[cfe-dev] Objective-C @defs support

[David Chisnall]

On 21 Jun 2008, at 19:24, Chris Lattner wrote:

> On Jun 18, 2008, at 7:28 AM, David Chisnall wrote:
>> This patch adds parsing and code generation support for @defs  
>> directives (which allow you to create a C structure with the same  
>> layout as an Objective-C object).
>
> The patch looks great, some minor issues:
>
> +++ lib/Parse/ParseDecl.cpp	(working copy)
>
> Please update the grammar to indicate the changes you're making to  
> the parser.

Done.

> @@ -736,18 +736,33 @@
> +    if(!Tok.is(tok::at)) {
>
> Space after the 'if' please.

Done.

> +      ConsumeToken();
>
> Not ok.  This consumes the token after the @ even though you don't  
> know what it is.  If it is <eof>, you're hosed :).  You need to  
> think of a recovery strategy here, what should happen if an "@" is  
> seen without defs after it?

Fixed

> +      if (!Tok.is(tok::l_paren)) Diag(Tok,  
> diag::err_expected_lparen);
> +      ConsumeParen();
>
> Likewise, and below.  Also, please use ExpectAndConsume where  
> possible.
>
> +      llvm::SmallVector<DeclTy*, 16> Fields;
> +      Actions.ActOnDefs(CurScope, Tok.getLocation(),  
> Tok.getIdentifierInfo(),
> +          Fields);
> +      FieldDecls.insert(FieldDecls.end(), Fields.begin(),  
> Fields.end());
> +      ConsumeToken();
> +      if (!Tok.is(tok::r_paren)) Diag(Tok,  
> diag::err_expected_rparen);
> +      ConsumeParen();
> +    }

Done.  Lots of errors happen now with nonsense input.

> +++ lib/Sema/SemaDecl.cpp	(working copy)
>
> +static void CollectIvars(ObjCInterfaceDecl *Class,
> +    llvm::SmallVector<Sema::DeclTy*, 16> &ivars) {
> +  if(Class->getSuperClass()) {
> +    CollectIvars(Class->getSuperClass(), ivars);
> +  }
> +  for (ObjCInterfaceDecl::ivar_iterator iter=Class->ivar_begin() ;
> +      iter!=Class->ivar_end() ; iter++) {
> +    ivars.push_back(*iter);
> +  }
>
> Can this be: ivars.append(Class->ivar_begin(), Class->ivar_end())?

Yes - it was a loop so I could dump the ivar names as they were found  
and check they were in the right order.

> +}
> +
> +void Sema::ActOnDefs(Scope *S, SourceLocation DeclStart,  
> IdentifierInfo
> +      *ClassName, llvm::SmallVector<DeclTy*, 16> &Decls) {
> +  ObjCInterfaceDecl *Class = getObjCInterfaceDecl(ClassName);
> +  if (!Class) Diag(DeclStart, diag::err_undef_interface, ClassName- 
> >getName());
> +  CollectIvars(Class, Decls);
> +}
> +
>
> comments please!

Added.

> +++ lib/CodeGen/CodeGenTypes.cpp	(working copy)
> @@ -149,7 +149,7 @@
>
> What is this patch doing?  It seems unrelated.

Hmm, in fact it might be.  Ignore it for now (or commit it  
separately).  It's just a few cleanups of the CodeGenTypes stuff.

defs.m is the file I am using to test this.  It outputs two lines in  
the form %d: %d.  If both numbers on each line match then it worked.   
Compile with ccc defs.m -lobjc.  Advice on how to turn this into an  
automated clang test welcome...

David
-------------- next part --------------
A non-text attachment was scrubbed...
Name: defs.diff
Type: application/octet-stream
Size: 6361 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20080621/ed6b5312/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: defs.m
Type: application/octet-stream
Size: 433 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20080621/ed6b5312/attachment-0001.obj>