Re: NetBSD 4.0 "opcode not supported on this processor:mips1"...

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

> > - "./configure --disable-atomic" on scratch build as pkgsrc does
> 
> So -disable-atomic disables the cpu check then?

No.

It may configure bind not to use atomic ops as its name.
asm code lines you got errors are written to provide
atomic ops for newer mips, but old mips1 cpus don't have
asm instructions for atomic ops and bind people don't bother
to implement alternatives or proper autodetect code for
such ancient ones, I guess.

NetBSD/mips still supports such older mips1 cpus as well as
it still supports rather slow cobalts, so disabling it may
be a reasonable choice.

> My complaint was not that it took a long time to build because the  
> system was slow - that I've willingly experienced before - my  
> complaint was that the build time with pkgsrc was long because it  
> built perl which I don't think is necessary to build bind96.  Or is  
> perl now a requirement for bind96?

Well, I wonder if you actually need bind96, not 9.5.1-P2 in base..

configure script in bind-9.6.1-P1 says:
>> # Perl is optional; it is used only by some of the system test scripts.
and maybe pkgsrc is designed "generic enough" for all possible users.

If you don't want options which requires perl but
need slight optimization which requires some quirks for -mips2
(or -march=vr5000 -mabi=32 as kernel) flags, please try
to dig out configure script etc.

> Besides - I think they look much better than your average server. :-)

Hehe, I have many m68k machines much slower than cobalts
(one of them takes more than 100 hours to build pkgsrc perl5)
while they don't look so much better ;-)
---
Izumi Tsutsui