<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">On 1/26/2017 9:00 AM, Peter Bel wrote:<br>
    </div>
    <blockquote
cite="mid:CAKzDyZsd-hZeY1Mo3biXT=RVKhAgjrH_ibYs_RnQP5-cJ5PzWA@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>Sorry for putting this topic up, but can I ask for your
          advice once again? <br>
          I'm getting some very strange errors and I can't really
          understand why. In my <span
            style="font-family:monospace,monospace">applyFixup() </span>I
          have fixup value adjustment, and for fixup <span
            style="font-family:monospace,monospace">PCREL24 </span>I
          have the following simple code:<br>
        </div>
        <div><br>
          <span style="font-family:monospace,monospace">case
            Epiphany::fixup_Epiphany_PCREL24<br>
              Value = Value << 7;<br>
          </span></div>
        <div><span style="font-family:monospace,monospace">  break;</span><br>
        </div>
        <div><br>
        </div>
        <div>When I'm trying to compile a sample binary (any, even
          without this fixup type), it "might fail" on <span
            style="font-family:monospace,monospace">MC/ELFObjectWriter.cpp::executePostLayoutBinding():392</span>,
          which is called from <span
            style="font-family:monospace,monospace">MCAssembler.cpp::layout()</span>.
          If fails on the following line:<br>
        </div>
        <div><span style="font-family:monospace,monospace">const auto
            &Alias = cast<MCSymbolELF>(A);<br>
            <br>
          </span></div>
        <div>"Might fail" - because it doesn't fail every time. I can
          change some other completely unrelated piece of code - and
          suddenly it'll fail. If I'll change it to smth else, e.g. <span
            style="font-family:monospace,monospace">Value = Value</span>,
          it won't fail, but it still "might fail" after some other
          change. Sometimes it can fail on different lines, sometimes it
          can even pass through on the first run and fail on the second.<br>
          <br>
          I've checked that it fails because <span
            style="font-family:monospace,monospace">MCSymbol &A </span>has
          <span style="font-family:monospace,monospace">A.Kind </span>different
          from 2 (<span style="font-family:monospace,monospace">SymbolKindELF</span>),
          usually it's 1 or 0 (COFF or Unknown).<br>
        </div>
      </div>
    </blockquote>
    <br>
    This sounds like memory corruption of some sort... have you tried
    asan and/or valgrind?<br>
    <br>
    -Eli<br>
    <pre class="moz-signature" cols="72">-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project</pre>
  </body>
</html>