Hello all,<div><br></div><div>test_math is currently failing on the Debian/alpha buildbots (trunk and py3k).</div><div>I&#39;ve been trying, unsuccessfully, to figure out what&#39;s going wrong, and I&#39;m</div><div>starting&nbsp;to run out&nbsp;of ideas, so I thought I&#39;d ask the list for help to see if</div>
<div>anyone has any useful&nbsp;suggestions.</div><div><br></div><div>Details of the failure: &nbsp;running the following code:</div><div><br></div><div>from math import log</div><div>ar = 9.88e-324</div><div>x = log(ar)</div><div>
<br></div><div>produces:</div><div><br></div><div>ValueError: math domain error<br></div><div><br></div><div>Somehow, it looks like errno is getting set to something nonzero in</div><div>math_1_from_whatever in mathmodule.c, but I really can&#39;t figure out</div>
<div>how.</div><div><br></div><div>I&#39;ve tried adding &#39;-mieee&#39; to the gcc compile flags, and I&#39;ve added a bunch</div><div>of autoconf tests to verify that log(9.88e-324) succeeds, produces</div><div>roughly the right result, and doesn&#39;t set errno to anything nonzero. &nbsp;</div>
<div>All the autoconf tests that should pass do. &nbsp;So&nbsp;I can&#39;t find anything</div><div>wrong with the libm implementation of log.</div><div><br></div><div>test_math is fine on Tru64/alpha.</div><div><br></div><div>Does anyone have access to a Linux/alpha machine, and a few minutes</div>
<div>to figure out exactly what&#39;s failing? &nbsp;Or any suggestions about what might</div><div>be failing. &nbsp;I&#39;m open to wild ideas at this point... :-)</div><div><br></div><div>Mark</div><div><br></div>