I have my doubts that this feature ever worked 100% correctly.  It probably stopped functioning at all when we switched to using exceptions to handle interrupts and errors instead of setjmp/longjmp.

RE comment #8: how does it work for you?  It doesn't work for me.

Something like the attached change might be a starting point, but my limited testing shows that there are definitely problems with this change so I think it will have to wait until after version 7.

Setting release to "dev".


(file #52325)

It doesn't work for me with Octave 4.2.2 on Windows 10. Maybe it was broken on that platform for a different reason.

I went back to version 3.8.2 and it worked there.

The idea is to enter the debugger at the next available point after the interrupt is noticed by the interpreter.  But I don't think it has worked properly for a long time, probably since we started throwing exceptions to interrupt the interpreter.

I recently discovered the existence of debug_on_interrupt which I wanted to use as a way to interrupt temporarily a calculation and inspect variables before proceeding.

Unfortunately I was not able to have it work: whether debug_on_interrupt is true or false, Ctrl+C interrupts the script and returns to the prompt. For example I saved the following function in a file toto.m (which attached to this report), and then ran it but Ctrl+C interrupts it rather than leading to debug prompt.

function retval = toto (input1, input2)
  debug_on_interrupt (true, "local");
  while (true)
    a = 2;
  endwhile
endfunction