Re: nmap 5 + 64-bit Solaris x86

[David Fifield <david () bamsoftware com>]

On Thu, Dec 02, 2010 at 12:38:11AM -0500, John Cecere wrote:
> I found a bug in compiling nmap 5.21 on Solaris x86. I'm not sure if it affects SPARC as well.
>
> In file included from nbase.h:452,
>                  from snprintf.c:66:
> nbase_ipv6.h:143: error: conflicting types for 'inet_ntop'
> /usr/include/arpa/inet.h:69: error: previous declaration of 'inet_ntop' was here
>
> The reason for the conflict seems to be that the last arg to
> inet_ntop is defined as a size_t in nbase_ipv6.h whereas Solaris has
> it defined as a socklen_t in arpa/inet.h. In 32-bit mode, these data
> types are compatible, but in 64-bit mode they're not:
>
> # gcc -o tst tst.c
> # ./tst
> sizeof socklen_t = 4
> sizeof size_t    = 4
>
> # gcc -m64 -o tst tst.c
> # ./tst
> sizeof socklen_t = 4
> sizeof size_t    = 8
>
> Defining HAVE_INET_NTOP doesn't fix this as this only affects the prototype definition.
>
> Hacking the last arg as a socklen_t in both the prototype and
> function almost allowed me to get it to build a 64-bit binary, but
> the final linking appeared to ignore my setting for CXXFLAGS and
> barfed on an "invalid ELF class" error. I had to set
>
> CXX="g++ -m64"
>
> to force it to do the final linking as 64-bit.
>
> BTW, I've been a fan of nmap for many years. Thanks for a great piece of software.

Thanks for reporting this and for doing the research. Can you send me
your config.log file? I'm interested in seeing these output lines:

checking for inet_ntop... yes
checking for inet_pton... yes

On my GNU/Linux system, inet_pton.c never even gets built because it's
not necessary. Possibly inet_ntop is defined but not inet_pton.

David Fifield
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/