Re: fsck seg fault failure on vmware -i386?

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

    Date:        Sun, 7 Feb 2010 15:51:03 -0000
    From:        yancm%sdf.lonestar.org@localhost
    Message-ID:  
<65a39cab0c8978c78a5576b1fca89f18.squirrel%webmail.freeshell.org@localhost>

  | OK, managed to bring the failed state HD to the repaired/consistent
  | system, dropped to single user to unmount /usr and created a new
  | core dump with debug fsck_ffs... same gdb output as before for me.

Your output looked more rational, I think we'll just ignore what I was
seeing, and write that off as system version incompatibility weirdness.

But with those values, I can't even imagine what is causing the core dump.

Are you in a position to actually run fsck under gdb, with a breakpoint on
ctime(), and then step through the localtime() call and see what happens?

If/when it returns NULL, which is what we're presuming, what test is it
that causes that to happen?  - I am guessing that the call sequence will be,
ctime -> localtime -> localsub -> gmtsub -> timesub ... if timesub()
returns NULL (or is about to), the values in *tmp and any other variables
in use about the place where it happens (and of course, what test is
causing the NULL return) would be useful to know.   This is when /usr
isn't there of course.

  | Question: Is there any way to unmount /usr in a networked multi-user
  | state?

Only by killing everything that was started that's using it - that's
probably more trouble than it is worth.   But if you just need to get the
network working so you can connect while in single user more (without /usr)
that can be done, with a little fiddling, depending upon just what you
need to be running.  You can manually ifconfig the interface, and start
whatever processes are needed (perhaps copying what is needed from /usr to
somewhere on root first).

kre