fl.source arg changes crash

Issue #16 resolved
st kr created an issue

i had a [fl.source~ /size 4096 1024] object in a patch by mistake. it is being driven by params. i changed it to [fl.source~ 4096 1024]. crash.

crash log and patch included in attached zip. it is a messy thing made to test another wrong messy abstraction i am trying to make. but hopefully useful.

Comments (10)

  1. Alex Harker repo owner

    Thanks. That was a tough one.

    It was actually to do with memory corruption from binary objects with mismatched length inputs. My loop tests in wrap mode were incorrect in the case that the larger length is not a multiple of the smaller...

    This patch just happened to demonstrate the bug when the vector from fl.recall~ changes size, but it could have appeared anywhere - fl.source~ was just inducing it much later in the network.

    Fixed for the next one (after several hours of head scratching).

  2. st kr reporter

    i attach another crash to this thread as it is likely just the same issue (same patch), but just in case, you should see. it is strange, sometimes the frame lib chain seems to just stop working - not passing frames, although msp is on and working.

  3. Alex Harker repo owner

    When you say 'just stops working', do you mean it is running fine then it is not (no change), or "I have patched it differently and then it stops working"?

    There are scenarios in which you'll get no output because something is wrong somewhere in the chain (that is normal). A working patch suddenly not working is not normal, but I need something I can reproduce.

  4. Alex Harker repo owner

    The crash looks the same/related. I'd ask you to firmly test on the next alpha, which I'll try to get out tonight.

  5. Alex Harker repo owner

    Also, I advise using fl.length~ and fl.now~ to debug whether the issue is time moving correctly or whether the frames are empty or not!

  6. Log in to comment