[Gluster-users] Unusual bug in glusterfsd

[Stuart Longland]

On Mon, Aug 25, 2008 at 11:13:13PM +0530, Anand Avati wrote:
> >
> >         The key error message seems to be this one (note the high negative
> > maximum number):
> > > 2008-08-24 18:11:45 E [protocol.c:342:gf_block_unserialize_transport]
> > server: frame size (168) > max (-2147458000)
> 
> 
> This seems to be a compiler or build related issue. What is your nickname on
> IRC? we could try to debug it online.. what architecture is this machine? I
> remember another user having this issue on a mips machine but we could not
> get around to actually debug it.
> 
> avati

Hi Anand,
	I just had a look when you mentioned this -- seems on my laptop
(working glusterfsd) I was using gcc 3.4.6 with vanilla specs.  On the
web server (broken glusterfsd) it was using hardened specs (with stack
protection and position independent executable options).  This is on
x86.

	I've just recompiled it using vanilla specs on the web server,
having noticed another more subtle data truncation bug with the client.  Having
tried connecting with another client on the network, I'm successfully able to
mount the volumes from the webserver without issues.

	I'll do some further investigations, but it seems the hardened toolchain
in Gentoo would be a prime culprit here.

Regards,
-- 
Stuart Longland (aka Redhatter)              .'''.
Gentoo Linux/MIPS Cobalt and Docs Developer  '.'` :
. . . . . . . . . . . . . . . . . . . . . .   .'.'
http://dev.gentoo.org/~redhatter             :.'

I haven't lost my mind...
  ...it's backed up on a tape somewhere.