Issue A-24

[Christophe de Dinechin]

Matt,


This proposal looks fine, but is a bit different that what I wanted to do. My
initial intent was to try to create a situation where (without modifying the
linker):

- If the complete type info is there, the linker or dld resolves to it for all
(pointer) typeinfo that use it.

- If the complete type info is not there, then we keep multiple copies, and only
in that case do you pay the price for name comparison.

I did not find a proper way to do that. So currently, I actually have no real
proposal, and yours seem interesting.


Christophe


Matt Austern wrote:
> 
> I think it does work, provided that a __class_type_info for an
> incomplete class can appear only as the end of a __pointer_type_info
> chain.

I believe this is true (maybe you need to include references)


> (Christophe suggested a new class, __incomplete_class_info.  I see
> no reason for it, though.  We don't need any information for
> incomplete classes that we don't already have in __class_type_info.)
> 
I agree, once you have the bit indicating incompleteness at the top level.


> The __class_type_info for an incomplete type is mangled differently
> than the typeinfo object for a complete type.

Why? Note: I see no reason either way, but we are a bit short in mangling
prefixen.