[Bug c++/96765] New: Base class constructor cast to derived should cause a warning

[Bug c++/96765] New: Base class constructor cast to derived should cause a warning

[jzwinck at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96765

            Bug ID: 96765
           Summary: Base class constructor cast to derived should cause a
                    warning
           Product: gcc
           Version: 10.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jzwinck at gmail dot com
  Target Milestone: ---

If I cast "this" in a base class constructor to a derived class type, there is
no warning even with -Wall -Wextra.  Such a cast is undefined behavior, and
seems like it should be diagnosed at compile time.

For example:

    struct Base
    {
      Base();
      int num;
    };

    struct Derived : Base
    {
      int calc() const { return 42; }
    };

    Base::Base()
      // UB: Derived not yet constructed
      : num(static_cast<Derived*>(this)->calc())
    {
    }

    int main()
    {
      Derived d;
      return d.num;
    }

It compiles cleanly with "-Wall -Wextra -Werror" in GCC 8 and 10, with and
without optimization.  I think it should produce a diagnostic such as "invalid
static_cast from type ‘Base*’ to type ‘Derived*’ before the latter is
constructed".

UBSan reports the undefined behavior if Base has a virtual method (e.g. if you
add "virtual ~Base() = default;"), but not with the code as written, making it
even more advantageous to diagnose at compile time.

Live demo: https://godbolt.org/z/x3fa4Y

[Bug c++/96765] Base class constructor cast to derived should cause a warning

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96765

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2020-08-24
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
I agree a diagnostic makes sense.

[Bug c++/96765] Base class constructor cast to derived should cause a warning

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96765

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement
   Last reconfirmed|2020-08-24 00:00:00         |2021-12-16

[Bug c++/96765] Base class constructor cast to derived should cause a warning

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96765

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu.org

--- Comment #2 from Jason Merrill <jason at gcc dot gnu.org> ---
Hmm, is the cast undefined, actually?  I thought you were right, but now can't
find any wording to that effect.  I see

[expr.static.cast] If the prvalue of type “pointer to cv1 B” points to a B that
is actually a base class subobject of an object of type D, the resulting
pointer points to the enclosing object of type D. Otherwise, the behavior is
undefined.

But this B is a base of a D.  Then

[class.cdtor] For an object with a non-trivial constructor, referring to any
non-static member or base class of the object before the constructor begins
execution results in undefined behavior.

but D() has begun execution.  Then

[class.cdtor] During the construction of an object, if the value of the object
or any of its subobjects is accessed through a glvalue that is not obtained,
directly or indirectly, from the constructor’s this pointer, the value of the
object or subobject thus obtained is unspecified.

but 'this' in the Base ctor seems to count as obtained indirectly from 'this'
in the Derived ctor.

Then [class.cdtor] paragraph 4 talks about virtual functions, which this is
not.

What rule makes this undefined?