Hey,
I had a weird problem using GiNaC with g++-4.8. It seems that
GiNac::symbol will be deleted twice if it's not dynamically declared.
In other words, some resources in it is free'd before the automatic
destructor is called when the scope finishes.
The minimum code to re-produce the issue is:
#include <ginac/ginac.h>
int main() {
GiNaC::symbol x("x");
return 0;
}
$ g++-4.8 -g -o main main0.cc -lginac -lcln && ./main
main(69018) malloc: *** error for object 0x7fff73275570: pointer being
freed was not allocated
It seems this issue only exists in g++-4.8. g++-4.7 is fine.
I briefly looked at symbol.h but did not manage to locate the problem.
I was hoping someone here more familiar with the code base could fix
it or provide some direction for me. Thanks.
--
Song Gao
http://song.gao.io/
Hi,
I have read the issue of multithreading from 2010 and I wonder whether
there has been attempt to make GiNaC thread-safe. I think that it
shouldn't be a big deal to make the shared data like the refcounter
atomic (e.g. using std::atomic< T >) or to the synchronize it's
manipulation (e.g. using std::mutex).
Best regards,
Florian
--
Florian Corzilius
RWTH Aachen University
Computer Science Department, Informatik 2
D-52056 Aachen, Germany
http://www-i2.informatik.rwth-aachen.de/i2/corzilius/
phone: +49 241 80-21244