On Fri, 23 Aug 2002, Shin'ichiro TAYA wrote:

> I think MAXDSIZ and MAXSSIZ in sys/arch/mac68k/include/vmparam.h is
> too small to build mozilla.
> Increse them and "ulimit -s unlimited -d unlimited" will help you.

I don't think that's it. "regchrome" consistently dumps core in the
same place, using only about 11mb. This is after mozilla is completely
built and installed. An interesting data point -- by the time
"regchrome" drops out, "top" shows that every other process on the
machine is paged out (but swap is not touched).

> > Is it really necessary to build mozilla with "-01" on most ports, and
> > with "-O0" on m68k? I think not. Usually, if there's a problem with
> > optimization levels on one file, we'll reduce it for that one file,
> > not for the whole program, and mozilla is slow enough, that every
> > little bit counts.
>
> You are right.
> But mac68k is *too* slow to investigate which file(s) should be
> compiled without optimize.
> I tried, but I give up.

At the time you reported testing, the default for m68k was -O0, so I
guess that doesn't help. I've since made it -O2 for all ports, and
marked NetBSD-*-m68k broken. What I intend to do, after I've finished
upgrading to 1.6, is to try building a cvs checkout of mozilla with a
cross-compiler, to really explore this.

Frederick