Re: nmap 6.47 compile failure on AIX 7100-03

[Daniel Miller <bonsaiviking () gmail com>]

Kevin,

I had already committed almost that exact fix in r33660. Now r33661 solves
an interaction I had with the older version of GCC that is available on the
AIX system I'm testing on. At this point, Nmap compiles but fails to scan:

./nmap --iflist

Starting Nmap 6.47SVN ( http://nmap.org ) at 2014-09-09 00:00 CDT
INTERFACES: NONE FOUND(!)
ROUTES: NONE FOUND(!)


I will continue to work on this as time permits.

Dan


On Mon, Sep 8, 2014 at 5:42 PM, Kevin Brott <kevin.brott () gmail com> wrote:

> Actually I have a fix for that already:
>
> # this patch fixes this issue ... AIX has a slightly different structure
> here ...
> ---CUT---
> *** nmap-6.46.orig/TargetGroup.cc       2013-09-11 12:06:20.000000000 -0700
> --- nmap-6.46/TargetGroup.cc    2014-08-13 15:25:00.529370342 -0700
> ***************
> *** 549,556 ****
> --- 549,561 ----
>   }
>
>   void NetBlockIPv6Netmask::apply_netmask(int bits) {
> + #ifdef _AIX
> +   const struct in6_addr zeros = { { { 0x00000000, 0x00000000,
> 0x00000000, 0x00000000 } } };
> +   const struct in6_addr ones = { { { 0xffffffff, 0xffffffff, 0xffffffff,
> 0xffffffff } } };
> + #else
>     const struct in6_addr zeros = { { { 0x00, 0x00, 0x00, 0x00, 0x00,
> 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00} } };
>     const struct in6_addr ones = { { { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
> 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff} } };
> + #endif
>     struct in6_addr mask;
>
>     if (bits > 128)
> ---CUT---
>
> On Mon, Sep 8, 2014 at 1:52 PM, Daniel Miller <bonsaiviking () gmail com>
> wrote:
>
> > On Tue, Sep 2, 2014 at 6:12 PM, Kevin Brott <kevin.brott () gmail com>
> > wrote:
> >
> > > Looks to possibly(?) be an issue with NEXTSA/ROUNDUP on this platform.
> > > Noted that 6.25 builds and runs on this platform without issue, but
> > > nothing
> > > more recent than that (including an SVN pull from today).
> > >
> > > oslevel 7100-03-03-1415 :: gcc (GCC) 4.4.7 :: GNU Make 3.82
> > >
> > > $ ./configure --prefix=/opt/nmap --mandir=/opt/nmap/man \
> > >   --with-liblinear=included --with-liblua=included \
> > >   --with-libdnet=included --with-libpcre=included
> > >   --with-libpcap=included && echo OK || echo FAIL
> > > ...
> > > Configuration complete.  Type make (or gmake on some *BSD machines) to
> > > compile.
> > > OK
> > >
> > > $ gmake
> > > ...
> > > gcc -DHAVE_CONFIG_H -I. -I../include -I../include -g -O2 -Wall -c
> > > route-bsd.c -o route-bsd.o
> > > route-bsd.c: In function 'route_msg':
> > > route-bsd.c:108: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:108: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:116: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:116: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:123: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:123: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:157: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:157: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c: In function 'route_loop':
> > > route-bsd.c:294: warning: implicit declaration of function 'getkerninfo'
> > > route-bsd.c:354: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:354: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:364: error: invalid type argument of '->' (have 'int')
> > > route-bsd.c:364: error: invalid type argument of '->' (have 'int')
> > > gmake[2]: *** [route-bsd.lo] Error 1
> > > gmake[2]: Leaving directory
> > > `/lppdir/build/nmap-6.47/libdnet-stripped/src'
> > > gmake[1]: *** [all-recursive] Error 1
> > > gmake[1]: Leaving directory `/lppdir/build/nmap-6.47/libdnet-stripped'
> > > gmake: *** [dnet_build] Error 2
> > >
> > > Searching the 'net seems to indicate an undefined function call, but
> > > determining exactly which one and why is beyond my current skillset.
> > >  Pointers?  Missed gcc or AIX networking FAQ? A polite(?!) sod-off?
> >
> > Kevin,
> >
> > Thanks for bringing this up. Turns out AIX defines RT_ROUNDUP in
> > net/route.h as:
> >
> > #define RT_ROUNDUP(sa)  ((sa)->sa_len > 0 ? \
> >                                 (1 + (((sa)->sa_len-1) |
> > (sizeof(int)-1))) : \
> >                                  sizeof(int))
> >
> > And we use RT_ROUNDUP as a replacement for ROUNDUP due to better 64-bit
> > handling on NetBSD (line 50),
> > But we are expecting a definition for ROUNDUP that fits the one on line
> > 55:
> >
> > #define ROUNDUP(a) \
> >   ((a) > 0 ? (1 + (((a) - 1) | (RT_MSGHDR_ALIGNMENT - 1))) :
> > RT_MSGHDR_ALIGNMENT)
> >
> > On line 60, NEXTSA is defined:
> >
> > #define NEXTSA(s) \
> >   ((struct sockaddr *)((u_char *)(s) + ROUNDUP((s)->sa_len))
> >
> > So when we run this through the C preprocessor (cpp -I../include
> > route-bsd.c), we get this expansion:
> >
> > sa = ((struct sockaddr *)((u_char *)(sa) + (((sa)->sa_len)->sa_len > 0 ?
> > (1 + ((((sa)->sa_len)->sa_len-1) | (sizeof(int)-1))) : sizeof(int))));
> >
> > The "->sa_len" is being applied twice! Naturally, this results in a type
> > error. As a solution, I added a check for __NetBSD__ to ensure we only use
> > RT_ROUNDUP on that platform. Should be fixed in r33657.
> >
> > Unfortunately, we still have problems:
> >
> > g++ -c -I./liblinear -I./liblua -I./libdnet-stripped/include -I./libpcre
> > -I./libpcap -I./nbase -I./nsock/include -DHAVE_CONFIG_H
> > -DNMAP_NAME=\"Nmap\" -DNMAP_URL=\"http://nmap.org\";
> > -DNMAP_PLATFORM=\"powerpc-ibm-aix7.1.0.0\"
> > -DNMAPDATADIR=\"/usr/local/share/nmap\" -D_FORTIFY_SOURCE=2 -g -O2 -Wall
> > -fno-strict-aliasing   TargetGroup.cc -o TargetGroup.o
> > TargetGroup.cc: In member function 'virtual void
> > NetBlockIPv6Netmask::apply_netmask(int)':
> > TargetGroup.cc:551: error: too many initializers for 'u_int32_t [4]'
> > TargetGroup.cc:552: error: too many initializers for 'u_int32_t [4]'
> > make[1]: *** [TargetGroup.o] Error 1
> >
> > Stay tuned for more fixes!
> >
> > Dan
>
>
>
> --
> # include <stddisclaimer.h>
> /* Kevin  Brott <Kevin.Brott () gmail com> */
_______________________________________________
Sent through the dev mailing list
http://nmap.org/mailman/listinfo/dev
Archived at http://seclists.org/nmap-dev/