There is good chances that the bug is in the user code (like calling
msg_comm_test twice or something similar), and there is nothing we can
do reliably from simgrid land if the user land builds such crap.
See also
https://gforge.inria.fr/tracker/index.php?func=detail&aid=14327&group_id=12&atid=165
action, action->comm.refcount, (int)action->state);
if (action->comm.refcount <= 0) {
action, action->comm.refcount, (int)action->state);
if (action->comm.refcount <= 0) {
- xbt_backtrace_display_current();
- xbt_die("the refcount of comm %p is already 0 before decreasing it. "
- "That's a bug!", action);
+ xbt_backtrace_display_current();
+ xbt_die("The refcount of comm %p is already 0 before decreasing it. "
+ "That's a bug! If you didn't test and/or wait the same communication twice in your code, then the bug is SimGrid's...", action);
}
action->comm.refcount--;
if (action->comm.refcount > 0)
}
action->comm.refcount--;
if (action->comm.refcount > 0)