<div dir="ltr">Another common source of apparently random results is uninitialized variables. I don't think it is in ctest itself, as this is the first we've heard of something like this on this mailing list (although anything's possible...)<div>
<br></div><div>Have you run your test code through valgrind to see if it catches any uninitialized data access?</div><div><br></div><div><br></div><div>David</div><div><br><br><div class="gmail_quote">On Fri, Oct 17, 2008 at 11:37 PM, Ian Mackenzie <span dir="ltr"><<a href="mailto:ian.e.mackenzie@gmail.com">ian.e.mackenzie@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><div class="Ih2E3d">On Fri, Oct 17, 2008 at 10:09 PM, Alan W. Irwin<br>
<<a href="mailto:irwin@beluga.phys.uvic.ca">irwin@beluga.phys.uvic.ca</a>> wrote:<br>
> One potential source of randomness is failing or iffy hardware.<br>
><br>
> Have you kept track of such things as systematic changes in system<br>
> temperature under the same load conditions? I must clean out the dust from<br>
> one of my boxes every 6 months or I see dramatically increasing temperatures<br>
> (like 10 C) under heavy system load. The other box has much better air flow<br>
> design so the the problem isn't as critical there.<br>
><br>
> Have you tried running memtest?<br>
><br>
> What happens if you try running your ctests on an entirely different<br>
> machine?<br>
<br>
</div>Thanks for the suggestions! I don't think system temperature has much<br>
to do with it, though - I did a bunch more tests keeping track of<br>
temperature at the same time (acpi -t && ctest) and there didn't seem<br>
to be a correlation between temperature and errors...both cores were<br>
pretty steady around 56-57 C. I also just ran memtest, which reported<br>
no errors.<br>
<br>
I don't have a convenient way of testing on another machine right now;<br>
while I suspect the problem is somehow dependent on my machine and<br>
might very well not show up on another, the errors also seemed to<br>
start cropping up as soon as I upgraded to 2.6.2, so I think it's<br>
somehow related to CMake. It might just be a packaging thing with my<br>
distribution, but I tried with the binary version of CMake available<br>
on the CMake web site as well and I get the same sort of errors (even<br>
after wiping the build directory clean and rebuilding everything with<br>
the binary version of CMake and then testing with the binary version<br>
of CTest).<br>
<font color="#888888"><br>
-Ian<br>
</font><div><div></div><div class="Wj3C7c">_______________________________________________<br>
CMake mailing list<br>
<a href="mailto:CMake@cmake.org">CMake@cmake.org</a><br>
<a href="http://www.cmake.org/mailman/listinfo/cmake" target="_blank">http://www.cmake.org/mailman/listinfo/cmake</a><br>
</div></div></blockquote></div><br></div></div>