> Why not extend the callback to include the error code of the writev() and
> then call the callback /after/ the writev()s?
> Hmm, on second thought I guess the writes can return EAGAIN? Nonetheless,
> why not wait until the write has been known to have either succeeded or
> failed?
Our goal was just to log outgoing DNS queries. But your suggestion make sense
from my point of view. Also I think beside return code of write we can
return "struct in_addr" of DNS server which we tried.
Received on 2009-05-05