[4.0 PATCH] Fix for PR21951

[Jan Hubicka]

> On Sat, 11 Jun 2005, Dan Kegel wrote:
> > C++ programs compiled with -fno-exceptions fail with gcc-4.0.0
> > because of false positive warnings about control reaching
> > the end of a inlined function without returning a value
> > (see http://gcc.gnu.org/PR21951).
> 
> Yes, that's a bad one.
> 
> > 	PR c++/21951: avoid appearing to hit end of function when compiled
> > with -fno-exceptions
> > 	* libstdc++-v3/include/bits/stl_vector.h
> > 	* libstdc++-v3/include/bits/stl_uninitialized.h
> > 	* libstdc++-v3/include/ext/rope
> > 	* libstdc++-v3/include/ext/memory
> > 	* libstdc++-v3/include/ext/hashtable.h
> 
> Either this, or someone tries to convince Honza to backport his fixes
> from mainline. (I believe it was Honza who fixed this by virtue of his
> inline work, but I may be wrong.)

I don't really know I fixed such a problem at all, but if this was a
part of CFG inliner work (where I really moved this warning around that
probably resulted in better diagnostics), there is no chance to make
similar patch for release branch :(

Honza
> 
> Gerald