On Thu, Jun 22, 2000 at 07:31:18PM +0100, Alan Cox wrote:
> > One could make a case for the statement that every heuristic
> > in the kernel is wrong. And here we have a heuristic for guessing
> > whether the partition table was written using 512-byte or 2048-byte
>
> Why is it a heuristic. The rule should be simple enough
>
> if (partiton > sizeof disk && scaled && partition fits on disk if not scaled)
> try unscaled
>
> I'd say its a fairly good rule.
I agree.
> I need to figure why its not doing the right thing.
Maybe I told you the answer already.
: The heuristic itself is reasonable enough. But since 2.2.15
: did not yet have this heuristic, it follows that in all cases where
: the heuristic applies and 2.2.15 was OK, 2.2.16 will break.
In other words, nothing is wrong with the heuristic,
but when people partition their disk with some fdisk,
nobody knows which version they use, nobody knows whether
they use the -b 2048 flag, they just get something,
correct or not. And once they have something that works
under 2.2.15 while the heuristic applies then it certainly
doesnt work under 2.2.16 (because the starting point of all
partitions is changed by a factor 4; in particular, it does not
stay the same).
Now you ask: but wasn't then the old partition table broken?
Quite possibly, yes. And if he did mke2fs on it, does not the
ext2 filesystem extend beyond the end of the disk?
Maybe. He should certainly check.
> And if I cant figure that in time for .17 I agree entirely its
> better out than in
Yes perhaps. Note that I only expressed a weak opinion:
: So, I would not be unhappy if this code were removed again.
The code is a kludge, and it does some good, and it does some bad.
There is unrest in all cases, whatever we do.
Andries
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
