Jon said:
Thanks Paul, that's a neat tool to keep on hand. Unfortunately
(fortunately?) it didn't find any bugs, but it did put up an interesting
piece of information just as it was starting.
Apologies for the poor quality of this, it only flashed on the screen
for a moment:
http://i.imgur.com/Cgv7uMA.jpg
Maybe I have found a clue....
Jon
Something about the partition table entries themselves ?
I wish I could understand how that stuff all works,
because I've run into a situation before where something
broke like that, and I couldn't tell using PTEDIT32 or
other similar tools, exactly what was broken.
When your WinXP is running, try recording a screenshot of
PTEDIT32, when it displays that particular disk.
In this sample picture, is a Win2K disk. It's legacy aligned,
and all numbers on the right hand side should be divisible
by 63. Newer Linux and Vista/7/8 will use megabyte alignment.
Notice how my fake head count and sectors are 255 and 63
respectively, on a 500GB disk. Once you get above a certain
disk size, CHS geometry bitfields aren't wide enough to
properly represent the size, so fake "max" values are
substituted. Modern stuff knows it should use LBA when
seeing bogus CHS numbers. The stuff on the right is LBA related.
http://i41.tinypic.com/16kxrb9.gif
See what you disk looks like, with this (free) tool. For
the latest OSes, you would select "Run as administrator" to
avoid getting an error 5. It might just work by clicking
it in WinXP (after you unzip it of course).
ftp://ftp.symantec.com/public/english_us_canada/tools/pq/utilities/PTEDIT32.zip
In addition to that definition, it should be noted that
the virtual file system size can be smaller than the
physical size declared in that table. One user tried to
shrink a partition using Disk Management in a modern
OS, the virtual portion shrank but the physical definition
did not. Half the disk was "wasted", because of that bug.
Linux is king at doing that sort of thing, as for
some reason (on purpose), users are allowed to have
unaligned usage (virtual smaller than physical).
When a fraction of a cluster is involved, it's normal
for the virtual file system definition (defined by the
file system header), to be smaller than the physical space.
A fraction of a cluster can't be used to hold data. I'm
referring to more gross errors, where the virtual file
system definition is only half the size of the physical
space as defined in the partition table.
I think your error is referring to the stuff PTEDIT32
displays. That's a guess.
If the virtual definition was bigger than the physical,
I expect the result would be more immediate. And permanent.
Paul