URL with '-' in host not detected by autolink plugin

[mpeuss]

Steps to reproduce

1. enable the autolink plugin
2. paste a url with a '-' in the hostname (e.g. ​http://example-site.org/) into the editor text field

Expected result

url string is converted to a clickable link

Actual result

url is not detected

[Jakub Ś]

Problem can be reproduced from CKEditor 4.5.0.

[Jakub Ś]

#13364 we had similar but it has turned out there was rather a problem in user implementation.

Are you by any chance using any non-standard package with third-party plugins not created by CKSource?

[Jakub Ś]

Other issues for autolink plugin: #13569, #14769, #14903, #16685.

[Jakub Ś]

#14903 was marked as duplicate.

​In search of the perfect URL validation regex might help with finding a good regexp.

[Jakub Ś]

#16842 was marked as duplicate.

[Marek Lewandowski]

It's a super trivial bug, let's have it fixed.

[Tade0]

Done.

Changes pushed to ​branch:t/14769.

[Tade0]

Added manual test.

Changes pushed to ​​branch:t/14769.

[Tomasz Jakut]

The fix fixes the issue with not recognizing links with hyphen in them, but it also links invalid links, e.g. ​http://-invalid-

I'm not sure if it's easily fixable, but we could also consider ​replacing our current regex with another one. The questions are:

• Do we want to be 100% compliant with the spec?
• Do we want to include big regex instead of our current small one?

@m.lewandowski @Tade0 WDYT?

[Tade0]

I think most instances of pasting such an invalid link will be people testing if this particular URL is transformed.

In other words: There are significantly more valid urls with hyphens pasted each day into CKE than weird, invalid links.

Originally I had chosen this particular regex because it strikes a good balance between shortness and correctness.

[Marek Lewandowski]

Replying to Tade0:

I think most instances of pasting such an invalid link will be people testing if this particular URL is transformed.

In other words: There are significantly more valid urls with hyphens pasted each day into CKE than weird, invalid links.

Originally I had chosen this particular regex because it strikes a good balance between shortness and correctness.

+1 it make sense. I can imagine being unable to paste links with hypens to be far more irritating, than letting some edge cases through.

[Tomasz Jakut]

In that case, fix is good enough. Merged with ​git:f8e830b.