libgcc move to the top level

[Paolo Bonzini]

> I've declined to make this change.  I think it's a bad idea.

Ok.

>>> +$(patsubst %,%.vis,$(LIB1ASMFUNCS)): %.vis: %_s.o
>>> +	$(gen-hide-list)
>> Replace .o with $(objext).
> 
> OK, though there's little point - nothing anywhere in all of gcc
> sets it to anything else.

Yes, it's just the inconsistency that bothered me...

>>> +libgcc.a libgcov.a libunwind.a libgcc_eh.a:
>>> +	-rm -f $@
>>> +
>>> +	objects="$(objects)";	
>> No blank lines (i.e. no lines with the tab only).
> 
> It doesn't have a tab.  You don't need one; completely blank lines are
> ignored by make.  Do you really want me to remove the blank lines?  I
> think they make it more readable.

No, thanks for educating me this time. :-)

> 2) The extra-parts variable is necessary because all isn't the only
> thing to depend on it - the shared libraries do too and there's an
> extra enable_shared conditional there.  It would be possible to rework
> this, but I think it's much simpler the way it is.  There's an easy
> way to disambiguate it though; I renamed extra-parts: to
> libgcc-extra-parts:.

Right, much better.

> Here's the incremental diff.  I built and inspected libgcc on
> x86_64-linux.

Okay, if you can run a few more bootstraps (I'd say, at least 
powerpc-linux and powerpc64-linux, I'll take care of powerpc-darwin), I 
have no problems with the merge.

Maybe we should disable gcc-only bootstrap for a week before merging the 
branch, so that people running automated testers know that they're doing 
the wrong thing.  In the meanwhile, it'd be great if you convince one or 
two arch maintainers to help moving stuff to libgcc, or even better work 
on some documentation.

Arnaud, are there any problems with Ada and configuring with 
--enable-bootstrap?  I think I fixed them all a while ago, but it's 
better to double check.

Thanks again for the work,

Paolo