A few scattered old bug reports

[Russ Allbery]

Anonymous Coward <newsb'stard at Free-N0RP.netSCUM.dk> writes:

> bin/cnfsstat -a does not give a percentage-like indication of buffer
> usage, but rounds things to a whole cycle:

>  Buffer FIVE, size:  41.0 GBytes, position:  29.0 GBytes  0.00 cycles
>   Newest: 2000-07-31 14:32:07,    0 days,  0:00:34 ago
>   Oldest: 2000-07-19 17:01:07,   11 days, 21:31:34 ago

Are you still having this problem?  This seems to work fine for me.

> I'm not sure if I'd rather see the %, or rather have the number of
> cycles reflect this (0,74 cycles in this case) -- probably the latter or
> else unneeded precision is being wasted.  Also, um, why the precision of
> 29,0 GB and 41,0 GB rather than, say, 41,4 GB, or whatever?

This works fine for me too....

> Second, in nnrpd, in SetupDaemon, the value of ClientHost which
> appears in the overview log message is not set until later in the
> StartConnection call, after SetupDaemon returns...

Fixed a while back.

> And finally, I noticed that somehow, ExitWithStats seems to be called
> twice for a small number of connections -- right at the point of the
> last nnrpd wedgelock/ovdb failure, but also once before then without the
> catastrophic following.  Still, I could not see in the k0deZ how one
> would pass through ExitWithStats more than once.

This one I have no clue about.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>