Ok, I shrunk the images down so they are smaller in byte size, but they are 
still fairly large in dimension.  You can download the two of them here:

http://www.bluelavalamp.net/hugin/enfuse-openmp-segfault/

I did 20 runs of enfuse and got 8 segfaults.  I then re-compiled adding in 
-g as you said and ran in gdb.  Here's the output then it segfaulted:

(gdb) run
Starting program: /usr/local/bin/enfuse-test 0.jpg 1.jpg -o 
crash_small_test.tif
warning: Could not load shared library symbols for linux-vdso.so.1.
Do you need "set solib-search-path" or "set sysroot"?
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
enfuse: info: input image "0.jpg" does not have an alpha channel;
enfuse: info: assuming all pixels should contribute to the final image
enfuse: info: input image "1.jpg" does not have an alpha channel;
enfuse: info: assuming all pixels should contribute to the final image
enfuse: warning: no usable resolution found in first image "0.jpg";
enfuse: warning:   will use 300 dpi
enfuse: info: loading next image: 0.jpg 1/1
[New Thread 0x7fffcb88c700 (LWP 2233)]
[New Thread 0x7fffcb08b700 (LWP 2235)]
[New Thread 0x7fffca88a700 (LWP 2237)]
enfuse: info: loading next image: 1.jpg 1/1

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffca88a700 (LWP 2237)]
operator() (this=0x89cc00 <Twister>) at 
/usr/include/boost/random/mersenne_twister.hpp:454
454         UIntType z = x[i];
(gdb)

Hope this helps!
Kevin


On Wednesday, November 28, 2012 10:59:19 AM UTC-5, cspiel wrote:
>
> Kevin -
>
> Am Mittwoch, 28. November 2012 12:54:03 UTC+1 schrieb kevin360:
> > I reduced the number of input images to 2 and it still segfaults.
>
>         Nice job!  Two is perfect for testing.
> If the images aren't too large, do you think it
> makes sense to supply them to us?  I'd like to
> reproduce your findings on my system.
>
>
> > Next I reduced the image size which caused the segfaults to go away.
> > I increased the size and it came back.  I'm using images that are
> > 9,000 pixels on the long side right now.  Doing 40 runs with 2 of
> > those images it segfaulted 12 times.  Turning off openmp from the
> > shell (OMP_NUM_THREADS=1) and it completes every run.
>
>         The OMP_NUM_THREADS=1 test proves that
> either the OpenMP synchronization is faulty or
> at least one thread tramples on another one's
> memory.  None of which I have ever heard before
> for Enfuse code as its OpenMP code is
> extraordinarily simple.
>
>
> > Next I tried to compile in the debug support but that crashes the
> > compile.  Just to be sure I re-ran configure and just removed the
> > --enable-debug, compiles fine.  Run configure again putting back in
> > --enable-debug, make clean, make and it crashes.  Here's the command
> > where it crashes:
>
>         Doh, I should have been more specific.
> I meant: ensure that the "-g" option is passed
> to the compiler _and_ the linker when I said
> "compile with debug support".  This makes
> backtraces much easier to read.
>
> Whether you configure with or without debug
> support, i.e. `--enable-debug' or
> `--disable-debug' should not matter.
>
>
> One more idea: If you have a suitable
> "libgomp.so" (i.e.  configured with
> `--disable-linux-futex') at hand, you can run
> Enfuse under the supervision of Helgrind (after
> preloading the correct "libgomp.so").  No idea,
> whether this will unveil the cause of the
> problem, though.
>
>
> HTH,
>         Chris
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"Hugin and other free panoramic software" group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/hugin-ptx

Reply via email to