IDNs are not properly converted to ASCII on automatic redirects

[acelaya]

Guzzle version(s) affected: 6.5

Description

I have been waiting for this PR to be merged, #2286, which brings IDN support to guzzle.

The thing is that I have just tested v6.5, which includes it, and I have realized that the idn_to_ascii function is only applied to the first domain provided by the user, but if for some reason some other IDN is returned in the Location header and guzzle is configured to automatically follow redirects, it does not apply it to those, causing some domains to not properly resolve.

How to reproduce

I have some internationalized domain names that can be used to test, which I also use for my e2e tests.

These are all the use cases I cover:

• IDN redirecting to regular domain: http://tést.shlink.io -> https://shlink.io
• Regular domain redirecting to IDN: http://test.shlink.io -> http://tést.shlink.io
• IDN redirecting to IDN: http://téstb.shlink.io -> http://tést.shlink.io

For now, I just disable automatically following redirects and use a recursive function.

[gmponos]

Will this solve the problem? #2424

[acelaya]

Yeah, I guess it should :)

I can try if my tests pass using your fork.

[acelaya]

Hey @gmponos.

I have tried installing your version on my project, by adding your fork as a VCS composer repository, but then, when I use the dev-issue#2423 version constraint, it thinks I want to install commit with ID 2423 from branch issue, and fails.

I have tried escaping/encoding the # char in different ways, but it always fails to resolve to the proper branch.

Could you rename the branch so that it does not include special chars?

[GrahamCampbell]

You could just edit the vendor folder with the changes from the PR. Would be faster...

[acelaya]

You are completely right. I overengineered the whole process 😅

[acelaya]

@gmponos I just tested your changes by copy-pasting the code as @GrahamCampbell suggested, and I can confirm it works 😄

[gmponos]

Btw I think I fixed that on the related PR #2424

I also pushed a branch issue2423 without # in case you want to retest it :)

[acelaya]

Hey @gmponos

I tested again with your branch, and it continues working. Thanks!

[Nyholm]

This is fixed by #2424

Will be released in 6.5.1