Kill the deferred string table

[Gerald Pfeifer]

On Sun, 27 Apr 2003, Zack Weinberg wrote:
> As suggested earlier, this patch eliminates the deferred string table;
> instead, I set SYMBOL_REF_DECL for constants to point to the tree for
> that constant, which allows mark_constant to do the right thing with
> STRING_POOL_ADDRESS_P constants.
> [...]
> At -O2 both function calls are eliminated, and so is the string
> constant

Zack, would this warrant mention in gcc-3.4/changes.html (and if so,
would you mind adding something)?

I don't know how likely that may be, but some users might have relied
on this in some way (though I definitely don't see any reason we should
add a flag or something to retain the old functionality!) and, after all
it's also a relevant optimization that GCC used to lack.

Gerald