Re: Lisp watchpoints

[Stefan Monnier]

>> The let-binding is currently trapped, but not the corresponding
>> unbinding.  Is it better to skip both?  As far as I can tell, it's not
>> feasible to trap the unbinding.

The unbinding currently doesn't check CONSTANT_P, but that's just an
optimization that becomes invalid for trapped variables.  I don't see
why we couldn't trap the unbinding just like we trap the binding (at
the cost of an extra check of the `trapped' attribute).

In terms of debugging, I think trapping let-binding and let-unbinding is
definitely useful, so I think it's worth the effort to try and catch
both cases.

> I should clarify that this only applies to dynamically-bound
> variables, lexical variables are never trapped.

Yes, that's normal and right.  Fundamentally, lexical variables are
anonymous, anyway.


        Stefan