Re: [PATCH 15/16] gdb: make cmd_list_element var an optional union

[Simon Marchi via Gdb-patches <gdb-patches@sourceware.org>]

> I took a quick look, and that seems reasonable to me.  I don't think it
> will conflict (in design) with my upcoming patch.  My upcoming patch
> makes it so that some settings don't use the storage that is in
> cmd_list_element at all, but instead provide some getter/setter
> callback.  This is to fix display of settings where the "source of
> truth" is not cmd_list_element::var.
> 
> With my patch, the code at various places becomes, conceptually:
> 
>   if (cmd_list_element::var is set)
>     use cmd_list_element::var
>   else
>     use the getter/setter
> 
> I think it would be nice to make the first case (using
> cmd_list_element::var for storage) just a specific case of using the
> getter/setter.  That is, when we create a setting that uses
> cmd_list_element::var for storage, we install a getter and a setter than
> get and set the value in cmd_list_element::var.  But I haven't gotten
> there yet.  You'll probably have some good ideas for achieving this :).
> 
> But before posting that patch, I'd like to decide how the current patch
> series will end up, because that will very much affect how the next
> patch will look like.

I realize I wasn't accurate here.  The storage isn't actually in
cmd_list_element::var, cmd_list_element::var only contains pointers to
the storage.  But the rest still stands.

Simon