<div dir="ltr">+kcc, samsonov (please don't remove people from CC)<div class="gmail_extra"><br></div><div class="gmail_extra">You mean in the presence of threads? There's no such option because it's not supposed to interfere with the symbolizer. If it does then it's a bug, someone from our team will follow up on this tomorrow.</div>
<div class="gmail_extra"><br></div><div class="gmail_extra">Sergey</div><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Nov 15, 2013 at 2:01 AM, Rick Foos <span dir="ltr"><<a href="mailto:rfoos@codeaurora.org" target="_blank">rfoos@codeaurora.org</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    <div>Thank you Sergey!<br>
      <br>
      Address Sanitize running alone on a server is stable without the
      symbolizer option. It is running all the tests in a reasonable
      amount of time, and there are no llvm-symbolizer tasks.<br>
      <br>
      The problem is coming from Threads, and I'm trying to prove that
      now. <br>
      <br>
      If threads runs clean by itself alone on a server, there is an
      interaction with both address and threads running at the same
      time.<br>
      <br>
      Is there a similar feature to disable symbolizer in threads?<br>
      <br>
      Best Regards,<br>
      Rick<div><div class="h5"><br>
      <br>
      On 11/14/2013 03:51 PM, Sergey Matveev wrote:<br>
    </div></div></div><div><div class="h5">
    <blockquote type="cite">
      <div dir="ltr">ASAN_OPTIONS=symbolize=false<br>
      </div>
      <div class="gmail_extra"><br>
        <br>
        <div class="gmail_quote">On Fri, Nov 15, 2013 at 1:14 AM, Nick
          Kledzik <span dir="ltr"><<a href="mailto:kledzik@apple.com" target="_blank">kledzik@apple.com</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div style="word-wrap:break-word"><br>
              <div>
                <div>
                  <div>On Nov 14, 2013, at 9:07 AM, Rick Foos <<a href="mailto:rfoos@codeaurora.org" target="_blank">rfoos@codeaurora.org</a>>
                    wrote:</div>
                  <br>
                  <blockquote type="cite">
                    <div bgcolor="white" link="blue" vlink="purple" style="font-family:Helvetica;font-size:12px;font-style:normal;font-variant:normal;font-weight:normal;letter-spacing:normal;line-height:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px" lang="EN-US">

                      <div>
                        <div><span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)">Status:
                            System in swap overnight. Stopped both
                            buildmaster and slave. 187 llvm-symbolizer
                            tasks were still running. Tasks did not stop
                            after</span></div>
                        <div><span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)"> </span></div>
                        <div>
                          <span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)">Retried
                            this morning, no other workload, 8
                            llvm-symbolizer tasks consuming 100% on each
                            cpu</span></div>
                      </div>
                    </div>
                  </blockquote>
                  <div><br>
                  </div>
                </div>
                <div>Doesn’t that mean that Asan found some problems,
                  but is stuck trying to symbolicate the backtraces?  
                  Is there a way to run Asan and *not* symbolicate?  </div>
                <div><br>
                </div>
                <div>This also seems like a bug (infinite loop?) in
                  llvm-symbolizer.  </div>
                <span><font color="#888888">
                    <div><br>
                    </div>
                    <div>-Nick</div>
                  </font></span>
                <div>
                  <div>
                    <div><br>
                    </div>
                    <br>
                    <blockquote type="cite">
                      <div bgcolor="white" link="blue" vlink="purple" style="font-family:Helvetica;font-size:12px;font-style:normal;font-variant:normal;font-weight:normal;letter-spacing:normal;line-height:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px" lang="EN-US">

