Faster compilation speed

[Richard Kenner]

    It also removes a pernicious variety of bug that often caused nasty memory
    corruption in earlier versions of GCC. 

Yes, but whenever that happened, it represented a scoping problem in the
front end.  If the entities in question only involved constants, switching to
GC indeed "removed" the bug.  But in most of these cases, the problem could
also occur where non-constants are involved.  In that case, what we've done
is to replace a memory corruption problem in the compiler which causes a
crash with a bug that generates subtly wrong code.  Not a good trade, in my
opinion.  In most cases, though, what this does is that it makes the scoping
bug become latent.