Re: [nmap-svn] r14271 - nmap-exp/ithilgore/ncrack/nbase

[ithilgore <ithilgore.ryu.l () gmail com>]

Fyodor wrote:
> On Tue, Jul 14, 2009 at 07:30:28AM -0700, commit-mailer () insecure org wrote:
> > Log:
> >
> > OK the nbase configure straight from Nmap trunk won't do
> > out-of-the-box but for some reason the autoconf tool gives quite a
> > different output configure between ncrack-nmap. I will have to examine
> > what is the cause of that.
>
> Well, Nbase really should be able to handle being under a different
> parent directory.  So if it doesn't, I'd consider it a bug.  If you
> make a patch which makes /nbase generic, that would be great.  It
> beats having a slightly changed nbase just for ncrack.

Here's what I have found so far:

For some reason that I am not currently aware of, compilation of Ncrack requires
that the -l linking flags of nsock and nbase be in that order: first nsock and
then nbase (meaning -lnsock -lnbase). However, Nmap has them in the order: first
nbase and then nsock. If Ncrack specifies them in the order -lnbase -lnsock then
the linking phase fails complaining that no symbols can be found for some
functions that are called in nsock but have been defined in nbase (like
get_random_bytes).

Putting that aside, with the current order (-lnsock -lnbase), during the first
compilation of Ncrack from a brand-new svn checkout, happens the following:
Normally by typing ./configure the configure scripts of *all* libraries are
being run (nbase, nsock and opensshlib in that order). However, when you then
call 'make' to start the compilation procedure, the configure scripts of nbase
and nsock are called again (something which shouldn't be really happening). In
particular, this is the output of when makes start running:

cd nbase && make


make[1]: Entering directory `/home/ithilgore/la/ncrack/nbase'


autoconf


./config.status --recheck


running CONFIG_SHELL=/bin/sh /bin/sh ./configure  --disable-option-checking
--prefix=/usr/local --cache-file=/dev/null --srcdir=.  --no-create
--no-recursion
checking for gcc... gcc


checking for C compiler default output file name... a.out


checking whether the C compiler works... yes


checking whether we are cross compiling... no


checking for suffix of executables...
....
....

As you see, autoconf is called there again, something which as a result creates
a new configure script for nbase. I think this has something to do with the
./config-status --recheck line.

The above behaviour of configure scripts running two times, happens only with
the order of -lnsock -lnbase. However, as I mentioned aboved, changing that
order creates linking problems. So, we've got 2 different paths to choose:

1) Solve the linking problems when specifying -lnbase -lnsock. Everything else
seems to work fine from there on.
OR
2) See why the configure scripts run twice when specifying -lnsock -lnbase and
find a workaround.

I would be really glad, if someone, who has a little more experience on these
slightly obscure building systems, could shed some light on this matter.

-- ithilgore






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