[Darwin] Patch, take 2: enable linkonce support for Darwin

[Matt Austern]

This is very similar to my previous patch.  I've tried to address 
Richard's stylistic comments (target hook versus target macro, for 
example).

I also found one place where my previous patch wasn't quite right: on 
Darwin we need to make sure that explicit template instantiations are 
not marked coalesced.  This is tied up in the whole 
static-archive-table-of-contents mess.  I've fixed the bug (see 
maybe_make_one_only), and I have a test for this whole corner case.  
That test is not part of this patch, because it involves multiple 
compile and link steps and the creation of a static archive; I don't 
know how to put such a complicated test into dg.  When and if I learn 
how to do that, I'll submit it as a separate patch.

Still bootstrapped on OS X, still fixes a bunch of test cases in g++.dg 
and doesn't cause any regressions in the C, C++, or libstdc++ test 
suites.

OK to commit to  mainline?


			--Matt

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: weak-darwin-3.txt
URL: <https://gcc.gnu.org/pipermail/gcc/attachments/20040201/f4bbe341/attachment.txt>