Bug in except handlers: the caught exception symbol should be cleared at the end of the except clause.

[tristanlatr]

Hello @serge-sans-paille, hope you are well.

I've just found out that the reasoning around the try:/except Exception as e: nodes is boggus, here is the relevant part of the documentation:

When an exception has been assigned using as target, it is cleared at the end of the except clause.
This is as if

except E as N:
    foo

was translated to

except E as N:
    try:
        foo
    finally:
        del N

This means the exception must be assigned to a different name to be able to refer to it after the except clause. Exceptions are cleared because with the traceback attached to them, they form a reference cycle with the stack frame, keeping all locals in that frame alive until the next garbage collection occurs.

Reference: https://docs.python.org/3/reference/compound_stmts.html#the-try-statement

So a reproducer for this issue is the following code (which directly taken from our test suite):

try: 
    open("/dev/null")
except Exception as e: 
    pass
e

This should generate an unbound identifier warning for e.

[serge-sans-paille]

That's a very surprising behavior! I have a patch, still need to double check it. And indeed it's very similar to #112

[tristanlatr]

Hey @serge-sans-paille, what is the idea behind your patch? I could try to implement it if you lack of time. Thanks

[serge-sans-paille]

I'm trying to solve #112 first, and my current approach would be to add a list that contains the currently deleted nodes, per level/scope

[serge-sans-paille]

Approach changed: I'll consider an ast.Del as an ast.Load (as usual) plus a silent special store which, if reached, is ignored. It seems to blend elegantly in current approach, more about this later :-)