Re: [objc-improvements] Just remembered something about multiplemethod signatures...

[David Ayers]

Richard Frith-Macdonald wrote:

> 1. If the compiler has no signature for a method ... it should 
> generate an error.
> 2. If the compiler has multiple conflicting signatures ... it should 
> generate an error.
> 3. If the compiler has multiple signatures which would generate the 
> same code, it should generate a warning.
>
> It sounds like telling the difference between (2) and (3) may be too 
> hard (I certainly don't know how to code that comparison) ... in which 
> case we have to assume (2) for now, but making the distinction between 
> the two cases would be a very good thing.
This is a compromise I can definitely conceed to.  But before giving up 
on whether or not we're able to tell the difference between (2) and (3), 
I would like to thow up two suggestions for grabs, that effectively 
distinguish between 'conflicting' and 'compatible' prototypes in my 
previous mail. 

a) My first thought was to check return value and arguments in turn, to 
assert that all corresponding components agree on the same size as 
determined by sizeof().

This should effectively keep generating "working" code for 'compatible' 
but 'conflicting' prototypes.  The issue this appoach still raises is 
that we would have to 'construct' a prototype to pass to the C part of 
the processing, which may otherwise still emit an error (NSSize/NSPoint 
example) about conflicting types.  This construction of a C-type, may in 
the end be the complexity, that renders this approach not worth the 
trouble or even impossible.  (Besides the fact that others have already 
effectively veto'ed this :-) )

b) A simpler assumption would be to assume 'id' for all occurances of 
object references.  I think this is the 'merge' that Alex had been 
refering to (and has also been proposed by Nicola in the meantime).  We 
would still be 'constructing' a prototype, but it should be (relatively) 
trivial to implement after Alex's patch has been commited.  And I 
believe that will cover almost all potentially 'legitimate' cases of 
'conflicting' but 'compatible' prototypes.

If we cannot find a consensus, I can also conceed to retaining the old 
behavior.

Cheers,
David