PHPUnit has Argument injection via newline in PHP INI values that are forwarded to child processes

Impact

PHPUnit forwards PHP INI settings to child processes (used for isolated/PHPT test execution) as -d name=value command-line arguments without neutralizing INI metacharacters. Because PHP's INI parser interprets " as a string delimiter, ; as the start of a comment, and most importantly a newline as a directive separator, a value containing a newline is parsed by the child process as multiple INI directives.

An attacker able to influence a single INI value can therefore inject arbitrary additional directives into the child's configuration, including auto_prepend_file, extension, disable_functions, open_basedir, and others. Setting auto_prepend_file to an attacker-controlled path yields remote code execution in the child process.

Sources of INI values that participate in the attack:

• <ini name="…" value="…"/> entries in phpunit.xml / phpunit.xml.dist
• INI settings inherited from the host PHP runtime via ini_get_all()

Threat Model

Exploitation requires the attacker to control the content of an INI value read by PHPUnit. In practice this means write access to the project's phpunit.xml, the host php.ini, or the PHP binary's environment. The most realistic exposure is Poisoned Pipeline Execution (PPE): a pull request from an untrusted contributor that modifies phpunit.xml to include a newline-containing INI value, executed by a CI system that runs PHPUnit against the PR without isolation. A malicious newline is not visibly distinguishable from a legitimate value in a typical diff review.

Affected component

PHPUnit\Util\PHP\JobRunner::settingsToParameters().

Patches

The fix has two parts:

1. Reject line-break characters

Because a newline or carriage return in an INI value has no legitimate use and is the primitive that enables directive injection, any PHP setting value containing \n or \r is now rejected with an explicit PhpProcessException. This follows the same "visibility over silence" principle applied in CVE-2026-24765: the anomalous state fails loudly in CI output rather than being silently sanitized, giving operators an opportunity to investigate whether it reflects tampering, environment contamination, or an unexpected upstream change.