Peter Simons writes:

> >Number:         6185
> >Category:       misc
> >Synopsis:       Warning when using ntohs() and friends
> >Confidential:   no
> >Severity:       non-critical
> >Priority:       medium
> >Responsible:    misc-bug-people (Misc Bug People)
> >State:          open
> >Class:          change-request
> >Submitter-Id:   net
> >Arrival-Date:   Mon Sep 21 04:20:00 1998
> >Last-Modified:
> >Originator:     Peter Simons
> >Organization:
> 	
> 	 int i = ntohs(htons(0));
> 
> with egcs 1.1 will give the following warnings:
> 
>  | simons@peti:/tmp$ gcc -ansi -pedantic test.c -o test
>  | test.cpp: In function `int main()':

This patch fixes this problem.  There was also a send-pr for this as
kern/3562, but its proposed fix reduces the produced code efficiency.
This fix is less pedantic though :) (That is, it can miss
pedantic-warning generating code passed as a parameter to htons() and
not generate a warning.  I doubt anyone cares. :)

According to kern/3562 it's a problem for pc532.  I don't have that
file here, but the fix will be pretty similar.

--- sys/arch/i386/include/endian.h.orig Sat Aug 15 23:12:00 1998
+++ sys/arch/i386/include/endian.h      Sat Sep 26 18:23:04 1998
@@ -116,14 +116,14 @@
 
 #if defined(_KERNEL) && !defined(_LKM) && !defined(I386_CPU)
 #define        __byte_swap_long_variable(x) \
-({ register in_addr_t __x = (x); \
+__extension__ ({ register in_addr_t __x = (x); \
    __asm ("bswap %1" \
        : "=r" (__x) \
        : "0" (__x)); \
    __x; })
 #else
 #define        __byte_swap_long_variable(x) \
-({ register in_addr_t __x = (x); \
+__extension__ ({ register in_addr_t __x = (x); \
    __asm ("rorw $8, %w1\n\trorl $16, %1\n\trorw $8, %w1" \
        : "=r" (__x) \
        : "0" (__x)); \
@@ -131,7 +131,7 @@
 #endif /* _KERNEL && ... */
 
 #define        __byte_swap_word_variable(x) \
-({ register in_port_t __x = (x); \
+__extension__ ({ register in_port_t __x = (x); \
    __asm ("rorw $8, %w1" \
        : "=r" (__x) \
        : "0" (__x)); \