C++ lexical analysis rework

[Eric Botcazou]

> On the other hand, it's quite possible that *I* care a great deal.
> Apple uses stabs.

That was my impression too. :-)  Does Apple use the DBX_USE_BINCL feature?

> The one thing that puzzles me is why we haven't noticed this problem in
> our internal testing.  What's the symptom of this bug?

No B_INCL/E_INCL lines in the STABS info, see my previous message.   And 
indirectly PCH failures in the C++ testsuite, but you might not see them 
because of your -gused stuff.

-- 
Eric Botcazou