Paul Bijnens wrote:
> >Upgraded one box to 2.4.4p3. amreport still shows the
> >bogus labels when run against a log generated under
> >2.4.4p2. Tonight I'll see what it does when it creates
> >the log itself. :-)
> 
> Do you mean you upgraded a client?  Just the amanda server is fine.
> The 2.4.4p3 server is compatible with older clients.

Sorry, yes, I meant I upgraded the server. ("one box" here
being one of two amanda servers.)

> >(Does amreport pull its stats from anywhere but the logfile?)
> 
> It also needs the amanda.conf and the disklist file.

Yup, but there are no stats there.

> I should delve into the source code to find out what exact information
> it gets from there (e.g. columnspec, mailadress come to mind).

My columnspec is unspecified (ie, default). (That was one
of the things I initially suspected -- a missing column width
causing a snprintf() to get confused, for example. Doesn't
appear to be the case.)

> If you like, you may send me such a log file, so I could have a look
> at the strange things it calculates.

We'll see what happens in tonight's run with 2.4.4p3 :-)
I did manage to whittle down the logs (trimming all but
one DLE from them) to a state where I could trigger the
bogus behaviour by changing only limited info. But I
didn't completely exhaust all the options there before
deciding that the upgrade to 2.4.4p3 was perhaps a better
idea. :-) It *looked* like it was getting the "orig-kb"
value from thin air, as changing that had no effect on
the label (but changing the "kb" did.) That's why I
asked about amreport pulling stats from other places.

But I'll explore that idea further tomorrow, if 2.4.4p3
still misbehaves.

The configs I'm using have grown up from (I think) 2.4.1
or thereabouts. They've been around (and been hacked up
occasionally) for several years. So there's possibly
something missing that 2.4.4p[23] expects to see...