CC-ing to the ML.
Hackie:
> Ok, looks like a reasonable way.
>
> You solved it by allocating an array. That's something I wanted to avoid.
>
> What I don't like is the new bsize, which is set to the maximum. The
> bigger the value, the lower the precision. If there is any squashfs with
> its default 128k blocks in your branches, all 4k block numbers of other
> branches are divided by 32 and rounded. That's not really a problem, but
> also not nice when there are other ways. And especially not better than
> using 512.
I don't know why you like 512 so much.
If we choose fixing the aufs bsize, I'd say the bsize on the first
branch is best.
Now you wrote the multiplication variant and I wrote the division
variant, which means we can choose any value as the aufs bsize!
So, I'd like to implement such like this.
- the aufs bsize is the one on first branch.
- when adding in the sum mode,
+ if the branch bsize is greater than the aufs bsize, then we
calculate the number of blocks by division.
+ if the branch bsize is smaller than the aufs bsize, then we
calculate the number of blocks by multiplication.
How do you think?
J. R. Okajima
----------------------------------------------------------------------
> Ok, looks like a reasonable way.
>
> You solved it by allocating an array. That's something I wanted to avoid.
>
> What I don't like is the new bsize, which is set to the maximum. The
> bigger the value, the lower the precision. If there is any squashfs with
> its default 128k blocks in your branches, all 4k block numbers of other
> branches are divided by 32 and rounded. That's not really a problem, but
> also not nice when there are other ways. And especially not better than
> using 512.
>
> I just found a new variant which works with only one loop and uses the
> lowest bsize.
> Didn't have a chance to test it yet, but at least it compiles, and a
> test calculation in ooo worked. Apply the patch after all 4 of us.
>
>
> btw. You were right with the problem of patch 1. Was a beginner mistake.
> One should never use -f when there is no real reason for it...
>
> Daniel
------------------------------------------------------------------------------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here
http://p.sf.net/sfu/sfd2d-msazure
