[Motmot] Segmentation faults

[Andrew Straw]

John Bender wrote:
> Just for kicks, I thought I'd try this on my machine, too (I haven't
> been doing long runs yet). Instead of segmentation faults, I get
> "illegal instructions" after 1-30 minutes.
> 
> Program received signal SIGILL, Illegal instruction.
> [Switching to Thread 0x42d5c950 (LWP 18392)]
> 0x0000000042d5ac38 in ?? ()
> (gdb) bt
> #0  0x0000000042d5ac38 in ?? ()
> #1  0x00007fa17641ba01 in raw1394_loop_iterate () from /usr/lib/libraw1394.so.8

OK, that looks like some nasty thing in the linux 1394 driver. There are 
known bugs lurking in there... Which is why the linux kernel folks have 
re-worked the firewire infrastructure, but I don't think this 1394 
driver will be making it to even the upcoming Ubuntu release (9.04). 
That being said, it's possible this bug has been fixed.

> Also, remember I said I got segfaults at camera startup occasionally
> (right after clicking "OK" to initialize camera)? Here's one:
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7fd2629286e0 (LWP 18074)]
> 0x00007fd261b94330 in memset () from /lib/libc.so.6
> (gdb) bt
> #0  0x00007fd261b94330 in memset () from /lib/libc.so.6
> #1  0x00007fd253015073 in ?? () from /usr/lib/dri/fglrx_dri.so

And this is coming from the fglrx driver... Not a lot we can do there, 
other than ask ATI to fix it. Or you can get an nVidia card. Possibly 
it's fixed in newer releases of Ubuntu.

> I wish I could help you debug this stuff, but I have no idea where to
> start with your code. If it helps, you should still have access to
> this machine.

Well, I'm afraid both of these are out of scope for my abilities, so 
there's not much I can do. I will say that things have been very stable 
with the 1394 cameras on 32 bit linux, so if the illegal instruction 
thing becomes a problem for you, I'd suggesting moving your acquisition 
to a 32bit installation.

I am working slowly but steadily on getting the whole motmot stack into 
shape for the next Ubuntu release, which, as I said, might fix a couple 
of these bugs. It's difficult to be sure without trying it out, however...

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