Re: [PATCH] corrupted free inodes list

[Pavel Tsekov <ptsekov gmx net>]

On Fri, 22 Dec 2006, Thomas Dickey wrote:

> On Fri, 22 Dec 2006, Leonard den Ottolander wrote:
>
> > Hi Mikulas,
> >
> > On Wed, 2006-12-20 at 02:15 +0100, Mikulas Patocka wrote:
> > > --- maybe you could change double to "long long" but I'm not sure if it
> > > exists on all machines --- a configure test would probably be needed.
> >
> > Yes. Using floats for discrete counters is not such a good idea.
>
> It's likely to be slower, but if you don't exceed the precision of the
> mantissa, you won't lose information.
>
> The general reader, taking into account the comment about "discrete" is
> likely to assume that you meant information would be lost.  Since that's
> not the case (for the assumed 48 bits), it's worth clarifying the comment.

I am working to get this fixed. I plan to use code from
coreutils's df to calculate the percent. I also resurrected
the files fsusage.[ch] where originally the get_fs_usage()
function resided. fsusage.[ch] come from gnulib/coreutils/filutils
and its maintainers had switched to uintmax_t to collect
the disk usage statistic for a long time now. I'll let you
know when I am done. If someone is willing to help take a
look at show_dev() in coreutils's source code. I am also
wondering whether we won't to import human.[ch] from gnulib.
IMO, it makes sense, but still this code does bit more than
what we need but maybe in the long term its worth to rely
on a known good and supported code instead of a homebrew
solution.

Please, let me know what do you think.

Thanks!