[Octave-bug-tracker] [bug #64749] VM seems to be out of sync in very rare situations for function parameters

[Petter]

Follow-up Comment #1, bug #64749 (project octave):

Thanks for the test.


Trace:
sp  : 0x7f92901b1088
bsp : 0x7f92901af1e0
sp i: 981
sp ii: 1018
ip  : 61942
code: 0x7f92b773aec0
data: 0x7f921fabf7f0
ids : 0x7f921fec6860
fn  : fname
Next op: 140 // WIDE

Trace:
sp  : 0x7f92901b1078
bsp : 0x7f92901af1e0
sp i: 979
sp ii: 1016
ip  : 61947
code: 0x7f92b773aec0
data: 0x7f921fabf7f0
ids : 0x7f921fec6860
fn  : fname
Next op: 140 // WIDE


Something seems strange here. There should be some opcode using a slot after
each WIDE.

Is the function's bytecode 61947+ bytes long? If that is the case it might be
some 2^16-1 issue.

I am thinking about manipulating the bytecode compiler to always emit wide
slots and run the test suite, to see if it catches something. It would maybe
be a good idea to add some halt opcode or something and map it to zero and pad
the bytecode with zeroes.


    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?64749>

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