Ananth Ananthanarayanan wrote: > > I sent that mail again to kaio@xxxxxxxxxxx, but > still haven't received even my copy, so I'm > trying to send it to you directly. Please > let me know if you get this message. > > thanks, > > ananth. > > ------------------------------------------------------------------------ > > Chris Smith's program to test KAIO & sockets brought > out a problem in the interaction between KAIO code > and socket code. In a way the problem is similar > to the one already fixed in the PIPE code. This > time, though, I think it should be OK to change > the KAIO code that calls into the socket layer > (also the same KAIO code that calls into PIPE). > > The chages are only a couple of lines of code, so > I'm going to append it to the end of this message. > > Finally, noticed that the socket test was calling > aio_return() without first checking aio_error(), > whereas, aio_error() should be called first, > and only when it returns 0 can aio_return be called. > Note that till the I/O complete aio_error() can > return EINPROGRESS, although in Chris' test aio_suspend > precludes this possiblity. Added the code for aio_error(). Thanks for pointing it out. > Chris, can you let me know if you are able to apply > the following fix & if it works for you? I applied the 2 line change you sent (didn't try the patch but I assume it's the same). It appears to work wonderfully. --Chris P.S.: I attached an updated version of the aio2.cpp code. |