Re: FD_ISSET replacement

[Simon Josefsson]

Bruno Haible <address@hidden> writes:

> Simon Josefsson wrote:
>> The sys_socket module generates this warning:
>> 
>> ../lgl/sys/socket.h: In function 'rpl_fd_isset':
>> ../lgl/sys/socket.h:158: warning: comparison between signed and unsigned
>> ../lgl/sys/socket.h:159: warning: comparison between signed and unsigned
>> 
>> I pushed the patch below.
>> 
>> However, is the code really appropriate?  I'm thinking of this comment:
>> 
>>  /* Re-define FD_ISSET to avoid a WSA call while we are not using
>>     network sockets.  */
>
> This comments means IMO that when someone is using select() on pipes and
> console handles, he should not make ws2_32 calls.

Ah, I see.

>>  static inline int
>> -rpl_fd_isset (int fd, fd_set * set)
>> +rpl_fd_isset (SOCKET fd, fd_set * set)
>>  {
>> -  int i;
>> +  u_int i;
>>    if (set == NULL)
>>      return 0;
>
> I agree with the signature change, because this FD_ISSET replacement is
> called by winsock-select.c. For the type of 'i', better use a portable type,
> such as 'unsigned int'. 'u_int' is not a portable type - it may be present in
> mingw and absent in MSVC's headers, who knows. Proposed patch below.

Actually, u_int is the official Windows API type for the struct member,
see:

http://msdn.microsoft.com/en-us/library/ms737873.aspx

I can't seem to find documentation for u_int, so I'm not sure if it is
guaranteed to be the same as 'unsigned int'.

/Simon

> Bruno
>
>
> --- lib/sys_socket.in.h.orig  2008-10-23 03:08:52.000000000 +0200
> +++ lib/sys_socket.in.h       2008-10-23 03:08:42.000000000 +0200
> @@ -122,7 +122,8 @@
>  static inline int
>  rpl_fd_isset (SOCKET fd, fd_set * set)
>  {
> -  u_int i;
> +  unsigned int i;
> +
>    if (set == NULL)
>      return 0;
>