PCRE regular expressions with JIT enabled gives different result

[antman3351]

Description

Hi,
The fallowing regex preg_match( '/<(\w+)[\s\w\-]+ id="S44_i89ew">/', '<br><div id="S44_i89ew">', $matches ); in PHP >= 8.1 gives different results if PCRE JIT is enabled or not. ( Note, I tried to simplify the regex to isolate the problem, that's why it doesn't make much sense ).

I had this problem with PHP 7.3 where I found a bug report that sated the problem was the PCRE library and not PHP.
I've recently migrated the code the regex is used in to PHP 8.2 and found there's still an issue, so I tested it on 3v4l.org and it seems to be fixed in PHP <= 8.0 and broken in 8.1.

I noted the issue on the PCRE2 github but the author doesn't think the issue is with any changes in the PCRE library: PCRE2Project/pcre2#261 (comment)

Code example: https://3v4l.org/189jh

Thanks,
Antonio

PHP Version

8.2.7

Operating System

No response

[nielsdos]

I can reproduce this locally. I'll run a bisect. This will also reveal if the problem is in PHP or in PCRE.

[nielsdos]

Bisect finished.

It points to 5d42900 and 79755ca.
This is indeed a pcre2lib problem in 2.37. Given that the pcre2lib maintainer says it does not reproduce in their testings on 2.42, I'd say the solution here is updating PHP's bundled pcre2lib to the latest version.

As a temporary workaround you can configure & build with the flag to use your system's pcre2lib, and hope that version is high enough to not contain the bug.

[hormus]

Or for compatibility

$equal = preg_quote('=');
$pattern = '/<(\\w+)[\\s\\w\\-]*id' . $equal . '"S44_i89ew">/';

// or for other regex
$equal = preg_quote('=');
$pattern = '/<(\\w+)[\\s\\w\\-]*id' . $equal . '"S44_123">/';

Pattern /<(\w+)[\s\w\-]+ id="S44_i89ew">/ not + but * and without white space id
for subject  <br><div id="S44_i89ew">

[antman3351]

Or for compatibility

$equal = preg_quote('=');
$pattern = '/<(\\w+)[\\s\\w\\-]*id' . $equal . '"S44_i89ew">/';

// or for other regex
$equal = preg_quote('=');
$pattern = '/<(\\w+)[\\s\\w\\-]*id' . $equal . '"S44_123">/';

Pattern /<(\w+)[\s\w\-]+ id="S44_i89ew">/ not + but * and without white space id
for subject  <br><div id="S44_i89ew">

Your regex doesn't cause the issue where there is a difference between JIT and non JIT.