[Octave-bug-tracker] [bug #59228] embedded.cc example crashes with segfault

[John W. Eaton]

Follow-up Comment #18, bug #59228 (project octave):

Markus, I don't think that the reference count for the static objects can ever
be zero but the static analyzer can't determine that.

But maybe I'm wrong.  We could test with an assert about the reference count
for the nil_rep object being zero.  If it is triggered, then I would like to
understand how and why that happens.

I would also like to move away from our custom reference counting and use
standard shared_ptr templates instead.  But in some cases, it seems tricky to
understand what is happening with reference counts.  For example, I was
looking at idx_vector yesterday and there is a lot of unexplained (or
insufficiently explained) magic in that set of classes.  Of course, this
indicates potential future maintenance trouble and I would like to improve
things to both avoid bugs and improve reliability.

When I've looked at the output of the clang static analyzer in the past, it
seemed to be so full of noise that it wasn't really helpful.  If I am just
misunderstanding the report, then I'd be grateful for someone to give me some
clues.

    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?59228>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/