A problem with the select(2) interface

[Erik Trulsson]

On Mon, May 14, 2007 at 10:29:15PM +0800, MQ wrote:
> Hi,
> I'm writing a network application these days, for compatibility, I choose
> the select(2) interface to do event polling. When using it, I'm curious
> about the fact that the type of the final parameter is `struct timeval *'. I
> skimmed through the codes in /sys/kern/sys_generic.c, there is nothing
> written to this address. So, I think if we can use `const struct timeval *'
> instead? This type can inform the users explicitly that we do NOT modify the
> timeval struct. But I'm not sure if this modification will conflict with the
> POSIX standard. Will you please tell me your considerations? Any suggestions
> will be appreciated.
> Thanks.

Some other implementations *do* write to that adress.
>From the select(2) manpage on a Linux system:

   On Linux, select() modifies timeout to reflect the amount of time not slept; most other  imple-
   mentations do not do this.  (POSIX.1-2001 permits either behaviour.) This causes problems both
   when Linux code which reads timeout is ported to other operating systems,  and  when  code  is
   ported  to  Linux  that reuses a struct timeval for multiple select()s in a loop without reini-
   tializing it.  Consider timeout to be undefined after select() returns.

The FreeBSD select(2) manpage says:

  BUGS
    Version 2 of the Single UNIX Specification (`SUSv2'') allows systems to
    modify the original timeout in place.  Thus, it is unwise to assume that
    the timeout value will be unmodified by the select() system call.





-- 
<Insert your favourite quote here.>
Erik Trulsson
ertr1013 at student.uu.se