                        <div>
                          <div><span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)">.
                              7 zombie tasks.</span></div>
                          <div><span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)"> </span></div>
                          <div>
                            <span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)">So
                              not quite ready this morning. If anyone
                              knows of an llvm-sanitizer issue like this
                              it would help.</span></div>
                          <div>
                            <span style="font-size:11pt;font-family:Calibri,sans-serif;color:rgb(31,73,125)"> </span></div>
                          <div>
                            <div style="border-style:solid none none;border-top-color:rgb(181,196,223);border-top-width:1pt;padding:3pt 0in 0in">
                              <div>
                                <b><span style="font-size:10pt;font-family:Tahoma,sans-serif;color:windowtext">From:</span></b><span style="font-size:10pt;font-family:Tahoma,sans-serif;color:windowtext"><span> </span><a href="mailto:llvm-commits-bounces@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">llvm-commits-bounces@cs.uiuc.edu</a><span> </span>[<a href="mailto:llvm-commits-bounces@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">mailto:llvm-commits-bounces@cs.uiuc.edu</a>]<span> </span><b>On
                                    Behalf Of<span> </span></b>Rick Foos<br>
                                  <b>Sent:</b><span> </span>Wednesday,
                                  November 13, 2013 1:42 PM<br>
                                  <b>To:</b><span> </span>Sergey
                                  Matveev; Shankar Easwaran<br>
                                  <b>Cc:</b><span> </span><a href="mailto:llvm-commits@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">llvm-commits@cs.uiuc.edu</a>;
                                  Galina Kistanova<br>
                                  <b>Subject:</b><span> </span>Re: [lld]
                                  r194545 - [PECOFF] Fix
                                  use-after-return.</span></div>
                            </div>
                          </div>
                          <div> </div>
                          <div>
                            <div>Sorry for the delay,<span> </span><br>
                              <br>
                              Our problem with running the sanitizers is
                              that the load average running under Ninja
                              reached 146 and a short time after a
                              system crash requiring calling someone to
                              power cycle the box...<br>
                              <br>
                              The address sanitizer by itself leaves a
                              load average 40. This means the OS over
                              100% utilization, and is thrashing a bit.
                              Load Average doesn't say what exactly is
                              thrashing.<br>
                              <br>
                              Ninja supports make's -j, and -l options.
                              The -l maximum load average, is the key.<span> </span><br>
                              <br>
                              The load average should be less than the
                              total number of cores (hyperthreads too)
                              before Ninja launches another task.<span> </span><br>
                              <br>
                              A Load Average at or lower than 100% 
                              technically should benefit performance,
                              and maximize throughput. However, I will
                              be happy if I don't have to call someone
                              to power cycle the server :)<br>
                              <br>
                              So the maximum load average of a 16 core
                              machine with hyperthreads is 32 (keeping
                              it simple). This needs to be passed to all
                              make's and Ninja build steps on that slave
                              to maximize throughput.<br>
                              <br>
                              For now, I'm looking at a minimal patch to
                              include jobs and a new loadaverage
                              variable for the sanitizers.<span> </span><br>
                              <br>
                              Longer term, all buildslaves should define
                              maximum loadaverage, and all make/ninja
                              steps should pass -j, and -l options.<br>
                              <br>
                              Best Regards,<br>
                              Rick<br>
                              <br>
                              On 11/13/2013 11:21 AM, Sergey Matveev
                              wrote:</div>
                          </div>
                          <blockquote style="margin-top:5pt;margin-bottom:5pt">
                            <div>
                              <div>+kcc</div>
                            </div>
                            <div>
                              <p class="MsoNormal">
                                 </p>
                              <div>
                                <div>On Wed, Nov 13, 2013
                                  at 6:41 AM, Shankar Easwaran <<a href="mailto:shankare@codeaurora.org" style="color:purple;text-decoration:underline" target="_blank">shankare@codeaurora.org</a>>
                                  wrote:</div>
                                <div>Sorry for another
                                  indirection. Rick foos is working on
                                  it. I think there is some good news
                                  here :)<br>
                                  <br>
                                  Cced Rick + adding Galina,Dmitri.<br>
                                  <br>
                                  Thanks<br>
                                  <br>
                                  Shankar Easwaran</div>
                                <div>
                                  <div><br>
                                    <br>
                                    On 11/12/2013 8:37 PM, Rui Ueyama
                                    wrote:</div>
                                  <p class="MsoNormal">Shankar tried to
                                    set it up recently.<br>
                                    <br>
                                    <br>
                                    On Tue, Nov 12, 2013 at 6:31 PM,
                                    Sean Silva <<a href="mailto:silvas@purdue.edu" style="color:purple;text-decoration:underline" target="_blank">silvas@purdue.edu</a>>
                                    wrote:</p>
                                  <p class="MsoNormal">Sanitizers?<br>
                                    <br>
                                    There have been a couple of these
                                    sorts of bugs recently... we really<br>
                                    ought to have some sanitizer bots...<br>
                                    <br>
                                    -- Sean Silva<br>
                                    <br>
                                    <br>
                                    On Tue, Nov 12, 2013 at 9:21 PM, Rui
                                    Ueyama <<a href="mailto:ruiu@google.com" style="color:purple;text-decoration:underline" target="_blank">ruiu@google.com</a>>
                                    wrote:</p>
                                  <p class="MsoNormal">Author: ruiu<br>
                                    Date: Tue Nov 12 20:21:51 2013<br>
                                    New Revision: 194545<br>
                                    <br>
                                    URL:<span> </span><a href="http://llvm.org/viewvc/llvm-project?rev=194545&view=rev" style="color:purple;text-decoration:underline" target="_blank">http://llvm.org/viewvc/llvm-project?rev=194545&view=rev</a><br>

                                    Log:<br>
                                    [PECOFF] Fix use-after-return.<br>
                                    <br>
                                    Modified:<br>
                                       
