Closed seanmonstar closed 8 months ago
Hi, it appears that h3
doesn't directly depend on hyper
. What actions do we need to take for h3
? Can I provide any help for it?
h3
depends on http
, which defines the types for hyper
. h3
is already updated in https://github.com/hyperium/h3/commit/5c161952b02e663f31f9b83829bafa7a047b6627 but it still needs to be released on crates.io
Just checking for planning purposes, any idea how long until this ships? We'd like to upgrade to hyper 1, axum 0.7 and reqwest in one go, and reqwest is the odd one out.
I've been pushing commits to #2059 as I make progress, in case any wants to watch or help.
Just checking for planning purposes, any idea how long until this ships? We'd like to upgrade to hyper 1, axum 0.7 and reqwest in one go, and reqwest is the odd one out.
In case it helps, I've got code that uses hyper
1, axum
0.7 and current reqwest
and works fine. It's not ideal because I've got to do some conversions myself because of mismatched http
versions, but for what I've needed it's not been too hard.
Looks like hyper-rustls
is making solid progress in https://github.com/rustls/hyper-rustls/pull/232, but h3
is not yet released. Is it getting time to ask for one?
hyper-rustls 0.26 now seems to support hyper 1.0.
Made a lot of progress today, most tests are passing. Seems the unstable h3 feature is blocked, since Quinn currently depends on rustls 0.21 (https://github.com/quinn-rs/quinn/pull/1715), but hyper-rustls depends on rustls 0.22.
@seanmonstar merging #2059 is expected to be a breaking change? Asking because I'm working on a service that depends on reqwest
and we plan to update to hyper v1.
It's recommended to wait for this to be closed, or the public APIs will not be changed?
I skimmed through the PR and the only "breaking" change I found is the temporary removal of http3 feature, which is gated behind reqwest unstable cfg so it's probably ok.
First of all, yay, it's done and merged!
I skimmed through the PR and the only "breaking" change I found is the temporary removal of http3 feature, which is gated behind reqwest unstable cfg so it's probably ok.
I'm skeptical. I thought many public types such as reqwest::headers::HeaderMap
were direct re-exports from http
. They're now compatible with http 1.0 types rather than http 0.2 types. Isn't that a breaking change already? I think this needs to be released as reqwest 0.12.0 rather than 0.11.whatever?
Correct, it is a breaking change. The exposed http
types and integration are going from 0.2 to 1.0.
Can I ask for a clarification about this? I see it's merged but not released yet but some versions with the old http
version were released.
Is the release blocked on something else (docs or something else?) or does this just wait for the release itself?
Reqwest 0.12 was released yesterday with this.
A tracking issue for upgrading reqwest to use hyper 1.0. It'll be a breaking change, and we should probably include some other breaking fixes that have piled up. But those can be tracked else where. This is just the items needed to update the hyper dependency.