Re: [RFC] bloody mess with __attribute__() syntax

[Josh Triplett <josht@linux.vnet.ibm.com>]

On Thu, 2007-07-05 at 17:43 +0100, Al Viro wrote:
> On Thu, Jul 05, 2007 at 08:36:35AM -0700, Josh Triplett wrote:
> > Wow.  Insane.  So these all declare the same type:
> > __attribute__((foo)) T *v;
> > T __attribute__((foo)) *v;
> > T *__attribute__((foo)) v;
> > ?  Specifically, they point to a foo-T, for convenient shooting?
> 
> They all give you foo-pointer-to-T.  
> 	T (__attribute__((foo)) *v);
> would give pointer-to-foo-T.

Doesn't that do exactly what we want, then?  If we say
T __attribute__((noderef)) *v;
, we want a noderef-pointer-to-T, not a pointer-to-noderef-T.  noderef
should modify a pointer.

> > context also represents a qualifier; the position of the qualifier should
> > determine things like whether you want to enforce the context when you access
> > a pointer or dereference a pointer.
> 
> Since __context__ is (sparse-only) keyword, we are not constrained by
> anything anyway.

No, I mean __attribute__((context(...))), which means something
different.  __context__() works as a statement statement changing the
context.  __attribute__((context(...))) works as an attribute modifying
a type to say that it requires a given context, and that
accessing/calling it changes the context.  Somewhat of an odd
distinction, but sparse currently works that way.

> > > Frankly, I would rather add a new primitive (__qualifier__) mirroring the
> > > __attribute__, but acting like real qualifiers do.  And switched the
> > > noderef et.al. to it.
> > 
> > Something like that sounds vaguely reasonable.  It should allow the same set
> > of attributes, and just change what they apply to.  To use your example,
> > T __qualifier__((foo)) *v;
> > and
> > T (__attribute__((foo)) *v);
> > would mean the same thing.
> 
> Yup, except that it would not accept storage-class-like attributes (e.g.
> always_inline).

So, to clarify what you mean by storage-class-like: __qualifier__ would
accept all attributes that modify a *type* rather than a *declaration*.

> And yes, __qualifier__((context(...))) probably might
> be a replacement for __context__, to reduce the number of primitives.

No, see above.  __context__ works as a statement, not as a way to modify
a type.

That said, "calling this function requires this context or changes the
context" and "accessing this variable requires this context or changes
the context" may represent different types of attributes.  I don't know.

- Josh Triplett