https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68727
Bug ID: 68727
Summary: [4.9 regression] invalid offsetof expressions accepted
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: msebor at gcc dot gnu.org
Target Milestone: ---
While coming up with a comprehensive test case for c++/68711 I noticed that GCC
silently accepts other invalid offsetof expressions that then cause the
ill-formed program to crash at runtime (below). Since this problem is not
caused by the merge of the C++ delayed folding branch or a regression from GCC
5.1, I raise it a bug of its own. While strictly speaking, GCC's behavior
isn't non-conforming because the expressions have undefined behavior, since GCC
otherwise makes an effort to detect and diagnose such constructs and in the
past did diagnose these, this is a defect in the implementation.
$ /build/gcc-trunk-svn/gcc/xgcc -B /build/gcc-trunk-svn/gcc -Wall -Wextra
-Wpedantic x.cpp
tmp$
tmp$ cat x.cpp && /build/gcc-trunk-svn/gcc/xgcc -B /build/gcc-trunk-svn/gcc
-Wall -Wextra -Wpedantic x.cpp && ./a.out
struct A { int i; };
struct B: virtual A { };
__SIZE_TYPE__ a[] = {
!&((B*)0)->A::i,
__builtin_offsetof (B, A::i)
};
int main ()
{
return a[0] + a[1];
}
Segmentation fault (core dumped)
$
The absence of diagnostics on these expressions is a regression from 4.9.3
which issues the following warnings. (It also successfully compiles the
program that then crashes at runtime).
x.cpp:5:19: warning: invalid access to non-static data member ‘A::i’ of NULL
object [-Winvalid-offsetof]
!&((B*)0)->A::i,
^
x.cpp:5:19: warning: (perhaps the ‘offsetof’ macro was used incorrectly)
[-Winvalid-offsetof]
x.cpp:6:31: warning: invalid access to non-static data member ‘A::i’ of NULL
object [-Winvalid-offsetof]
__builtin_offsetof (B, A::i)
^
x.cpp:6:31: warning: (perhaps the ‘offsetof’ macro was used incorrectly)
[-Winvalid-offsetof]
