External link icon not always shown after update to 7.x-1.19

Just to be a bit more specific. Do ALL links stop getting the extlink icon? Also do you have any errors in the console?

I just installed this module, seems only some links work, others don't.

I've tried some tests, there seems to be a pattern.

It never seems to work on domains like example.org.au, example.com.au (seems to be baised against Australian domains :).

Also some other domains it seems to fail on http://www.aussiepigs.com/lucent, but http://www.example.com/lucent will work fine (More evidence of anti australianism!).

Just to finally prove this clear discrimination:

www.auspigs.com - doesn't work
www.usapigs.com - does work!

Clearly some sort anti-aussie conspiracy here.

Can you post the html in a code block here? Also what are your settings and did you receive errors in your console?

Are any of you using any css exclude, regex exclude, or "Exclude links with the same primary domain."?

I just installed the module and maybe turned on open links in a new window, nothing else. Exclude links with the same primary domain is checked, but I don't recall checking it, think it was turned on by default.

Hmm, so I think you could be on to something though, I just turned it off and all the links work. So something askew there.

Can you apply this patch and see if it fixes the problem?

Thanks. I'll do a release soon.

As I originally reported in #2961407: Regressions in 1.19?, I think version 1.20 still isn't working correctly.

One of my main problems was that I was testing on a temporary domain where the site is actually on a sub-domain and I was getting confusing results.
That aside though, the regex looks strange to me.

It would be useful to have some comments in the code stating what the different blocks of regex are trying to match, eg. domain name, port, file extension, etc.
It would make it much easier for a new developer to come in and do work on the code.

The current version has this:

    var pattern = /^(([^\/:]+?\.)*)([^\.:]{1,})((\.[a-z0-9]{1,253})*)(:[0-9]{1,5})?$/;
    var host = window.location.host.replace(pattern, '$2$3');
    var subdomain = window.location.host.replace(host, '');

So if my page is at website.gov.au/some/path then I get this

• $0 = website.gov.au
• $1 = website.gov.
• $2 = gov.
• $3 = au
• $4 =

So host ends up being gov.au and the subdomain is empty, which is wrong.

If my page is at www.website.gov.au/some/path then I get this:

• $0 = www.website.gov.au
• $1 = www.website.gov.
• $2 = gov.
• $3 = au
• $4 =

So host ends up being gov.au and the subdomain is empty, which is wrong.

If you go back to 1.18 you have this:

  var pattern = /^(([^\/:]+?\.)*)([^\.:]{4,})((\.[a-z]{1,4})*)(:[0-9]{1,5})?$/;
  var host = window.location.host.replace(pattern, '$3$4');
  var subdomain = window.location.host.replace(pattern, '$1');

So if my page is at website.gov.au/some/path then I get this

• $0 = website.gov.au
• $1 =
• $2 =
• $3 = website
• $4 = .gov.au
• $5 = .au

So host ends up being website.gov.au and the subdomain ends up being empty, which is correct.

If my page is at www.website.gov.au/some/path then I get this:

• $0 = website.gov.au
• $1 = www.
• $2 = www.
• $3 = website
• $4 = .gov.au
• $5 = .au

So host ends up being website.gov.au and the subdomain ends up being www, which is correct.

Thanks for your input. Please see the discussion at #2914838: (Sub-)domain detection not reliable..

PS...
The issue where this bug was introduced was #2679977: New top level domains and pattern matching

The commit was: https://cgit.drupalcode.org/extlink/commit/?id=d6ab635

Will need to do a bit more work on the regex to fix both issues I assume.

If that's not trivial, the quick temporary fix would be to revert that commit, since it is worse with that commit for most (maybe all?) production use cases.