As I was investigating the message passing subsystem I noticed that my
own blocks had a “system” port that I did not manually register.
Looking through the code I found in block.cc the special message to put
the block in a finished state:
void
block::system_handler(pmt::pmt_t msg)
{
//std::cout << "system_handler " << msg << “\n”;
pmt::pmt_t op = pmt::car(msg);
if(pmt::eqv(op, pmt::mp(“done”))){
d_finished = pmt::to_long(pmt::cdr(msg));
global_block_registry.notify_blk(alias());
} else {
std::cout << “WARNING: bad message op on system port!\n”;
pmt::print(msg);
}
}
Maybe I don’t understand the reason for this, but it appears that this
code requires pmt::cdr(msg) to return a long int and then implicitly
casts to a bool. Wouldn’t it be better for this to be a bool already
such that a message could be constructed like
pmt::cons(pmt::mp(“done”), pmt::get_PMT_T())
or Python
pmt.cons(pmt.to_pmt(“done”), pmt.PMT_T) ?
Perhaps it really doesn’t matter. It just seems more natural,
particularly if one were to put a logical statement as the argument to
pmt::cons.
Jared.