Simon Josefsson <[email protected]> writes:
> Hmm, you make a good case for this, and I'm changing my position to
> neutral. What I think is more important is to align ping and ping6 more
> to reduce code duplication. If that happens to lead to using normal
> getaddrinfo() selection defaults for IPv4 vs IPv6, that is probably a
> good thing for system-wide consistency.
Speaking of code duplication, which I agree is good to get rid of.
I had a look at porting traceroute to support IPv6 a long time ago. But
see this comment:
/* #include <netinet/ip_icmp.h> -- Deliberately not including this
since the definitions in use are being pulled in by libicmp. */
#ifdef HAVE_NETINET_IP_VAR_H
# include <netinet/ip_var.h>
#endif
It would be nice to get rid of the headers in libicmp and add networking
headers like netinet/ip_icmp.h and netinet/icmp6.h to Gnulib. That would
make life much easier in my opinion.
It would also help for MinGW which does not have many networking
headers. For example, tftp.h, telnet.h, and the list goes on...
Collin