wp_texturize to defect certain links in comments and on page

[h4x3d]

the wp_texturize() function in formatting.php (includes) of wordpress breaks links of the domain format (number)x(number).

For example, a comment left by a domain such as www.h4x3d.com, which includes the (4)x(3) in the domain name, renders to www.xn--h43d-rma.com. This breaks the link.

It does occur on the page, but also with in the page comments.
It does also occur when the text with the special combination is not a link, but plain text.

[h4x3d]

graphic showing the error (wp_texturize)

[Nazgul]

This is caused by make_clickable being run after wp_texturize.

If you switch those around the link text still gets changed, but the link href works.

Attached patch does just that.

[Viper007Bond]

Looks good to me.

[intoxination]

Glad someone figured this out. It is one of those bugs that has been a nightmare. It was reported on #3777 also.

[rob1n]

Looks good to me, as well.

[rob1n]

I could reproduce this on a trunk install, but the patch didn't fix it.

[Nazgul]

Replying to rob1n:

I could reproduce this on a trunk install, but the patch didn't fix it.

What input did you use? I tried all the cases I could think of and got all working links with my patch.

Just curious what I missed.

[rob1n]

I put ​http://www.h4x3d.com/ (just text) in a comment, but it didn't turn out quite like what the description says. It was messed up, just that ​http://www.h4 and then the x was a multiplication sign, and the rest text.

I tried switching the priority to 9, but still nothing. I'll try again.

[rob1n]

Okay, tried again, works. Committing.

[rob1n]

(In [5236]) Run make_clickable before texturize. Props Nazgul. fixes #4116

[h4x3d]

thanks for all your efforts!

[h4x3d]

I was just trying out the 2.2er version of wordpress on my website (h4x3d.com) and found out that some instances (see attached image (50kb) or source code of ​http://wordpress.h4x3d.com)of 4x3 are turned into × which breaks the link to the style-sheet and so on.

I assume this is connected to the wp_texturize function. Can anyone help me?
I have set-up a test-install of wordpress 2.2 on ​http://wordpress.h4x3d.com

[h4x3d]

shows the error (both on site and source code)

[h4x3d]

(4)x(3) in the domain name renders to www.h4×3d.com. This breaks the link.
not sure which breakage I prefer, the wordpress 2.1 xn-rma break or this ×...

[Nazgul]

Try removing the , '/(\d+)x(\d+)/' and , '$1×$2' parts from the $dynamic_characters= and $dynamic_replacements= lines in the wp_texturize function.

[h4x3d]

Replying to Nazgul:
works like a charm nazgul. removed it from /wp-includes/formatting.php

is this likely to get fixed in a future release of wordpress?
thank a lot for your help already!

[h4x3d]

I put it to work on ​http://themes.h4x3d.com and left ​http://wordpress.h4x3d.com unfixed for later review (if anyone needs to check the defect)

[h4x3d]

thanks rob1n

[Nazgul]

Replying to h4x3d:

Replying to Nazgul:
works like a charm nazgul. removed it from /wp-includes/formatting.php

is this likely to get fixed in a future release of wordpress?
thank a lot for your help already!

I did some further testing. The code is there for a reason, it makes certain types of text look pretty, but it shouldn't be used on (part of) links. Therefore it seems theme related.

Could you attach the header.php from your theme here, so I can do some further testing?

I'm also lowering the priority, because you have a working workaround and not a lot of people are affected by this.

[Nazgul]

Duh... I actually have to change the priority. Just typing about it doesn't make it happen. ;)

[h4x3d]

header.php

[h4x3d]

sure thing nazgul,
but I think it is theme UNrelated, as it appears to happen even with the default wordpress theme. the link to the stylesheet is broken due to the 4x3 in the domain name and the filter changing it from the actual 4x3 into some fancy ascii multiply-sign which breaks it.

[h4x3d]

I uploaded the files as you wanted and added some more description in my posts above, how can I help additionally?

[Nazgul]

It took me a lot longer to track this bug down than I'd thought.

There is a bug which was introduced in changeset [4983], which causes all thing passed to bloginfo to be wptexturized, instead of just the non-url ones.

I've attached a patch which fixes this.

[rob1n]

Wow. Four characters. I'm amazed :).

[rob1n]

(In [5526]) Fix bloginfo() filtering when it comes to links. Props Nazgul. fixes #4116

[codein2]

I think it make more sense if the pattern look something like this: /(\d+)x(\d+[^{\w])/ in place of /(\d+)x(\d+)/}

because a 23x23 is math but a url like ​http://www.youtube.com/watch?v=irWR7F0x2uU isn't.

maybe there are better solution but this pattern broke some video/youtube plugins because the url looks like ..?v=irWR7F0×2uU wich a flashplayer didn't unterstand.

[markjaquith]

You know what? We've burnt a lot of cycles on this over the years. Poor h4x3d is always getting screwed by this. Why don't we just axe it adn leave mathematical forumas up to plugins?

[Otto42]

+1 to that. I don't see that converting number x number into number × number is particularly useful anyway.

[Nazgul]

+1 from me as well.

I've attached a path for it.

[mrmist]

See also #4539, #3810. It'd be nice if all the wp_texturizes fixes could be lumped together and sorted. Pattern-matching is not for me though.

[hakre]

Another long-player in the formatting bug league, +1 for getting a
definition, unit-tests and merging with related ones.

Versus: #3833

[hakre]

Related: #4152

[norbertm]

Both math and URLs containing (number)x(number) will be fixed by #4539. Submitting unit tests there together with a bunch of other tests for the same ticket.

[miqrogroove]

This was mostly resolved in [14144]

It would still be a problem where a domain name begins or ends with the 9x9 phrase.

[timhibberd]

This is still a problem.

I had to use

add_filter( 'no_texturize_shortcodes', 'shortcodes_to_exempt_from_wptexturize' );

because urls like ​http://xyz-300x200.jpg were being converted to ​http://xyz-300×200.jpg

[Mte90]

Tested now with 6.2 alpha, in a post content and in a comment with http://xyz-300×200.jpg.

The HTML generated in comment is <p><a href="http://xyz-300x200.jpg">http://xyz-300&#215;200.jpg</a></p>, so just sanitize the content but not the URL.

To me this ticket can be closed.

[audrasjb]

As per the last old tickets triage session, it appears thats the issue can be reproduced anymore.
Closing this as worksforme.

Feel free to reopen the ticket and to share reproduction steps if you find a way to reproduce the issue :)