> On another note, I have been trying to setup my build environment
> on a Sparc 5 running Solaris 2.4. I built my own binaries after applying
> the patches off ftp.fnet.fr (gcc 220.127.116.11 & binutil 18.104.22.168).
The patches are intended to be applied to the vanilla FSF versions of
GCC 2.7.2 (22.214.171.124 should work except harmless rejects, haven't checked).
But in no case use HJ Lu's binutils 126.96.36.199. It will generate unuseable
kernels with high probability.
> When I try and build a mips kernel (ralfs 2.1.14 version) the first
> problem I have is trying to build the new mkdep version. For some reason
> this uses a shared memory segment and is not well liked by solaris.
> If you do manage to get it to compile, (if I remember correctly, with a
> bit of help from the bsd compatibility libs) then it crashes the second
> time it is run :(
Yes, unfortunately you're right.
> Now if we force the issue a bit and touch .depend in the root of the source
> tree we can actually start compiling the kernel. Most of this works out
> okay except when it gets to the low level stuff and it gives me the
> following rude message:
> make: Entering directory
> mipsel-linux-gcc -D__KERNEL__ -I/opt/home/stuarta/mips/linux-2.1.14/include
> -I/opt/home/stuarta/mips/include -Wall -Wstrict-prototypes -O2
> -fomit-frame-pointer -G 0 -mno-abicalls -fno-pic -mcpu=r3000 -mips1 -c
> r3000.S -o r3000.o
> r3000.S: Assembler messages:
> r3000.S:141: Warning: Macro used $at after ".set noat"
> r3000.S:230: Warning: Macro used $at after ".set noat"
> r3000.S:277: Error: illegal operands `sw'
> r3000.S:337: Warning: Macro used $at after ".set noat"
> r3000.S:365: Warning: Macro used $at after ".set noat"
> r3000.S:526: Warning: Macro used $at after ".set noat"
> make: *** [r3000.o] Error 1
> make: Leaving directory
> make: *** [linuxsubdirs] Error 2
> Suspecting this was something to do with the way I installed the binutils,
> I grab the binutil binaries for solaris 2.5 and gave them a go but this
> didn't improve matters....
Well, the problem is the source. Paul's DECstation source tree is probably
completly out of sync with mine. We'll fix that rsn. Paul, maybe you
have a tree more suitable for DECstations for ftp at Softway?