Subject: Re: fsck
To: None <mycroft@gnu.ai.mit.edu>
From: David Gluss <david@pure.com>
List: tech-kern
Date: 10/26/1994 21:02:50
I give up.  My attempts to port fsck to other machines (so I could
Purify it) failed because the other machines have a different byte
ordering, so reading in stuff gave unintelligible numbers.

The fsck failure is possibly because the file system in question is "old",
and the code to handle that is not robust.  The failure is symptomatic
of a UMR, because it doesn't fail the same way twice.

I got some confirmation of that by converting all "malloc" calls to
"Malloc" calls and initializing the memory it returned, and verifying
the continuing nonrepeatability of things.

Still, this seems hard to explain.  Doesn't the program model presume
that memory acquired with sbrk return zero initially?  So when I quit
gdb and restart it, I should see repeatable behavior?

I can ship the failing file system to anyone who wants to try
their hand at it.  It's about 15 Mbytes uncompressed, 6.1M compressed.

--David Gluss    david@pure.com     408-524-3028