[COMMITTED] Fix c++-types-check conditionalization.

  Committed as obvious.

I noticed this because the power8-linux build slave hit it (and another
patch to follow).  It's nice that we fix all the gratuitous failures when
no C++ compiler is available.  But we do want the bots to have C++
compilers available, so the power8 bot should get whatever it's missing
installed.


Thanks,
Roland


2016-03-04  Roland McGrath  <roland@hack.frob.com>

	* Makefile ($(objpfx)c++-types-check.out): Fix conditionalization
	to test for empty $(CXX) rather than $(CXX) of "no".
  

Roland McGrath <roland@hack.frob.com> writes:

> Committed as obvious.
>
> I noticed this because the power8-linux build slave hit it (and another
> patch to follow).  It's nice that we fix all the gratuitous failures when
> no C++ compiler is available.  But we do want the bots to have C++
> compilers available, so the power8 bot should get whatever it's missing
> installed.

It was missing glibc-static and it should be fixed now.

Thanks!
  

On 03/04/2016 04:25 PM, Roland McGrath wrote:
> Committed as obvious.
> 
> I noticed this because the power8-linux build slave hit it (and another
> patch to follow).  It's nice that we fix all the gratuitous failures when
> no C++ compiler is available.  But we do want the bots to have C++
> compilers available, so the power8 bot should get whatever it's missing
> installed.
> 
> 
> Thanks,
> Roland
> 
> 
> 2016-03-04  Roland McGrath  <roland@hack.frob.com>
> 
> 	* Makefile ($(objpfx)c++-types-check.out): Fix conditionalization
> 	to test for empty $(CXX) rather than $(CXX) of "no".
> 
> diff --git a/Makefile b/Makefile
> index 7fc92ae..9c6918d 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -252,8 +252,8 @@ tests-clean:
>  	@$(MAKE) subdir_testclean no_deps=t
>  
>  tests-special += $(objpfx)c++-types-check.out $(objpfx)check-local-headers.out
> -ifneq ($(CXX),no)
>  
> +ifneq (,$(CXX))
>  vpath c++-types.data $(+sysdep_dirs)
>  
>  $(objpfx)c++-types-check.out: c++-types.data scripts/check-c++-types.sh
> 

This is the wrong way to fix this. You've hidden the failure :-(

Please see:
https://sourceware.org/ml/libc-alpha/2015-08/msg01280.html
https://sourceware.org/ml/libc-alpha/2015-08/msg01284.html
  

I correctly fixed the bug in the conditionalization.  I did not fix the
separate issue of this test not supporting UNSUPPORTED (which I had indeed
forgotten about).  Feel free to fix that one.