The C++ ABI defines order of virtual function pointers in such way: "2.5.2 Virtual Table Components and Order The order of the virtual function pointers in a virtual table is the order of declaration of the corresponding member functions in the class. There is an entry for any virtual function declared in a class, whether it is a new function or overrides a base class function, unless it overrides a function from the primary base, and conversion between their return types does not require an adjustment." *** G++ 3.1.1 compiler violates given rules *** Let's consider the example. --- fail.cpp ---------------------- struct base { short b; virtual int foo() {} }; struct derived: virtual base { int d; virtual int foo() {} virtual int bar() {} }; ----------------------------------- Note that function declaration order is "foo, bar" and 'base' is not the primary base class for 'derived', since it's not nearly empty class. New we check the virtual table for 'derived' class using "g++-3.1.1 -c -fdump-class-hierarchy fail.cpp". The virtual table is: Vtable for derived derived::_ZTV7derived: 9 entries 0 8 4 0 8 &_ZTI7derived 12 derived::bar() 16 derived::foo() 20 0fffffff8 24 0fffffff8 28 &_ZTI7derived 32 derived::_ZTv0_n12_N7derived3fooEv() We see that virtual functions appear in the order "bar, foo" which is wrong. It's expected to have functions in the declaration order as spcified by the ABI. Release: 3.1.1 Environment: Red Hat Linux release 7.1 (Seawolf) Kernel 2.4.2-2smp on a 2-processor i686 gcc version 3.1.1 Configured with: ./configure --prefix=/usr/local --program-suffix=-3.1.1 --enable-threads Thread model: posix How-To-Repeat: 1. Compile given testcase g++-3.1.1 -c -fdump-class-hierarchy fail.cpp 2. Look for 'derived' virtual table in the produced fail.cpp.class file
State-Changed-From-To: open->analyzed State-Changed-Why: confirmed. an ABI bug
Responsible-Changed-From-To: unassigned->nathan Responsible-Changed-Why: patching
State-Changed-From-To: analyzed->closed State-Changed-Why: now fixed