[patch] Finish cleaning up gimple.h

[Jeff Law]

On 11/15/13 12:41, Andrew MacLeod wrote:
> On 11/15/2013 01:32 PM, Jeff Law wrote:
>> On 11/15/13 09:47, Andrew MacLeod wrote:
>>> This patch finally gets gimple.h into its proper state of only having
>>> prototypes for gimple.c.
>>>
>>> The prototype for get_base_address was in tree.h, and it seemed more
>>> appropriate there since gimple.h is for gimple statements.  I moved the
>>> function to tree.c
>>>
>>> 2 new .h files were created,  cfgexpand.h and tree-nested.h.   I flushed
>>> both of those out to include all the exports for the files, so the
>>> 'build_addr' prototype was moved from tree.h and
>>> 'estimated_stack_frame_size' was moved from tree-inline.h. I looked at
>>> shuffling the exports to other files, but nothing jumped at me.
>>> estimated_stack_frame_size depends on too much other stuff in cfgexpand
>>> to move even though it only has one client.
>>> build_addr may belong somewhere else, but Im not sure where... tree.c as
>>> well maybe?  The fortran FE uses it as well...
>>>
>>> A few #includes needed to be added to .c files for the moved prototypes.
>>>
>>> I left the macro definitions for SCALE, LABEL and PERCENT at the bottom
>>> of gimple.h.  I was going to move them, but really have no idea where to
>>> put them. I thought about coretypes.h but that doesn't seem right
>>> either.  suggestions?
>> I think PERCENT is only used by tree-into-ssa.c, so shove it in there.
>>
>> SCALE and LABEL are only used by tree-cfg.c and tree-dfa.c, if that
>> helps you narrow down a potential location.
> SCALE is also defined and used in ggc-common.c, ggc-page.c and
> input.c... I'm still not sure where a *good* common place would be.
> There is no need for those 3 places to redefine it.... (the defs are all
> basically the same)
Ah, I'd filtered those others out because they had defintions in their 
.c files.

You could argue for system.h given it has MIN, MAX, CEIL (/me ducks).


Jeff