Crash on resuming during program run (16.12 regression)

[robhagemans]

10 for i = 1to 10000:print i: next

Ctrl+C during run

./pcbasic.py --resume
Exception in thread Thread-1:
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 801, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.7/threading.py", line 754, in run
    self.__target(*self.__args, **self.__kwargs)
  File "/home/rob/Projects/basic-project/pc-basic/pcbasic/main.py", line 157, in run_session
    session = state.zunpickle(state_file).attach(iface)
  File "/home/rob/Projects/basic-project/pc-basic/pcbasic/state.py", line 82, in zunpickle
    return pickle.loads(zlib.decompress(f.read()))
TypeError: Required argument 'object' (pos 1) not found

[robhagemans]

This may have to do with the inability to pickle memoryview objects, see https://bugs.python.org/issue27098

[robhagemans]

• Does not happen if we simply define a variable i.
• Does not happen if we allow the program to finish and then exit.
• Does happen if we include a STOP statement and exit after the break.
• Also happens if we interrupt PC-BASIC with Alt+F4 while the program is running.

[robhagemans]

Another suspect is the ByteStream class which is derived from _pyio.BytesIO

[robhagemans]

10 i=1: goto 10 does not have the issue. FOR loops only?

[robhagemans]

interpreter.py lines 385, 386:

        counter_view = self._scalars.view(varname)
        self.for_stack.append((varname, counter_view, stop, step, step.sign(), forpos, nextpos,))

This stores a view in a structure that is to be pickled - may be the cause?

[robhagemans]

The memoryview is the issue. Apparently I'm storing most variables simply as bytearrays, but in a FOR-loop I'm keeping a copy of the Value object, which contains a memoryview. Adding __setstate__ and __getstate__ to Value to convert the memoryview to bytearray and back avoids the crash, but cuts the connection between the loop variable and the view of it which is being incremented.

[robhagemans]

As far as I can see, the loop record contains the Value objects counter_view, stop and step. Not sure why the Scalars object itself pickles without problems in the absence of a FOR-loop.

[robhagemans]

Fixed by 35e5640