On Mon, Feb 21, 2000 at 05:17:45PM +0100, Marcus Brinkmann wrote:

> > prevents me from successfully cross-compiling serverboot.
> 
> Or any other static binary, most notably the linker and the root
> filesystems!

Actually, some limited testing (Mostly involving printf("Hello 
world!\n");) has shown me that not all static binaries fail.

> The error is in the specs file.

I've been thinking about this.  When I run a diff between 
/usr/local/lib/gcc-lib/i586-pc-gnu/2.95.2/specs and 
/gnu/lib/gcc-lib/i586-pc-gnu/2.95.2 the only difference is a '1' instead 
of a '0' in the cross-compile section.  This is consistant with other 
cross-compilers that I have set up.  The specs modifications in 
`make-cross' are only to point to the new paths that get setup.  This 
isn't an issue because I'm actually building the whole toolchain, not 
tying it to my Linux environment.

I beleive the C Library to be good, because it is the one that I copy 
into place on the system.  It could be that there is a file that I'm not 
generating that exists from the Debian setup, however, that is causing me 
grief.  Especially if it is only core Hurd binaries that cannot be built 
static.

[I haven't tried tying it to my Linux environment, although it would be an 
interesting experiement to try...  Maybe after my exam on Sat.]

> > Also note that 
> > you should cross-compile the Hurd with egcs.  Roland has been working on 
> > making pfinet cross-compile with gcc, but the lst time I tried the latest 
> > pfinet, I had to recover from a different OS... =(
> 
> It should be fairly stable by now.

No real luck so far, compiling natively with gcc-2.95.2.  pfinet will 
resolve but not connect.  (And I still haven't scribbled down the proc 
failure, but I always substitute a recent one compiled with egcs)

Oh hold on, there were some Glibc changes too, weren't there?  I'll build
a new one and try that. 

Tks,
Jeff Bailey

-- 
"Backwards compatibility is nice, but preserving every undocumented quirk
that nobody sane would use... Sorry, but we really need an addition to
errno.h: EBITEME. Exactly for such cases." 
 -- Alexander Viro

Reply via email to