[PATCH] D47630: [Sema] Allow creating types with multiple of the same addrspace.

[Bevin Hansson via Phabricator via cfe-commits]

ebevhan added inline comments.


================
Comment at: test/Sema/address_spaces.c:17
   int *_AS1 _AS2 *Z;  // expected-error {{multiple address spaces specified for type}}
+  int *_AS1 _AS1 *M;
 
----------------
Anastasia wrote:
> ebevhan wrote:
> > bader wrote:
> > > I think it might be valuable to give a warning or remark to user. 
> > > Using the same address space qualifier multiple times is not something OpenCL C developers are supposed to do.
> > > 
> > The test is obviously a bit contrived, but it could happen by mistake, or as a result of some typedef or macro combination. It also cannot go wrong, so there's no harm in it happening.
> > 
> > I see your point, though. A warning feels like a bit much, so I'm not sure what else to use. A note?
> Just checked for const qualifier we get a warning:
>   warning: duplicate 'const' declaration specifier [-Wduplicate-decl-specifier]
> 
> We could do the same... not sure if we could try to share the diagnostic as well.
I have a patch ready that adds a warning in the same warning group and uses that. I'm not sure about reusing the other one since address spaces don't have to be declaration specifiers.

The warning is 'multiple identical address spaces specified for type', similar to the error. Is that acceptable?


Repository:
  rC Clang

https://reviews.llvm.org/D47630