<div class="moz-cite-prefix">On 05/31/14 20:44, johnlumby wrote:<br /></div><blockquote
cite="mid:538A776C.2060006@hotmail.com"type="cite"><div class="moz-cite-prefix">On 05/30/14 09:36, Claudio Freire
wrote:<br/></div><br /><font face="Nimbus Roman No9 L">Good point. I have included the guts of your little test
program<br/> (modified to do polling) into the existing autoconf test program<br /> that decides on the <br /> #define
USE_AIO_ATOMIC_BUILTIN_COMP_SWAP.<br/> See config/c-library.m4.<br /> I hope this goes some way to answer your concern
aboutrobustness,<br /> as at least now if the implementation changes in some way that<br /> renders the polling
ineffective, it will be caught in configure.<br /><br /></font><br /></blockquote><font face="Nimbus Roman No9 L">I
meantto add that by including this test, which involves a fork(),<br /> in the autoconf tester, on Windows<br />
USE_AIO_ATOMIC_BUILTIN_COMP_SWAPwould always by un-defined.<br /> (But could then be defined manually if someone wanted
togive it a try)<br /></font><br />
