Re: [PATCH] gdb/python: new selected_context event

[Tom Tromey <tom@tromey.com>]

>>>>> "Andrew" == Andrew Burgess <aburgess@redhat.com> writes:

Andrew> This commit introduces a new Python event, selected_context.  This
Andrew> event is attached to the user_selected_context_changed observer, which
Andrew> triggers when the user changes the currently selected inferior,
Andrew> thread, or frame.

Thanks.  I think this approach seems fine.

FWIW this is https://sourceware.org/bugzilla/show_bug.cgi?id=24482

Andrew> Adding this event allows a Python extension to update in response to
Andrew> user driven changes without having to poll the state from a
Andrew> before_prompt hook, which is what I currently do to achieve the same
Andrew> results.

Yeah, that approach is ok-ish but also problematic.

I think we have similar issues in the TUI, see

https://sourceware.org/bugzilla/show_bug.cgi?id=32865

Andrew> I did consider splitting the user_selected_context_changed observer
Andrew> into 3 separate Python events, inferior_changed, thread_changed, and
Andrew> frame_changed, but I couldn't see any significant advantage to doing
Andrew> this, so in the end I went with just a single event, and the event
Andrew> object contains the inferior, thread, and frame.

This seems totally fine to me.

Andrew> Additionally, the user isn't informed about which aspect of the
Andrew> context changed.

This too.

Andrew> +The event is of type @code{gdb.SelectedContext} and has the following
Andrew> +attributes:

There's been a convention that an event type name ends in "Event".

Andrew> +    frame_obj = gdbpy_ref<> (Py_None);

This has to be gdbpy_ref<>::new_reference (Py_None);

Unfortunately even the safety series I am working on won't prevent this
kind of problem :-(

I guess we could do it by forbidding the use of plain "Py_None" somehow,
and then having a gdb variant that is clearly a borrowed reference.

Tom