Jason R Thorpe <thorpej@wasabisystems.com> writes:

> On Fri, Oct 19, 2001 at 03:56:07PM -0400, James Chacon wrote:
> 
>  > Any even then reading these changes that would affect the apps ability
>  > to compile, not run. The struct should still map to the same place in the ABI
>  > with these changes.
>  > 
>  > Making stuff not compile is a lot less worrisome when it's stuff that's 
>  > already violating specs.
> 
> I agree... 
> 
> So, really, I'd change _ub to be like this:
> 
> 	union {
> 		struct __sbuf _ubu_ub;
> 		void *_ubu_intcookie;
> 	} _ub_u;
> 
> to really insure that we break compilation of apps that use _ub.

Perhaps I'm missing something here, but Yamamoto-san's change simply
renames the structure member from _ub to _ext; isn't that sufficient?
(Besides, like Todd I can imagine a processor ABI breaking binary
compatibility by introducing this union.)


- Klaus