Conversation
| .package(url: "https://github.com/apple/swift-openapi-generator", from: "1.0.0"), | ||
| .package(url: "https://github.com/apple/swift-openapi-runtime", from: "1.2.0"), | ||
| .package(url: "https://github.com/swift-server/swift-openapi-async-http-client", from: "1.0.0"), | ||
| .package(url: "https://github.com/apple/swift-openapi-urlsession", from: "1.0.0"), |
There was a problem hiding this comment.
I think it's fine to keep it using AHC? Since it doesn't require the initial message hack?
There was a problem hiding this comment.
@czechboy0 we could, it doesn't matter in the end as there something with transport happening (both for vapor and hbv1), so we anyway need initial message
There was a problem hiding this comment.
there something with transport happening (both for vapor and hbv1)
Can you elaborate what you're seeing? Are you saying even when using AHC, you still don't get the stream established without an initial message?
There was a problem hiding this comment.
Can you elaborate what you're seeing? Are you saying even when using AHC, you still don't get the stream established without an initial message?
Yep, exactly. I will try to investigate, but for now think we can just update the example.
|
@czechboy0 will check, last time I've tried—it was a bit hard, cause it's hard to debug "not happening" events. |
Last week I've been checking bidirectional examples with latest Xcode 16 beta 3 and realised it uses HBv2, which is supported only by macOS 14+, and it's not perfect. After downgrading to HBv1—same problem with returning stream until some bytes are sent appeared. Haven't figured out yet why, maybe some tech inside uses URLSession. So, this PR fixes two things:
initial messagehack.It's working for now. I'll be checking from time to time status of Swift Server tools, maybe we can improve it in the future.