False Negative: CloseSql.ql misses leaked JDBC resources once the allocation site is split across helper calls or `getResultSet()`.

从 Likely Bugs/Resource Leaks/CloseSql.ql 开始，并将其获取处理与受影响的 PosCase1.java、PosCase6_Var3.java 和 PosCase6_Var4.java 示例进行比较。验证 checker 会报告泄漏的 Connection 以及间接获取的两个 ResultSet 实例，同时保持现有覆盖率。

Version
codeql 2.24.3

Checker

• Checker id: Likely Bugs/Resource Leaks/CloseSql.ql
• Checker description: This checker detects SQL resource objects (Connection, Statement, ResultSet) that are initialized locally and not guaranteed to be closed on method exit.

Description of the false negative

These cases still leak JDBC resources. One is the simplest possible leaked Connection. The other two leak ResultSet instances that come from a Statement parameter and are never closed before the method returns.

What changes between the samples is only how the resource is obtained: directly, through a helper method, or through Statement.getResultSet() after execute(...).

Affected test cases

PosCase1.java

The method opens a Connection and exits without closing it. This should be a baseline match for the rule.

PosCase6_Var3.java

The ResultSet comes back from a helper, but it still originates from the passed-in Statement and still leaks.

PosCase6_Var4.java

The ResultSet is retrieved through stmt.getResultSet() after execute(...). That is still a live SQL resource that needs to be closed.

Reproduction code

PosCase1.java

// Scenario 1: A java.sql.Connection is locally initialized, not assigned, not passed to a local constructor, has no parent, and is not closed.
package scensct.core.pos;

import java.sql.Connection;
import java.sql.DriverManager;
import java.sql.SQLException;

public class PosCase1 {
    public void test() throws SQLException {
        // Locally initialized Connection, assigned to a variable but not closed.
        Connection conn = DriverManager.getConnection("jdbc:example:db");
        // No close() call before method exit.
    }
}

PosCase6_Var3.java

// Scenario 6: A java.sql.ResultSet is locally initialized as a child of a non-locally-initialized Statement parameter, used directly, and not closed.
package scensct.var.pos;

import java.sql.ResultSet;
import java.sql.SQLException;
import java.sql.Statement;

public class PosCase6_Var3 {
    public void test(Statement stmt) throws SQLException {
        ResultSet rs = fetchResult(stmt);
        // rs not closed
    }

    private ResultSet fetchResult(Statement s) throws SQLException {
        return s.executeQuery("SELECT 1");
    }
}

PosCase6_Var4.java

// Scenario 6: A java.sql.ResultSet is locally initialized as a child of a non-locally-initialized Statement parameter, used directly, and not closed.
package scensct.var.pos;

import java.sql.ResultSet;
import java.sql.SQLException;
import java.sql.Statement;

public class PosCase6_Var4 {
    public void test(Statement stmt) throws SQLException {
        boolean hasResults = stmt.execute("SELECT 1");
        if (hasResults) {
            ResultSet rs = stmt.getResultSet();
            // rs not closed
        }
    }
}

Cause analysis

Likely Bugs/Resource Leaks/CloseSql.ql appears to be tied too closely to one acquisition shape. It handles some direct JDBC constructions, but coverage gets weaker once the returned resource is produced through a helper or through a second-stage API like getResultSet().

That leaves a real blind spot. In production JDBC code, ResultSet objects are often obtained indirectly rather than from a single inline executeQuery(...) call. The leak is still the same: the method owns a SQL resource and does not close it.