[Bug libstdc++/81078] dynamic_cast to virtual base produces the wrong answer

[arthur.j.odwyer at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81078

--- Comment #3 from Arthur O'Dwyer  ---
Yes, this is a libstdc++ issue.

I'm not 100% sure that "the RTTI [generated by GCC] is correct," because I
don't know how to use GCC with libc++; but yeah, there's definitely at least
some problem with libstdc++ here. Observe that Clang-with-libstdc++ fails,
Clang-with-libc++ succeeds: https://godbolt.org/z/PMq6rW9cE

[Bug libstdc++/81078] dynamic_cast to virtual base produces the wrong answer

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81078

Andrew Pinski  changed:

   What|Removed |Added

   Last reconfirmed||2021-08-18
 Status|UNCONFIRMED |NEW
 Ever confirmed|0   |1

--- Comment #2 from Andrew Pinski  ---
(In reply to Arthur O'Dwyer from comment #0)

you need to use libc++ to get the answer.
This means the RTTI is correct and it is libstdc++ issue?

[Bug libstdc++/81078] dynamic_cast to virtual base produces the wrong answer

[arthur.j.odwyer at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81078

--- Comment #1 from Arthur O'Dwyer  ---
I've found that the misbehavior actually depends on the *ordering* of bases in
the class graph, not just the shape of the graph.

// https://wandbox.org/permlink/DaKQxTc5Ldqj9FlY

#include 

struct Sideboard { virtual ~Sideboard() {} };
struct Shared { virtual ~Shared() {} };
struct Public : public virtual Shared {};
struct Protected : protected virtual Shared {};

template struct Main : Ts... {};

template
void test() {
Main m;
Sideboard *sb = 
Shared *sh = 
if (dynamic_cast(sb) != sh) {
puts(__PRETTY_FUNCTION__);
}
}

int main() {
test();
test();
test();
test();
test();
test();
}


OUTPUT is:
void test() [with Ts = {Public, Protected, Sideboard}]

That is, the failure case seems to be when (and only when):
- the base being casted "from" is laid out after any base descended from "to",
AND
- the last base (in lexical order) descended from "to" is not publicly
descended from "to".