Re: fsck seg fault failure on vmware -i386?

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Fri, 12 Feb 2010 22:38:43 +0000 (UTC)
    From:        Eduardo Horvath <eeh%NetBSD.org@localhost>
    Message-ID:  <Pine.NEB.4.64.1002122226270.26167%mail.netbsd.org@localhost>

  | All the makefile magic is /usr/share/mk.

Yes, it turns out I was looking at a much older version of that
(much older...) so ignore most of what I said about what works or
doesn't in my last message.   The how to check stuff was still OK
though I think.

  | MKDEBUGLIB      Build *_g.a debugging libraries, which are compiled
  |                 with -DDEBUG.

Except that we want (need anyway) just one debug library, and we don't
want -DDEBUG, which wile it would probably be harmless, might potentially
alter the code that is compiled so the debugging that needs to be done
might be a waste of time - we want to find out what happened, and for that
we need the code to be as close to the same as it was before as possible,
just with the extra symtab noise.

  | DBG is supposed to be used for the compiler optimization flags.

A strange name for that use...


  | I'd recommend:
  | 
  | make DBG="-g -O2" MKDEBUGLIB=yes LDSTATIC=-static

Except that this build is using build.sh - which would mean, for that

        build.sh -V DBG='-g -O2' -V MKDEBUGLIB=yes -V LDSTATIC=-static \
                distribution

except that we don't want -DDEBUG to exist, so no MKDEBUGLIB, we don't need
-static, so omit LDSTATIC, and we only need libc and fsck_ffs compiled with
-g (and in libc, really only localtime.c)   COmpiling the rest with -g
wouldn't hurt, but it would make a lot bigger result (lots more symbol tables.)

kre