[vta, vta4.4] merged with trunk and 4.4 @149247, updated VTA patchset

[Richard Henderson]

On 07/07/2009 02:10 AM, Alexandre Oliva wrote:
> On Jul  6, 2009, Richard Guenther<richard.guenther@gmail.com>  wrote:
>> Why do we have the VAR_DEBUG_VALUE tree code?
>
> It's a subcode of GIMPLE_DEBUG.  Currently, it's the only valid subcode.
> I'm leaving the door open to add other kinds of GIMPLE_DEBUG stmts, such
> as stmts to note relevant points of the source code that should be
> marked with is_stmt in .loc tables, so as to get nicer single-stepping
> in optimized programs.  This hasn't been implemented yet, of course.

It doesn't need to be a tree code to be a subcode.  Indeed, I don't
think you should be creating a tree code just for this purpose.

I'd suggest leaving the subcode zero until you do add other kinds of
GIMPLE_DEBUG statements.  At which point see enum gf_mask and do what
needs to be done for whatever kind of statement you're inventing.


r~