Re: [cgraph] move eh marking to except.c

[Jan Hubicka <jh at suse dot cz>]

> On Thu, Sep 11, 2003 at 01:31:58AM +0200, Jan Hubicka wrote:
> > > One, moves it out of language code, which is good.  Two, moves it to
> > > where we know if the data is actually used, which is also good.
> >
> > Won't this break when the outputing of variable will refernece other
> > functions that are not accessible previously so already freed out of
> > datastructure, or is this impossible?
> 
> Eh?  Why wouldn't it work?  Won't we notice some function 

Because in unit-at-a-time we first decide what functions needs to be
emit and do intraprocedural optimizations based on it.  It is critical
that we don't add new functions into the list during the later expanding
pass.  cgraph_analyze_function is supposed to notice all places that
would later result in the references.

assemble_name marking is used mostly to allow some stuff to be output before
cgraph_global_info_ready is set (originally before varpool code has been
added variables were such unknown garbage and still this is how variables are
output, but in future I plan to build complete dependency graph and do
the analysis, so when we optimize away reference to variable, we will be
able to optimize away everything referenced by that variable)

Honza
> referenced by the variable when the reference is emitted
> in assemble_integer?
> 
> If it doesn't, it should.
> 
> Also, I'm thinking that this patch too was wrong, and that
> we should consider backends that don't use at least
> assemble_name as buggy.
> 
> 
> r~