[Motmot] Segmentation faults

[Andrew Straw]

I have been debugging this here, too. The bottom line is that I still
don't know what is going on, but that it does seem to have a 100%
correlation with the Prosilica GigE drivers being in use. Therefore, I
have just made a new libcamiface package without Prosilica support and
an optional libcamiface-prosilica that adds in the Prosilica support if
needed. I'm testing this package out now, and if it goes 24 hours
without seg faulting, I'll upload these packages to my repository and
deal with the Prosilica-induced issue without worrying that I am
inhibiting experiments around the continent...

-Andrew

John Bender wrote:
> I made a 32-bit Hardy install on the same machine to see if that was
> the issue. Now I'm getting the same backtrace that Jon is reporting. I
> will await the outcome of his testing!
>
> JB
>
> On Tue, Mar 31, 2009 at 2:06 PM, John Schneider <jzschneider at gmail.com> wrote:
>   
>> Thanks, didn't notice the old packages!
>> Unfortunately these are brand new machines that I never bothered to
>> register with computing services, so they have only had one kernel
>> (2.6.24-23), and they don't have internet.
>> Update:
>> Older kernel/newer fview - Segfaulted overnight
>> New kernel/old motmot packages - still running strong
>> I will try to update the motmot package one step at a time (unless you
>> know of a reason that will not work) to try to narrow it down.
>>
>> Jon
>>
>> On Mon, Mar 30, 2009 at 6:10 PM, Andrew Straw <astraw at caltech.edu> wrote:
>>     
>>> John Schneider wrote:
>>>       
>>>> Hey Andrew,
>>>>
>>>> I'll downgrade some computers tonight with that earlier kernel and try
>>>> some long term acquisition. I'm also going about it another way: I'm
>>>> currently running one computer with the current kernel but old fview
>>>> packages (generated by apt-on-cd) that I had laying about, too early
>>>> to tell if it fixes anything, but could narrow things down as well.
>>>> Will update tomorrow,
>>>>         
>>> Hi Jon,
>>>
>>> That sounds good, but you don't have to downgrade from CD -- I purposely
>>> don't delete the old packages from my server. So you should be able to
>>> downgrade by manually installing the correct .debs directly. For
>>> example, if you go to http://debs.astraw.com/hardy/?C=M;O=D you'll get
>>> the .debs listed in reverse chronological order. From that, I can see
>>> that installing
>>>
>>> http://debs.astraw.com/hardy/libcamiface_0.4.dev716-0ads1_i386.deb
>>> http://debs.astraw.com/hardy/camiface_0.4.dev716-0ads1_i386.deb
>>> http://debs.astraw.com/hardy/python-motmot-camiface_0.4.1~dev-r637-0ads1_i386.deb
>>> http://debs.astraw.com/hardy/python-motmot-fview_0.5.2~dev-r715-0ads1_all.deb
>>>
>>> should restore the .debs to the situation throughout most of February.
>>> (The most recent package of those is from 6 Feb.)
>>>
>>> Finally, to run the older kernel, you don't have to install any new/old
>>> packages: it's just a matter of rebooting.
>>>
>>> -Andrew
>>> --
>>> Andrew D. Straw, Ph.D.
>>> California Institute of Technology
>>> http://www.its.caltech.edu/~astraw/
>>> _______________________________________________
>>> Motmot mailing list
>>> Motmot at code.astraw.com
>>> http://code.astraw.com/cgi-bin/mailman/listinfo/motmot
>>>
>>>       
>> _______________________________________________
>> Motmot mailing list
>> Motmot at code.astraw.com
>> http://code.astraw.com/cgi-bin/mailman/listinfo/motmot
>>
>>     
> _______________________________________________
> Motmot mailing list
> Motmot at code.astraw.com
> http://code.astraw.com/cgi-bin/mailman/listinfo/motmot
>
>   


-- 
Andrew D. Straw, Ph.D.
California Institute of Technology
http://www.its.caltech.edu/~astraw/