The vulnerability exists in the QLExpressEngine.doEval function, which evaluates QL expressions without proper sandboxing. The security patch 978f316c38b3d68bb74d2489b5e5f721f6675e86 confirms this by adding a mitigation directly to the QLExpressEngine class. The patch introduces a static initializer that calls QLExpressRunStrategy.addSecurityRiskMethod(InitialContext.class, "doLookup"). This action explicitly blocks the use of the doLookup method, which is a well-known vector for JNDI injection attacks. The vulnerability description and the patch evidence clearly point to doEval as the function that processes the malicious input (the expression), making it the vulnerable function. An attacker could have previously passed a malicious expression to this function to trigger a JNDI lookup and achieve remote code execution.