[Top] [All Lists]

Re: file corruption during emacs build on XFS logical volume

To: Sean Neakums <sneakums@xxxxxxxx>
Subject: Re: file corruption during emacs build on XFS logical volume
From: Steve Lord <lord@xxxxxxx>
Date: 03 Jan 2002 08:49:37 -0600
Cc: Linux XFS <linux-xfs@xxxxxxxxxxx>
In-reply-to: <6ubsgbq013.fsf@zork.zork.net>
References: <6u4rm4r53e.fsf@zork.zork.net> <1009995505.14223.9.camel@jen.americas.sgi.com> <6uy9jgpn2x.fsf@zork.zork.net> <6uu1u4pf48.fsf@zork.zork.net> <1010013019.1281.6.camel@jen.americas.sgi.com> <6upu4spccv.fsf@zork.zork.net> <1010013997.1315.11.camel@jen.americas.sgi.com> <6ulmfgpau5.fsf@zork.zork.net> <6uheq4p9ny.fsf@zork.zork.net> <3C33AF81.AC235AD0@idcomm.com> <6ubsgbq013.fsf@zork.zork.net>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Thu, 2002-01-03 at 03:03, Sean Neakums wrote:
> begin  D. Stimits quotation:
> > Sean Neakums wrote:
> >> I've been unpacking the source fresh for each try.
> >
> > FYI, file compression is extremely sensitive to marginal ram. If
> > unpacking has anything to do with it at all, and not the source of
> > the unpack or destination, you should double-check ram with
> > memtest86 or something similar. Let it run for many hours.
> I ran two passes of memtest-86 a few days ago, after I had that XFS
> filesystem shutdown.
> I'm not getting any errors at the unpack stage.  I mentioned that I
> was unpacking fresh for each try in order to establish that the tree
> itself was not b0rked.  Note from my previous messages that the builds
> have only failed when I built on XFS, and not when I built on ext2,
> and that builds on XFS have succeeded when I use 2.4.14-pre7.  This
> just doesn't feel random to me.

Since I am building on a redhat based box, I am doing rpm builds not
debian package builds, and the rpm version which appears friendly to
this box is emacs-20 not emacs-21. Having said that, emacs is building
OK for me here. It is difficult to tell if the problem you are seeing
is in the true emacs build, or in the package generation. Can you tell,
if it is the package generation then the rpm build is not a valid test
case here.

I do not think that the presence of LVM in the kernel will have any
effect on XFS, it is good to have a simpler test case though, there
are some subtle interactions between XFS and LVM.

I am off to find a combo where the rpm build will work for emacs-21.



Steve Lord                                      voice: +1-651-683-3511
Principal Engineer, Filesystem Software         email: lord@xxxxxxx

<Prev in Thread] Current Thread [Next in Thread>