[OpenAFS-devel] New unofficial OpenAFS 1.3.73 rpms for Fedora Core 2

[Miles Davis]

On Mon, Oct 25, 2004 at 02:49:59PM -0400, Matthew Miller wrote:
> On Mon, Oct 25, 2004 at 11:45:30AM -0700, Miles Davis wrote:
> > > <http://www.mattdm.org/misc/openafs/>, updated for 1.3.73. This release
> > > appears to actually work -- cool, and thanks everyone for all your work.
> > Oh, I like these...seems to make more sense to me compared with the 
> > existing RPMs. Very nice...
> 
> Thanks -- glad you find them of use.

No, thank you. I especially like the more Linux-like (or at least 
RedHat/Fedora-like) FSSTND layout. Though it may be confusing to people 
used to the OpenAFS or Transarc paths, it makes more sense for current 
systems.

> > > I'm posting this to the devel list because it's still a work in progress --
> > > obviously it's still the unstable branch, and there's the whole PAG issue.
> > > Plus, I haven't had a chance to test this with SMP yet.
> > Haven't encountered any problems on SMP yet.
> 
> Cool. I actually did build the packages on the site with SMP support -- I
> should change the comment on the site.


Still no SMP problems yet, though I haven't yet tried it out with my 
guaranteed-to-crash 1.2.11 software yet (see posts from me on the list 
from a while back for details).

I have seen a rather annoying problem on a couple of hosts -- occasional 
'cache corruption' -- that is, after doing something like a make of a 
large bit of software, something weird happens to files. They show up in 
an 'ls -l' as something like

?---------   ? ?---- ?------   ?--- ?-- ?-  ?--- file

Doing an fs flushvol solves is, and I haven't pinned down exact 
circumstances to reproduce it. I doubt it's specific to your RPM; has 
anybody else on the list seen this?

-- 
// Miles Davis - miles@cs.stanford.edu - http://www.cs.stanford.edu/~miles
// Computer Science Department - Computer Facilities
// Stanford University