[Octave-bug-tracker] [bug #53733] Panic/segfault when calling Java method after forge:im[read|write]

[Mike Miller]

Follow-up Comment #60, bug #53733 (project octave):

The GraphicsMagick code isn't in a loadable dldfcn oct file anymore, it's in
builtin functions now.

So based on what we've found so far, if I had to summarize this, it's that it
has to do with passing large arrays in to Java functions via JNI. There is
some size at which Octave goes into a 100% busy state (and never returns?
takes an exponentially long time?) but does not crash. And there is some
larger size at which Octave crashes immediately with a SIGSEGV.

For some as-yet-unknown reason, calling any GraphicsMagick functions in Octave
before the Java function is called drastically reduces these size limits.

Ideally, Octave would know about this limit or be able to detect this
condition dynamically and print an error message that some size threshold has
been crossed and avoid the segmentation fault.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?53733>

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