force_balance_tags breaks JavaScript

[yellowafterlife]

Input:

<?php
echo force_balance_tags('<script>console.log("extest:", 0<=1);</script>')
?>

Output:

<script>console.log("extest:", 0< =1);</script></script>

(no longer valid JavaScript)

Context:

I use tiny (<20KB) JS programs for interactive illustrations inside tutorials on a WP blog. These would most commonly break on pages where excerpts are shown (post list, search) and it took a little while to narrow this down to force_balance_tags running for !--more splitter in get_the_content.

Workarounds: enclosing JS code in a comment

<script><!-- code --></script>

which is a legacy part of spec ​https://html.spec.whatwg.org/multipage/scripting.html#restrictions-for-contents-of-script-elements

Could the script-tags be treated like the comments in this regard? This is a rather unobvious behaviour.

[soulseekah]

This is a very common issue among formatting functions in WordPress, wherein "stray" brackets break tag detection. This happens in scripts, styles and even regular HTML attributes with angle brackets in them. There are a bunch of tickets open regarding this.

The issue here is that <=1);</script> is detected as a tag.

Also there are no unit tests. Let's start writing some.

[soulseekah]

Tests for force_balance_tags

[dmsnell]

Hi @soulseekah - I'm exploring replacing the logic from force_balance_tags() with the new WP_HTML_Processor::normalize() and it seems to handle the cases you've raised here.

<?php
echo force_balance_tags( '<script>console.log("extest:", 0<=1);</script>' );
// <script>console.log("extest:", 0<=1);</script>

echo force_balance_tags( '<p>hello' );
// <p>hello</p>

echo force_balance_tags( '<p>hello</p>' );
// <p>hello</p>

since the logic is based on the HTML API it should be robust regardless of the input. no more guessing! as of WordPress 6.7.0 there will see be a marginal number of input HTML documents not supported by the HTML API; these will fall back to the legacy behavior of force_balance_tags() (or perhaps we'll replace the custom parsing with the Tag Processor which is also based on the HTML spec but doesn't know HTML structure).

​https://github.com/WordPress/wordpress-develop/pull/7409