Re: Warning with clang, error with gcc : attached files

[Bertrand Dekoninck]

On 2019-04-30 18:33:22 +0200 David Chisnall 
<gnustep@theravensnest.org> wrote:

> On 30/04/2019 15:34, Bertrand Dekoninck wrote:
> > OK, thanks for your quick answer. I don't see any either, but I 
> > could try 
> > to modify NSButtonCell.h locally to have this method public. Maybe I 
> > need 
> > more help.
>
> Normally, the fix for this kind of thing is to declare a category on 
> NSButtonCell that exposes the method that you want.  It's somewhat 
> fragile, 
> so best practice would be to have a dynamic check in your +initialize 
> that 
> checks that the method does exist.
>
> > Netherletheless, it doesn't explain why clang don't throw an error 
> > here. 
> > And I've got several warnings of this kind when compiling rik.theme. 
> > I 
> > suspect this is'nt the only private method that has be used is the 
> > rest of 
> > the code.
Thanks, I'll investigate this.

> Clang doesn't give an error because it's well-defined behaviour: the 
> method 
> will be assumed to return id and will be called assuming all 
> parameters are 
> of type id, unless otherwise specified.  That's almost certainly 
> wrong, which 
> is why it's a warning.
>
> In ARC mode, this is explicitly a failure (even if the calling 
> convention is 
> correct, the memory ownership semantics are unknown) and this is one 
> of the 
> many reasons why anyone using Objective-C in 2019 should be using ARC 
> for 
> almost everything.

If only I could... but Objc2 is still no an option on ppc linux and 
that's precisely why I want to port rik to objc1.

Bertrand