BitwiseSignCheck.qhelp

<!DOCTYPE qhelp PUBLIC
  "-//Semmle//qhelp//EN"
  "qhelp.dtd">
<qhelp>


<overview>
<p>This rule finds code that checks the sign of the result of a bitwise operation. Such a check may yield unexpected results. As an example, consider the following code that checks if the <code>n</code>th bit of a variable <code>x</code> is set: </p>

<pre>  x &amp; (1 &lt;&lt; n) > 0 </pre>

<p>If <code>x</code> is a 32-bit signed integer, the value of <code>x &amp; (1 &lt;&lt; 31)</code> is interpreted as a signed number. If <code>x</code> is negative (that is, its sign bit is set), and <code>n</code> is 31, then <code>x &amp; (1 &lt;&lt; 31)</code> evaluates to <code>0x80000000</code> (all bits zero except the sign bit). The sign check on this value fails, implying that the 31st bit of <code>x</code> is unset. This is clearly incorrect.</p>

</overview>
<recommendation>
<p>The above sign check should be rewritten as</p>

<pre>  x &amp; (1 &lt;&lt; n) != 0</pre>

</recommendation>
<references>

<li>
  Code Project: <a href="http://www.codeproject.com/Articles/2247/An-introduction-to-bitwise-operators">An introduction to bitwise operators</a>
</li>
<li>
  MSDN Library: <a href="https://docs.microsoft.com/en-us/cpp/c-language/signed-bitwise-operations">Signed Bitwise Operations</a>
</li>






</references>
</qhelp>