Re: [rfc] expose gdb values to python

[Tom Tromey <tromey@redhat.com>]

Tom> There are some tradeoffs here.  The "invisible access" approach is
Tom> convenient.  However, it runs into issues with odd programs -- say,
Tom> multiple inheritance where a given name refers to multiple fields.

Thiago> Adding a way to do casting in python solves this, right?

Yes.

Tom> I tend to like something toward the raw side, partly because any
Tom> cooked approach will still need some second way to deal with the
Tom> underlying explicit types.

Thiago> Sorry, I didn't understand "deal with the underlying explicit types".

That is just a funny way of saying, we'll still need the casting, and
of course some way to get the list of superclasses.

Tom> I propose we decide these questions and implement this before checking
Tom> in this patch.  The semantics of Value are critical.

Thiago> Right. I suggest people to take a look at the testcase to see
Thiago> what the syntax looks like, and speak up if there are any
Thiago> concerns or preferences.

The tricky cases all occur when using inheritance and having
conflicting names.

Tom