Re: [Xenomai] Debian squeeze packages preview for the amd64 architectures

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

On 01/08/2013 09:09 PM, John Morris wrote:

> On 01/08/2013 01:54 PM, Gilles Chanteperdrix wrote:
>> On 01/08/2013 10:12 AM, Leopold Palomo-Avellaneda wrote:
>>> Please Gilles, the linux image is produced using an standard stock version 
>>> (just patched with xenomai)?
>>
>>
>> Yes, this is vanilla linux 3.5.7 with the xenomai patch applied, and a
>> kernel configuration derived from debian configuration for linux 3.2,
>> enabling every new option (as a module if possible) in 3.5 compared to 3.2.
> 
> Hi Gilles,
> 
> I see that 3.5.7 is the most recent kernel in the 3.5 series.  Xenomai
> 2.6.2 (and also master, and adeos ipipe) ships patches labeled for
> 3.5.3, but you have presumably applied the 3.5.3 patches to the 3.5.7
> kernel.
> 
> Is it acceptable practice to apply the Xenomai patches to later
> sublevels in the same 3.x series like this, or is this best left to
> experts like you?  When doing this, have you generally found the
> resulting kernel to be reasonably stable?


I have upgraded from 3.5.3 to 3.5.7 because of an NFS bug which caused a
kernel oops on my workstation (I usually do not test xenomai on my
workstation, except that it is the easiest way to test debian packages).
The 3.5.7 kernel seems to be reasonably stable, though I still have an
unexplained bug on ARM omap3 (when compiled with CONFIG_THUMB2_KERNEL
and CONFIG_SLAB, disabling one of these options makes the bug go away),
but could not reproduce it on x86.

As for what is acceptable or not, what matters is your level of
exigence. What I do to consider that a kernel is stable is run the
xeno-regression-test script, which shakes the computer quite a bit, but
you may consider it is too much or not enough testing.

-- 
                                                                Gilles.