-current vs 0.53

Frediano Ziglio freddyz77 at tin.it
Fri May 2 19:13:14 EDT 2003


Il gio, 2002-09-05 alle 23:37, Brian Bruns ha scritto:
> 
> is errno 35 on your box EWOULDBLOCK? (check /usr/include/errno.h or the
> equivalent).  If so then something about setting the socket to
> non-blocking is possibly wrong, but a quick perusal of the code doesn't
> show any major changes since 0.53 in this area...any of you BSD folks care
> to comment?
> 
> Brian
> 

Good spot !
This may occur if you set a connect timeout and no other timeout.
If in your freetds.conf you set "connect timeout = 10" your socket will
be not blocking, so every read/write with no kernel buffer available
will report this EAGAIN. I don't know why it core...
I think best solution is to check buffer availability even if no timeout
(aka call always select before read). Best way to accomplish this is
join code with timeout and code without timeout...

Emile, is this spot right ? Can you try to remove timeout and retest ?

freddy77





More information about the FreeTDS mailing list