[openfabrics-ewg] SLERT pointer for iser testing

[Michael S. Tsirkin]

> Quoting Alex Tsariounov <alext at novell.com>:
> Subject: Re: SLERT pointer for iser testing
> 
> Hi,
> 
> On Wed, 2006-12-27 at 09:11 +0200, Erez Zilber wrote:
> > > Can you use the patch below to solve the problem with iSER? It
> [...]
> > >  case ${K_VER} in
> > > -    2.6.16.21-0.8-smp)
> > > +    2.6.16.21*)
> [...]
> > Michael, Vlad,
> > 
> > The patch above fixes a bug in OFED 1.1 installation scripts. iSER could
> > be installed only on a specific kernel version of SLES 10. The patch
> > allows it to be installed on any SLES 10 version (i.e SLERT and future
> > service packs).
> > 
> > This patch was accepted by Novell. Can you add it to OFED 1.1?
> 
> Well, as soon as I said that we would never change the suse minor number
> for a maintenance kernel release.....   The current SLES/D maintenance
> kernel found in the Novell updates site has the revision: 2.6.16.27-0.6
> 
> So, the test above will fail on that kernel and iser will yet again not
> be built.
> 
> It's probably best to use this test for SLES 10:
> 
>         case ${K_VER} in
>          -    2.6.16.21-0.8-smp)
>          +    2.6.16*)

But, it'll also catch all kernel.org kernels and any other 2.6.16 based distro,
so that's no good.
Something like 2.6.16.*-*.*) would be a bit safer.

I'm not strong on shell glob pattern but I think we even want something that
matches regexp 2\.6\.16\.?[0-9]+-[0-9]+\.[0-9]+-smp). It's a pity SLES does
not label its kernels more clearly. Something like
2.6.16.21-0.8-smp-sles10 would be much better (like fedora does).

-- 
MST