]> granicus.if.org Git - clang/commit
Revert the fix for PR8013.
authorDouglas Gregor <dgregor@apple.com>
Tue, 9 Nov 2010 20:03:54 +0000 (20:03 +0000)
committerDouglas Gregor <dgregor@apple.com>
Tue, 9 Nov 2010 20:03:54 +0000 (20:03 +0000)
commitef9b1497908ab1684e04f244289f1ebe8d44274e
tree1a81e2b97164f47d56727be0043fb602aaabb2ad
parentedfb7ecea0222be2796469f8c41b836129f833f8
Revert the fix for PR8013.

That bug concerned the well-formedness of code such as (&ovl)(a, b,
c). GCC rejects the code, while EDG accepts it. On further study of the
standard, I see no support for EDG's position: in particular, C++
[over.over] does not list this as a context where we can take the
address of an overloaded function, C++ [over.call.func] does not
reference the address-of operator at any point, and C++ [expr.call]
claims that the function argument in a call is either a function
lvalue or a pointer-to-function; (&ovl) is neither.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@118620 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Sema/SemaCodeComplete.cpp
lib/Sema/SemaExpr.cpp
test/CodeCompletion/call.cpp
test/SemaCXX/overload-call.cpp