Adding an attribute to clang

Hi Sean,

I've been working on a patch to implement -finstrument-functions in
clang, which includes a no_instrument_function attribute. (See [1] for
my earlier patch to LLVM, which Chris Lattner recommended pushing
entirely into clang). When I updated today, I found a conflict with your
attribute work, and saw the comment about not touching without
contacting you.

What's the plan with regard to adding support for new attributes while
your work is in progress? I'd ideally like to be able to send my patch
for review before the end of the summer -- do you have a sense of when
you'll be unfreezing Or is it sufficiently stable enough that
adding a new trivial attribute should be fine?

(As an aside, having written a patch to add an attribute, +1 to cleaning
up the attribute system, and of course I'm not suggesting this should in
any way block your work. I mostly just want to have a sense of whether I
should put this aside in favor of other projects for a while)


- Nelson Elhage