> In any case I'm trying to make Milo sources to serve as a framework
> for a future SGI bootstrap loader, but is it really usefull ?
I think we shouldn't answer this question before we've sawn SGI's
standalone shell. I remember some people on this list saying that they'd
like to have a standalone shell for their ARC machines.
> I end up using GNUism in Milo's Makefile is it an inconvient ? I'd like
> to ear your comments on this. My idea is to make the whole package ready
> to compile or cross-compile. So I guess using relying on GNU make for
> 'compilation' aka under Linux/MIPS is not a problem but for
> 'cross-compilation' ? I'll also try to make it a 'configure;make' package.
If you use autoconf you can live without gmake much easier. While beliving
that it's good not to rely on the GNU tools this is not really necessary.
After all they're available for free :-)
Hmm... if you want to configure the Milo-package (possibly supporting
multiple host environments in the future) we'll need a way for Milo to
identify which bootloader exactly is required. How about adding an entry
boot_method : ARC
boot_method : SGI
to /proc/cpuinfo so that Milo can autoconfigure completly for native
> Regarding kernel 1.3.91, I'm not able to launch a /bin/sh. This /bin/sh
> is ash - from sources tuned by Ralf available on ftp.fnet.fr - compiled in
> ELF against GNU libc 951212 - the lastest available Ralf's work. But
> the kernel_thread(do_shell......) in init/main.c:init() keeps getting launched
> and it says: 'child xxx died with code 000b'
> The do_execve() call in arch/mips/kernel/syscall.c:sys_execve() seems good
> I had also tried the ash binary from Ralf -> same thing... :(
> Any hints ?
Sorry, no. I've been running the latest kernels on Wayne's M700 with
success also. So it can't be a M700 specific problem. Could you send me
a more detailed description of your setup?