undefined references when using "compileall sourcesat once"

[Rob Landley]

On Tuesday 21 March 2006 8:47 pm, Rogelio Serrano wrote:
> > I wonder if we need an "experimental new features" menu.  Build
> > everything at once would certainly go in there, if so...
> >
> > > re,
> > >   walter
> >
> > Rob
>
> That option would be great.
>
> I think the crypt problem is a bug because when i used the busybox
> loginuitls there was no error. I was using shadow linked with
> libxcrypt. Dropping the lcrypt dependency seems better IMHO.

Currently, I'm trying to work my way through the logic of gnu crypt, or at 
least the man page thereof.  Busybox has built-in md5 code, and built-in sha1 
code.  The gnu version of crypt has support for $1$ handling crypt, but 
apparently not $2$ handling sha1.  I _was_ using the busybox versions for 
both, but I'm wondering if I should pass through $1$ to glibc?  (Not that 
this helps us for md5sum, so we'd still need that code in there...)

Sigh.  I generally find solving problems easier than understanding them in the 
first place.  Luckily this is _much_ easier to deal with than mount was. :)

Anyway, using the built-in md5 or sha1 code and configuring out support for 
DES, we can should be able to build without -lcrypt.  FYI.  It's not finished 
yet...

Rob
-- 
Never bet against the cheap plastic solution.