Logged In: YES 
user_id=85414

I've attached a new patch, which implements most of
your suggestions:

 o The Jump tests now have their own test case class
   in test_trace.py.
 o There's now a test for "lineno must be an integer".
 o I've added your nested-finally-jump test.
 o I've added a test for each of the four (four!! 8-)
   types of 'except' line.
 o I've clarified the docs.  I don't have a machine
   to build them on - are you in a position to check
   that they build and look sensible?
 o Jumps fail if the frame has no trace function set,
   and there's a new test in test_trace.py for that.
 o pdb complains if you try to do a jump in any frame
   other than the bottom one.

I'd have liked frame_setlineno to police the fact
that you can only jump in the bottom frame of a trace,
but I couldn't figure out how to do that.  It would
have implemented the final two points above in one go.

As far as I can see, you're right that you can't get
frame objects or set trace functions in restricted
mode.  It would be trivial to reject writes to
f_lineno in restricted mode (frame objects have an
f_restricted flag that frame_setlineno could look at)
but given that it would be impossible to test, I
haven't implemented it.

I don't get any char / unsigned char warnings under
MSVC6 (even on the highest warning level).  I've made
a guess at where they might be and 'fixed' them, but
I'd be grateful if you could check, and fix them if
they're still there.

I had a go at writing the finally-policing code in
pseudo-python, but it didn't come out significantly
simpler than the C - it has to work the same way as
the C in order to be useful, but that makes it pretty
much the same.  Hopefully the comment explains the
workings sufficiently well.