Re: [lto] PATCH: fill in code to merge declarations

[Diego Novillo <dnovillo at redhat dot com>]

Mark Mitchell wrote on 09/20/06 18:30:

> That's not the issue to which I'm referring: I was talking about global 
> data in GCC, not global data in the user's program.  I am concerned that 
> setting up the CFG (and preventing GCC from rebuilding it) may not be 
> the only bit of state that has to be reset when skipping the early 
> passes in the compilation process.
> 
Indeed it isn't.  I went through quite a bit of grief even when doing
the simpler creation of outline functions for parallel OpenMP regions
(move_sese_region_to_fn)

We have a fair amount of global state.  It's a bit better than it used
to be, thanks to the call graph framework.  But it's going to be an
interesting exercise.