http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51295
Bug #: 51295
Summary: [C++11][noexcept] Wrong c'tor exception-specification
with non-trivial d'tor
Classification: Unclassified
Product: gcc
Version: 4.7.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
AssignedTo: [email protected]
ReportedBy: [email protected]
CC: [email protected]
I became aware of the following problem, when attempting to create something
like boost::optional with all the C++11 bells and whistles.
gcc 4.7.0 20111119 (experimental) in C++11 mode with -Wall rejects the
following code:
//-------
struct B {
~B() {}
};
static_assert(noexcept(B{}), "Error"); // Line 5
struct B2 {
B2() noexcept {}
~B2() {}
};
static_assert(noexcept(B2{}), "Error"); // Line 12
struct D : B2 {
D() noexcept = default; // Line 15
};
//-------
"
main.cpp|5|error: static assertion failed: "Error"|
main.cpp|12|error: static assertion failed: "Error"|
main.cpp|15|error: function 'D::D()' defaulted on its first declaration with an
exception-specification that differs from the implicit declaration 'D::D()'
"
It seems that in some situations the compiler-deduced exception-specification
of special members gives incorrect values not following FDIS rules. The root of
the problem seems to be located when there exist user-provided destructors
without exception-specification, even though these should behave as if they
where declared as noexcept(true). In addition to that such destructors can
influence the deduced exception-specification of constructors, as shown for
type D.
A current workaround is to mark the destructors with noexcept as well. None the
less this is a rather huge problem, because due to the popular existence of
destructors without exception-specification this can lead to large number of
failures in test cases.