[SMT-COMP] SMT-COMP 2016: Application Track Testing

Armin Biere armin.biere at gmail.com
Mon Jun 13 02:18:34 EDT 2016


Yep, now with pthreads (ubuntu 16.04):

$ ./test-ulimit-unlimited.sh
+ cat
+ ulimit -s
8192
+ gcc -o /tmp/stack-overflow /tmp/stack-overflow.c -pthread
+ time -f %E /tmp/stack-overflow
+ tail -1
Command terminated by signal 11
0:00.45
261955
+ ulimit -s unlimited
+ ulimit -s
unlimited
+ time -f %E /tmp/stack-overflow
+ tail -1
Command terminated by signal 11
0:00.17
65347



And here is the modified script for testing:

#include <stdio.h>
#include <pthread.h>
pthread_t thread;
void f (int i) { printf ("%d\n", i); fflush (stdout); f (i+1); }
void * run (void * dummy) { f (0); }
int main () { pthread_create (&thread, 0, run, 0); pthread_join (thread,
0); }
EOF
ulimit -s
gcc -o /tmp/stack-overflow /tmp/stack-overflow.c -pthread
time -f %E /tmp/stack-overflow|tail -1
ulimit -s unlimited
ulimit -s
time -f %E /tmp/stack-overflow|tail -1



On Mon, Jun 13, 2016 at 2:05 AM, Mate Soos <soos.mate at gmail.com> wrote:

> Hi Clark,
>
> On 06/12/2016 06:06 PM, Clark Barrett wrote:
> > Hi all - I believe we had this exact problem in the 2014 competition.
> > The default stack size for a thread is 2 mb unless you explicitly set it
> > to be more.  We fixed it by using the boost threading library and
> > explicitly setting the stack size.  For the competition, we asked David
> > Cok to rerun our patched executable as a noncompetitive entry (i.e. just
> > for comparison purposes) which he did.  Perhaps a similar compromise
> > would be appropriate here.  Sorry you guys got bit by this - I know how
> > frustrating that is.
>
> Thanks! Yep, a bit of a pain, though I look a this more as a way to
> learn, which I (hopefully) seem to do. Next time we'll make sure to ask
> ulimit to be set to a fixed value (like 8M) rather than "unlimited"
> which seems pretty limiting, actually :) To be honest, I'm not sure it
> would have made such a difference to the end results. I still need to
> work harder on some of the multi-threaded stuff.
>
> Thanks for the kind email :)
>
> Mate
>
>
>


More information about the SMT-COMP mailing list