loads of XFS panics

[Tino Schwarze]

Hi Love,

On Wed, May 10, 2000 at 01:22:10AM +0200, Love wrote:
> > > they do not really that mean something is wrong for you, but
> > > rather that code doesn't do what other parts of it assume it will.
> >
> > hmmm. The bad thing is that this is a) very confusing b) renders the
> > system almost unusable with such a flood of messages everywhere. 
> >
> > Would an option "-debug=errors,warnings" passed to arlad fix these
> > issues?  My problem is that I'm maintaining an ARLA RPM and I cannot
> > live with thousands of non-critical "error" messages popping up
> > everywhere.
> 
> Eah, I know. Here is the deal: the error it warns about is VERY
> serious, it will crash in readpage/writepage if the correction-code
> didn't fix the error. So the error message should really say: "Ahhhhh
> the world is crashing down on our heads, but don't worry. I've fixed
> it.".
So this is actually more a reminder to developer to get rid of that
correction and fix the error itself than an error message?

> The message is there so we should remember to fix it. If you can give a
> simple way os reproducing the error, we'll happly fix the real error
> instead of the symtoms.
I'll have a look. Yesterday it looked like doing the following is
sufficient:
a) stop arla
b) remove arla cache
c) start arla
d) do ls in /afs/tu-chemnitz.de/home/urz/t/tisc
e) either do
  e.1) ls -l miktex
  e.2) ls -l /afs/tucz/ftp

At a first glance it looks like the issue pops up if there are links
involved. I'll see if I can check that today midday.

> > > So don't worry about and tell us to fix it. :-)
> > BTW: I could not find a list of supported debug options, a more
> > comprehenshive help screen would be helpful (I could also have a look
> > and add it myself)
> 
> : datan ; fs arladeb help help
> usage: arladebug [all, almost-all, errors, warnings, disconn, fbuf, messages, kernel, cleaner, callbacks, cache-manager, volume-cache, file-cache, initialization, connection, miscellaneous, venuslog, default, none
> ]
I thought more of an addition to "arla -help" and ARLA documentation.

> I have fixed the sillyness related to getting help in current arla.
:-)

> But these doesn't control the errormessage that you are after since it does
> a printk without looking at the debugging variable.
Bad luck. :-(

I'll play around a bit today and tell you if I found anything
interesting (e.g. a way to reproduce this)

Bye, Tino.

-- 
             * LINUX - Where do you want to be tomorrow? *