Xcode 13 GM not compiling ByteBuffer-Core

[Andrewangeta]

Expected behavior

Successful compilation

Actual behavior

Failed compilation.

Steps to reproduce

1. Download Xcode 13 RC
2. Try to compile

compile error:
Cannot convert value of type '(UnsafeMutableRawPointer) -> Void' to specified type '@convention(c) (UnsafeMutableRawPointer?) -> Void'
in ByteBuffer-Core.swift

SwiftNIO version/commit hash

version 232.2

Swift & OS version (output of swift --version && uname -a)

swift-driver version: 1.26.9 Apple Swift version 5.5 (swiftlang-1300.0.31.1 clang-1300.0.29.1)
Target: x86_64-apple-macosx12.0

[bdkjones]

I am also seeing this, as shown below. Xcode Version 13.0 (13A233).

Things I've tried:

1. Rebooting — (Don't laugh; rebooting has fixed Xcode many times over 15 years!)
2. Clearing package caches and re-downloading
3. Clean builds

Things That Worked:

1. Rolling back to Xcode 13.0 beta 5 (13A5212g)

[Lukasa]

Thanks folks, can you please file a Feedback Assistant report and write down the number here?

[bdkjones]

Is that....useful?

No offense, but Apple almost never replies to Radars and, when they do, it's 5 years later and the reply is: "We think this is no longer an issue. Please verify with the latest version of [platform] and update your issue if necessary." From the outside, Radar is a blackhole of wasted time and effort.

Since SwiftNIO is an Apple product (I think? It's on Apple's GitHub repo, at any rate) would it not be more efficient to get the Developer Tools team via internal channels?

[Lukasa]

Yes, it’s useful.

[bdkjones]

shrug, okay. Here's the number: FB9626044

[Lukasa]

Much appreciated, thank you!

[Lukasa]

Also filed as https://bugs.swift.org/browse/SR-15197.

[Joannis]

Note: NIO works correctly on iOS

[dweekly]

Also seeing this with XCode GM.

Notably, this causes the gRPC-Swift "hello world" experience to fail out-of-the box:

git clone https://github.com/grpc/grpc-swift && cd grpc-swift && swift run HelloWorldServer

[Joannis]

Note: NIO works correctly on iOS

not surprising. macOS just doesn't exist to apple, as can be seen by this issue.

While I understand the sentiment, it's quite a bold statement. Apple may not see macOS as a platform on the level of iOS, but can you really blame them? Also, the tools in which this bug appeared are a GM for good reason, it's quite difficult to see all the bugs. And the only reason we're confronted with this bug is due to an explicit definition of the function signature.

I don't think interacting with free in this fashion, where you want to support many platform, is a daily usecase for most people.

[Lukasa]

Ok, I think we can work around this nicely enough: see #1956.

[bdkjones]

@Lukasa that's awesome! Any chance this point release can go out today? It'll be at least a week or two before Xcode gets fixed. Thanks for the quick work!

[Lukasa]

At this stage today is probably not practical (the core devs are all UK-based and it's 9pm here, shipping a point release that late is just asking for the gods of irony to cause it to break someone's code in a fun exciting way). However, we had a pre-existing plan to ship a point release tomorrow UK time, so hopefully by the time you wake up tomorrow we'll have a point release out the door.

[Lukasa]

To tie a bow on this, the fix is available in SwiftNIO version 2.32.3.

[dweekly]

Confirmed fixed for me.