Subject: (a) new si driver; (b) make coredumping
To: None <port-sun3@NetBSD.ORG>
From: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>
List: port-sun3
Date: 10/31/1995 12:14:13
I've got a Sun-3/150 that I'm running NetBSD/sun3 on.

First point: I'm using a kernel with gwr's new si driver.  It has
worked fine for me, nary a peep of trouble, and is much faster than the
old one (I can actually fsck the /usr/src partition, which is about 2G,
in a tolerably short time now).  Thank you, gwr and whoever else is
responsible for getting that code working!

Second point: I'm currently doing a rebuild of the world.  Actually, I
did a rebuild of the world before, too; I'm doing _another_ rebuild,
for three reasons: (1) to make sure I've got a self-consistent system,
(2) as a test of the new si driver, and (3) to bring it right up to
-current; the previous build was from a tree a few weeks old.  The
reason I mention this is that the build of libc has crashed twice.
This would not be remarkable, except that the previous build, with a
kernel from back in June or July sometime, crashed in exactly the same
place, at least for the first failure.

	(cd /usr/src/include && make install)
	(cd /usr/src/lib && make depend && make && make install)
[...normal make depend messages...]
	===> libc
[...normal libc make output, until...]
	cpp -DPROF -DNLS -DYP -DLIBC_SCCS -DSYSLIBC_SCCS -I/usr/src/lib/libc/include -D__DBINTERFACE_PRIVATE -DPOSIX_MISTAKE -DFLOATING_POINT -I/usr/src/lib/libc/arch/m68k /usr/src/lib/libc/arch/m68k/string/memset.S |  as -o memset.po
	Segmentation fault - core dumped
	*** Error code 139
	*** Error code 1

I investigated, and it was make, of all things, that coredumped.  I
restarted by blowing away memset.po just in case and then doing
( cd /usr/src/lib && make && make install ) by hand; it picked up as
one would expect and ran along fine until

	nroff -mandoc /usr/src/lib/libc/sys/fsync.2 > fsync.cat2
	Segmentation fault - core dumped
	*** Error code 139

Again, it was make that dropped core.  I restarted by hand again,
blowing away fsync.cat2 first, and it picked up as one would expect.
It finished off libc, libcompat, libcrypt, libcurses, libedit, libkvm,
libl, and is as I write this is slogging through libm (it's at
"building standard m library").  Except for these coredumps from make,
I've seen no trouble at all.

Anyone have any idea what's with make?  I intend to do yet one more
build of the world once this one is done, and if the make installed by
this build doesn't die, I won't worry about it.

					der Mouse