[cfe-commits] [Patch] Improve diagnostics on gnu attributes

Ted Kremenek kremenek at apple.com
Wed Apr 13 21:03:11 PDT 2011

Hi Michael,

From looking at ParseDecl.cpp, ParseGNUAttributes, I think the difference between the parameter and the arguments is that the parameter is an identifier and the arguments are expressions.  Essentially, the parameter is the first argument that is guaranteed to be an identifier.

Instead of doing:

+  if (Attr.getParameterName() || Attr.getNumArgs() != 0) {

how about something like:

  if (Attr.hasParameterOrArguments())

which would just an inline method in Attribute that expands to the logic you wrote.  Right now you're basically copy-pasting a "complex" predicate in multiple places, and that is the perfect use of an inline method that summarizes your intent and makes the client code simpler.  


On Apr 10, 2011, at 9:17 PM, Michael Han wrote:

> Consider this code:
> int a;
> __attribute__((noreturn(a))) void foo();
> Clang does not emit any warning or errors on this code; gcc would report wrong number of arguments for the attribute specified.
> The attached patch fixed this by emit a “takes no argument” error for clang.
> Alternatively parser can be updated so the first identifier immediately following the attribute name could be treated as argument, instead of “parameter” (what is the difference between them?). This would require a lot more work to update the parser and the AttributeList so I am not sure if that is the right thing to do for now. Please review the patch thanks
> Cheers
> ~Michael
> <attributes.fix.patch>_______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110413/6ab36eba/attachment.html>

More information about the cfe-commits mailing list