[LLVMbugs] [Bug 14531] New: Clang asserts when calling-convention attributes are used on function pointers in C++ C-tors

bugzilla-daemon at llvm.org bugzilla-daemon at llvm.org
Thu Dec 6 14:48:05 PST 2012


             Bug #: 14531
           Summary: Clang asserts when calling-convention attributes are
                    used on function pointers in C++ C-tors
           Product: clang
           Version: 3.1
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Frontend
        AssignedTo: unassignedclangbugs at nondot.org
        ReportedBy: benjamin.j.panning at intel.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

Consider the following code:

typedef __attribute__((stdcall)) void StdcallFnTy();
typedef __attribute__((fastcall)) void FastcallFnTy();

class Foo
  Foo(StdcallFnTy* f) { }
  Foo(FastcallFnTy* f) { }
// *** END CODE

This produces the following assertion when Clang attempts to merge c-tor

Assertion failed: Entry->isDeclaration() && "definition already exists for
alias", file \llvm\tools\clang\lib\CodeGen\CGCXX.cpp, line 161
clang: error: clang frontend command failed with exit code 3 (use -v to see

Interestingly, if the calling-convention attributes are replaced by other
non-calling convention attributes, Clang reports an error:

  "constructor cannot be redeclared"

It appears that the front end parses the calling-convention attribute and
differentiates between the types StdcallFnTy and FastcallFnTy, but does not
correctly differentiate between the types when creating c-tor aliases.  I'm
assuming this is because Clang is grabbing the global value by mangled name and
the mangled names are the same for both c-tors.

I'm not sure what the C/C++ spec calls for on this, but I think it would make
sense to do one of the following:

  1. Report the same "constructor cannot be redeclared" error for c-tors whose
type differ only by calling convention.
  2. Handle the calling convention when creating c-tor aliases so that the
above code works.

Either one of these would make me happy.  Asserts make me sad. :(

Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

More information about the llvm-bugs mailing list