why include/bits/std_XXX.h?

[Benjamin Kosnik]

Again, I ask you to provide an independent way of verifying your 
assertion about make deleting like-named header files. 

To the best of my knowledge, you have never done so. 

Because of this, I consider elements of the current header design 
suspect. Namely, the existence of include/bits/std_xxxx.h, where xxx 
is the name included in paragraph 2 of [17.4.1.2 - Headers]. Note, 
explicitly excluded is paragraph 3, "C++ Headers for C Library 
Facilities." (Because "C" compatibility headers are still a 
work-in-progress, additional flexibility is required for them.)

I consider the existence of these files predicated on an assumption that 
is demonstratedly false. Furthermore, I believe I have demonstrated this 
in my last email.

> Nobody has claimed to have fixed the bug, verified that it was fixed, 
> or even verified that any GNU make maintainer ever admitted that it 
> is a bug.

Please prove the existence of this bug. Then, prove existence in the 
libstdc++ Makefiles, since that is the domain under discussion. I haven't 
seen evidence of either case.

> It should work correctly.  

That's not in dispute. I'm asking you to re-justify your initial 
assumptions about header design. I don't consider this unreasonable. 

If they no longer apply, let's simplify.

-benjamin