> >There is another problem with the ELF binutils 2.6. It is currently not
> >possible to compile a kernel with them. I'll take a look at this
> >buglet rsn. This bug is the only reason why we still cannot make
> >a.out what it actually already is - history ... To avoid for all
> Are you speaking about the same kind of problem I got by trowing away
> HJs binutils ? I was unable to compile my i486 kernel in ELF due to
> the -k (if I recall well) option added by him to objdump (aswell as another
> option I don't recall at the moment). The fix was simple : I had grabbed
> HJs objdump.c from binutils-2.5.2.l2?? and compiled it with a binutils 2.6
> "workbench", so I end up with a lobjdump (l for Linux :) , oh yes I did the
> same for encaps program (not present in binutils 2.6). So I just change
> the definition of the OBJDUMP variable in the i386 kernel Makefiles and tada
> it works : I had compiled 1.3.42 till 1.3.45 this these.
No, these special tools and patches aren't required to build a Linux/MIPS
kernel into ELF. Using the normal binutils 2.6 with my patches for Linux
flavoured a.out binaries on MIPS and some bugfixes for the assembler I can
compile and link the kernel into ELF format - no problems. Then I use
ld to convert the generated vmlinux into an a.out file for Milo, put it
on a boot disk and boot this kernel. An guarantee to crash your machine
as good as installing NT on an Oily :-) No idea what the cause of the
problem is - I only spent five minutes yet on that bug.
> >Last week I used my MIPS box for a (successfull :-) crypto-attack trying
> >to break a PGP key brute force. So you see it's slowly getting usable.
> Hey you had fullfilled you dream you can stop working on Linux/MIPS :)
Not yet - my own DES implementation that I did years ago will probably be
boosted by a factor of 3 or so by going to 64 bit. Then I'll be satisfied.
Oh, and anyway the overhead of an operating system is far too high. Why
wasting memory and CPU power for something like that (says it and continues
to punch cards ;-)
> cheers, Stoned.
> PS/ still making lines on my Olivetti console to find out where I fucked up
> my NASP code in the kernel :(
Ts, ts ... That's why I left a certain piece of code called beep() in the
kernel. Still works when almost everything else is dead.