[patch/rfc] Moving constraint definitions to the machine description

[Joern RENNECKE]

DJ Delorie wrote:

>>I don't see why you would want to change a standard letter
>>assignment.  For a new port, you can use multi-letter constraints
>>for the less frequently used stuff.
>>    
>>
>
>The m32c port uses very few of the standard constraints.  For the most
>part, none of them are useful.  Most constraints are multi-letter
>ones, and I tried to use the leading letter to group them.  I *really*
>wanted to use 'M' for memory constraints, but couldn't, so I used 'S'
>instead.
>  
>
All right, so maybe it makes sense in such special circumstances.  I'd 
still prefer
if the port writer / maintainer has do do something special to re-designate
standard constraint letters, e.g. add a force flag, or use a different 
define_* name
(or redefine_constraint ?) that expresses that something special is 
going on.
This not only prevents accidential redesignation of a standard letter 
(except for the
mindless copy-an-arcane-and-totally-different-port-as-a-template 
scenario ;-) ,
but also allows a quick grep for this construct to check for any nonstandard
uses of standard letters.