Subject: Re: Another vote to pull up /src/sys/arch/m68k/m68k/db_trace.c 1.24
To: matthew green <firstname.lastname@example.org>
From: Rob Healey <email@example.com>
Date: 04/05/1999 18:27:21
> there are *always* a lot of bootstrapping problems building NetBSD
> sources from an earlier release (be it 1.3B -> 1.3K, or 1.3 to 1.4,
> or whatever). surely you need to install new versions of several
> other tools (flex and yacc come to mind, due to the WARNSification)?
1.4 Kernel panic nailed me before I got that far. Looks like a
general m68k issue with the pool code that UVM uses when paging
starts in earnest; possible m68k code generation problem striking
flex and yacc are used to build executable code, I can understand
their need for a bootstrap requirement. I question the REQUIREMENT
for a bunch of doc files during initial bootstrapping, especially
when people with older versions of info in their /usr/local might
get bitten like I did!
> I've been able to build new versions of NetBSD/Amiga from the old
> since 0.9 days or so and I found the requirement for GNU info just to
> bootstrap irritating...
> well, you can always disable .info processing with `make MKINFO=no'.
> The suspected pool panic has stopped the process for now but I'd still
> would like to know why the build process for BSD now has an extra GNU
> dependancy that it doesn't REALLY need... B^(. Can we add a
> MKINFO=no or something to the basic bootstrap to forego the need for
> info during initial bootstrapping?
> oh, gee.. you meantioned exactly what exists but didn't try it?
Bzzzzzzzzzzzt, but thank you for playing! I DID try it since I
saw it in other Makefiles but the various bootstrap make's still
bombed for want of the info system. For the bootstraping of the
compiler, includes, etc. it seemed to have no effect.
See my other message for rant on requiring extraineous stuff for