                                     lld/trunk/lib/Driver/WinLinkDriver.cpp<br>
                                    <br>
                                    Modified:
                                    lld/trunk/lib/Driver/WinLinkDriver.cpp<br>
                                    URL:<br>
                                    <a href="http://llvm.org/viewvc/llvm-project/lld/trunk/lib/Driver/WinLinkDriver.cpp?rev=194545&r1=194544&r2=194545&view=diff" style="color:purple;text-decoration:underline" target="_blank">http://llvm.org/viewvc/llvm-project/lld/trunk/lib/Driver/WinLinkDriver.cpp?rev=194545&r1=194544&r2=194545&view=diff</a><br>

                                    <br>
==============================================================================<br>
                                    ---
                                    lld/trunk/lib/Driver/WinLinkDriver.cpp
                                    (original)<br>
                                    +++
                                    lld/trunk/lib/Driver/WinLinkDriver.cpp
                                    Tue Nov 12 20:21:51 2013<br>
                                    @@ -842,7 +842,7 @@
                                    WinLinkDriver::parse(int argc, const
                                    cha<br>
                                    <br>
                                          case OPT_INPUT:<br>
                                           
                                    inputElements.push_back(std::unique_ptr<InputElement>(<br>
                                    -          new PECOFFFileNode(ctx,
                                    inputArg->getValue())));<br>
                                    +          new PECOFFFileNode(ctx,<br>
ctx.allocateString(inputArg->getValue()))));<br>
                                            break;<br>
                                    <br>
                                      #define DEFINE_BOOLEAN_FLAG(name,
                                    setter)       \<br>
                                    @@ -892,9 +892,11 @@
                                    WinLinkDriver::parse(int argc, const
                                    cha<br>
                                        // start with a hypen or a
                                    slash. This is not compatible with
                                    link.exe<br>
                                        // but useful for us to test lld
                                    on Unix.<br>
                                        if (llvm::opt::Arg *dashdash =
                                    parsedArgs->getLastArg(OPT_DASH_DASH))
                                    {<br>
                                    -    for (const StringRef value :
                                    dashdash->getValues())<br>
                                    -      inputElements.push_back(<br>
                                    -        
                                     std::unique_ptr<InputElement>(new
                                    PECOFFFileNode(ctx, value)));<br>
                                    +    for (const StringRef value :
                                    dashdash->getValues()) {<br>
                                    +    
                                     std::unique_ptr<InputElement>
                                    elem(<br>
                                    +          new PECOFFFileNode(ctx,
                                    ctx.allocateString(value)));<br>
                                    +    
                                     inputElements.push_back(std::move(elem));<br>
                                    +    }<br>
                                        }<br>
                                    <br>
                                        // Add the libraries specified
                                    by /defaultlib unless they are
                                    already<br>
                                    added<br>
                                    <br>
                                    <br>
_______________________________________________<br>
                                    llvm-commits mailing list<br>
                                    <a href="mailto:llvm-commits@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">llvm-commits@cs.uiuc.edu</a><br>
                                    <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" style="color:purple;text-decoration:underline" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a></p>

                                  <div>
                                     </div>
                                  <p class="MsoNormal"> </p>
                                </div>
                                <div>
                                  <span><span style="color:rgb(136,136,136)">--<span> </span></span></span><span style="color:rgb(136,136,136)"><br>
                                    <span>Qualcomm Innovation Center,
                                      Inc. is a member of Code Aurora
                                      Forum, hosted by the Linux
                                      Foundation</span></span></div>
                                <div>
                                  <div><br>
                                    <br>
_______________________________________________<br>
                                    llvm-commits mailing list<br>
                                    <a href="mailto:llvm-commits@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">llvm-commits@cs.uiuc.edu</a><br>
                                    <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" style="color:purple;text-decoration:underline" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a></div>

                                </div>
                              </div>
                              <div>
                                 </div>
                            </div>
                            <div><br>
                              <br>
                              <br>
                            </div>
                            <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'">_______________________________________________</pre>
                            <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'">llvm-commits mailing list</pre>
                            <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'"><a href="mailto:llvm-commits@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">llvm-commits@cs.uiuc.edu</a></pre>

                            <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'"><a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" style="color:purple;text-decoration:underline" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a></pre>

                          </blockquote>
                          <div>
                            <br>
                            <br>
                            <br>
                          </div>
                          <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'">-- </pre>
                          <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'">Rick Foos</pre>
                          <pre style="margin:0in 0in 0.0001pt;font-size:10pt;font-family:'Courier New'">Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation</pre>
                        </div>
                        _______________________________________________<br>
                        llvm-commits mailing list<br>
                        <a href="mailto:llvm-commits@cs.uiuc.edu" style="color:purple;text-decoration:underline" target="_blank">llvm-commits@cs.uiuc.edu</a><br>
                        <a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" style="color:purple;text-decoration:underline" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a><br>
                      </div>
                    </blockquote>
                  </div>
                </div>
              </div>
              <br>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
    <br>
    <pre cols="72">-- 
Rick Foos
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation</pre>
  </div></div></div>

</blockquote></div><br></div></div>