Re: [PATCH 1/3] gdb/python: add type traits check for all PyObject sub-classes

[Andrew Burgess <aburgess@redhat.com>]

Tom Tromey <tom@tromey.com> writes:

>>>>>> "Andrew" == Andrew Burgess <aburgess@redhat.com> writes:
>
> Andrew> The problem is that Python is written in C, and PyObject_New doesn't
> Andrew> call any constructors for `some_new_type`, nor for any of the fields
> Andrew> within `some_new_type`.
>
> Since working on the Python safety series, I have been wondering if we
> could remedy this.  That is, give gdb's classes constructors and
> destructors and arrange for these to be called in-place in tp_init /
> whatever the destroy one is.
>
> This way we could use idiomatic gdb code, which would be safer.
>
> I was planning to investigate this more deeply once Matthew Longo's work
> was done, since I didn't want to touch all the type objects and cause
> conflicts.
>
> Andrew> +
> Andrew> +template <typename T>
> Andrew> +struct is_python_allocatable {
> Andrew> +  static constexpr bool value =
> Andrew> +    std::is_trivially_destructible_v<T> &&
> Andrew> +    std::is_trivially_copyable_v<T>;
>
> Brace and operator placement.

Fixed.  Thanks.

>
> Otherwise this looks good to me.
>
> If the tp_init experiment works out I suppose we may end up reverting
> this.  But IMO it's best not to wait on something that might not
> happen.

ACK.  If/when you have a better solution I'm happy for this series, at
least the static_assert bits and the associated test, to be removed.
This was really just me trying to find a way to "test" the fix that is
in patch #2.  If we can come up with a solution that means we can use
C++ objects and the constructos/destructors "just work" then I'm 100% in
favour.

For now though, I've gone ahead and pushed this series.

Thanks,
Andrew

>
> Approved-By: Tom Tromey <tom@tromey.com>
>
> Tom