kses can cut unicode text from tags attributes

[humator]

The following code

$string = preg_replace('/\xad+/', '', $string); # deals with Opera "feature"

in function wp_kses_bad_protocol can also break unicode letters. E.g. Russian word Экран was cut by this function by error (this word was in "alt" attribute). As a result the whole post was cut after that place.

The problem is that 0xAD symbol can appear in unicode text.

I would fix this bug, but unfortunately I don't know what Opera "feature" is mentioned here.

[takayukister]

Hi humator,

From ​kses 0.2.1 release note:

For some reason, the Opera developers decided to make chr(173) a whitespace
character in URL protocols, both when it occurs raw and in an entity. kses
now handles this.

Japanese users also have the same problem.

kses has another issue. See #5917 I opened. The problematic 0xAD stripping you mentioned here is run in the bad-protocol check action. So if #5917 is fixed, this #6233 should be fixed at the same time.

[ryan]

See #9823