC: Mutually-excusive, global features with some discussion but an actual solution in stable Rust seems years away. 😟 So once again, I think the pragmatic choice here is to just merge this and live with Tokio being the choice for Masonry too. |
DJMcNab
left a comment
There was a problem hiding this comment.
Taking another look at Kaur's message (I had originally assumed it raced with mine), I realise that his analysis is correct, and so we should land this. This is an annoying situation, but I agree that this is less bad.
Thanks!
I added
rfdusingtokioto my dependencies, and got a crash in my logs.It seems like this happens when
accesskit_unixdoes not have thetokiofeature, butzbusdoes.Since
xilemalready depends ontokioI thought it would make sense to choose it as an executor for accesskit as well.