Re: fsck seg fault failure on vmware -i386?

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

    Date:        Tue, 2 Feb 2010 23:31:16 -0000
    From:        yancm%sdf.lonestar.org@localhost
    Message-ID:  
<6dec5851c8865e4697948584e22b547d.squirrel%webmail.freeshell.org@localhost>

Sorry, busy week, no time for reading NetBSD e-mail ...

  | Is this what you are looking for?:

Yes, exactly (I didn't really neeed &t, but that's harmless).   But now
I'm more confused, as I can't thing of any reason, in any timezone, that
there would be a problem converting 1264115119 to a struct tm, whether or
not the zone file was available.

That's Thu Jan 21 23:05:19 GMT 2010 (5 hours earlier, in -0500 zone)
which, as best I can see is a totally unremarkable time, nothing at all
like I was expecting to see.

You're obviously using a current system (the existence of __ctime50()
shows that), I wonder if perhaps you somehow have a fsck that was compiled
with a 32 bit time_t?

I see Christos has been helping look into this (when I search ahead in my mail
for the rest of the thread), I'm going to hunt in the fsck binary and
core file you made for him, and see if I can see some plausible cause
for why this fails.

The recompile (that worked) with localtime.c included in the fsck
sources would perhaps be explained if somehow the number of bits in
a time_t was being mixed up - recompiling localtime then would make a
32 bit version (matching fsck) which would then just all work.

If I see something in those files, I will certainly let you know, but
whatever happens here, this problem has certainly led to a bug fix in
NetBSD's fsck, and one in the Olson tz code (new version will be released
with a fix - different to the one NetBSD adopted in its libc) as well.

kre