MS VC++ 6.0 object memory layout

[ankit.ankit]

Referring to the following links.
1.
http://groups-beta.google.com/group/...6539a90194bfbc
2.
http://msdn.microsoft.com/archive/de...angrayhood.asp

Q1. Can somebody tell what does the phrase
"non-virtually-inherited parts" as referred in John Gray's post at the
end of the page referred by 1st link

Also, for class M: public X, public Y
where X and Y both have virtual functions (and thus have vfptrs),
class M will contain two vfptrs (one each for embedded instances of X
and Y).

Q2. Each class has only one vftable asociated with it. So, do the two
vfptrs (as above) point to the same vtable? The rule of one vtable is
violated when doing virtual inheritance of a base class with virtual
functions and derived class adding new virtual functions.

Q3. Also, the vtable for M will have two entries for the same func().
When doing a lookup for func(), which of the two entries is refrred to.

[msalters]

an*********@gmail.com schreef:

Also, for class M: public X, public Y
where X and Y both have virtual functions (and thus have vfptrs),
class M will contain two vfptrs (one each for embedded instances of X
and Y).
Probably three, as M might add virtual functions of its own.
Q2. Each class has only one vftable asociated with it. So, do the two
vfptrs (as above) point to the same vtable? The rule of one vtable is
violated when doing virtual inheritance of a base class with virtual
functions and derived class adding new virtual functions.
What rule? It probably wouldn't work, either. X and Y are unrelated,
and have different virtual functions. I.e. the first virtual function
will differ, and __vtable[0] will then be different.
Q3. Also, the vtable for M will have two entries for the same func().
When doing a lookup for func(), which of the two entries is refrred to.

Name lookup doesn't happen at runtime in vtables, but at compile time.

HTH,
Mchiel Salters