Inconsistent handling of IDN vs non-IDN hosts in URL()

[nreilly]

Describe the bug
URLs created with long hostnames has inconsistent behaviour between IDN and non-IDN hostnames. For hostnames longer than 63 characters, non-IDN hostnames work fine, but IDNs result in URL returning Nil.

To Reproduce
Here's some sample code to show the issue:

import Foundation

func letters(_ count: Int) -> String {
  return String(repeating: "a", count: count)
}

func extraLetters(_ count: Int) -> String {
  let face = "🧐"  //  = xn--9u9h
  switch count {
  case 8: return face
  case 0...12: return letters(count)
  case 13...36: return letters(count - 10) + face
  default: return letters(count - 11) + face
  }
}

func host(_ host: String) -> String {
  return URL(string: "https://\(host)/")?.host(percentEncoded: false)
    ?? "<No URL>"
}

func printComparison(_ count: Int) {
  let prefix = String(format: "%03d: ", count)
  print("\(prefix)\(host(letters(count)))")
  print("\(prefix)\(host(extraLetters(count)))")
}

for comparison in [8, 30, 63, 64, 128, 256] {
  printComparison(comparison)
}

And here's the output:

008: aaaaaaaa
008: xn--9u9h
030: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
030: xn--aaaaaaaaaaaaaaaaaaaa-uv13t
063: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
063: xn--aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa-r3017c
064: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
064: <No URL>
128: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
128: <No URL>
256: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
256: <No URL>

Expected behavior
Consistency between IDN and non-IDN handling. I realise each section in a hostname is meant to be up to 63 octets (characters) (RFC1123), so I can understand that URL returns nil for IDN names, but I don't know why it then works for non-IDN.

Configuration (please complete the following information):

• Version 26.0 (17A321) / 26.0.1 (25A362) / swift-driver version: 1.127.14.1 Apple Swift version 6.2 (swiftlang-6.2.0.19.9 clang-1700.3.19.1)
• Swift 6.2-Nightly via SwiftFiddle

Additional context
Also posted on the swift forums

[jrflat]

Thanks for pointing this out and providing the repro case! I think it's a good idea to maintain consistency between IDN and non-IDN handling. For compatibility, I don't think we can enforce domain name lengths in the non-IDN case, but I can look into allowing UIDNA_ERROR_LABEL_TOO_LONG | UIDNA_ERROR_DOMAIN_NAME_TOO_LONG when IDN-encoding to ASCII.