[Bug c++/69261] [6 Regression] Copying char arrays during constexpr evaluation does not work reliably
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69261
--- Comment #4 from Jens Auer ---
It produces the correct results if you change foo to not use constexpr:
void foo()
{
auto const s1 = s( "bla" );
auto const s2 = s( "blu" );
string_constexpr<7> const s1s2 = concat(s1,s2);
auto constexpr c = concat("bla", "blu");
std::cout << s1.data << std::endl << s2.data << std::endl << s1s2.data <<
std::endl << c << std::endl;
}
$ ./a.out
bla
blu
blablu
blablu
[Bug c++/69261] New: Copying char arrays during constexpr evaluation does not work reliably
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69261
Bug ID: 69261
Summary: Copying char arrays during constexpr evaluation does
not work reliably
Product: gcc
Version: 5.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: jens.auer at cgi dot com
Target Milestone: ---
Created attachment 37331
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=37331=edit
Preprocessed source code
I was playing around with string processing for constexpr values, and wrote the
following program to concatenate two strings at compile time:
#include
#include
#include
using std::size_t;
template
struct string_constexpr
{
constexpr string_constexpr() = default;
template
constexpr string_constexpr( char const ()[M] ):
data{0}
{
static_assert( M <= N, "size!" );
for(size_t i=0; i != M; i++)
{
data[i] = d[i];
}
}
char data[N];
};
template
T& operator<<(T& stream, string_constexpr const& str)
{
return (stream << str.data);
}
template
constexpr string_constexpr s( char const ()[N] )
{
string_constexpr c{};
for(size_t i=0; i != N; i++)
{
c.data[i] = d[i];
}
return c;
}
template
constexpr auto concat(string_constexpr const& s1, string_constexpr const&
s2)
{
string_constexpr<N+M-1> s( s1.data );
for(size_t i=0; i != M; i++)
{
s.data[N+i-1] = s2.data[i];
}
return s;
}
template
constexpr auto concat(char const ()[N], char const ()[M])
{
string_constexpr<N+M-1> tmp{x};
for(size_t i=0; i != M; i++)
{
tmp.data[N+i-1] = y[i];
}
return tmp;
}
void foo()
{
auto constexpr s1 = s( "bla" );
auto constexpr s2 = s( "blub" );
string_constexpr<8> constexpr s1s2 = concat(s1,s2);
auto constexpr c = concat("bla", "blub");
std::cout << s1.data << std::endl << s2.data << std::endl << s1s2.data <<
std::endl << c << std::endl;
}
int main()
{
foo();
return 0;
}
Executables compiled with GCC and clang produce different outputs, and I think
clang is right:
$ g++ -std=c++1y static_strings.cpp
$ ./a.out
bla
blub
blablub
$ clang++ -std=c++1y static_strings.cpp
$ ./a.out
bla
blub
blablub
blablub
[Bug c++/69263] New: internal compiler error: in cxx_eval_store_expression
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69263
Bug ID: 69263
Summary: internal compiler error: in cxx_eval_store_expression
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: jens.auer at cgi dot com
Target Milestone: ---
Compilation of the following program with gcc 6.0.0 201601 (from the online
compiler http://melpon.org/wandbox,
http://melpon.org/wandbox/permlink/CMreVotFuJMgnuI7) crashes:
#include
#include
#include
using std::size_t;
template
struct string_constexpr
{
constexpr string_constexpr() = default;
template
constexpr string_constexpr( char const ()[M] ):
data{0}
{
static_assert( M <= N, "size!" );
for(size_t i=0; i != M; i++)
{
data[i] = d[i];
}
}
char data[N];
};
template
T& operator<<(T& stream, string_constexpr const& str)
{
return (stream << str.data);
}
template
constexpr string_constexpr s( char const ()[N] )
{
string_constexpr c{};
for(size_t i=0; i != N; i++)
{
c.data[i] = d[i];
}
return c;
}
template
constexpr auto concat(string_constexpr const& s1, string_constexpr const&
s2)
{
string_constexpr<N+M-1> s( s1.data );
for(size_t i=0; i != M; i++)
{
s.data[N+i-1] = s2.data[i];
}
return s;
}
template
constexpr auto concat(char const ()[N], char const ()[M])
{
static_assert(N == 4 && M == 4, "");
string_constexpr<N+M-1> tmp{x};
for(size_t i=0; i != M; i++)
{
tmp.data[N+i-1] = y[i];
}
return tmp;
}
void foo()
{
auto constexpr s1 = s( "bla" );
auto constexpr s2 = s( "blub" );
string_constexpr<8> constexpr s1s2 = concat(s1,s2);
auto constexpr c = concat("bla", "blub");
std::cout << s1.data << std::endl << s2.data << std::endl << s1s2.data <<
std::endl << c << std::endl;
}
int main()
{
foo();
return 0;
}
g++ prog.cc -Wall -Wextra -std=gnu++1z
prog.cc: In function 'void foo()':
prog.cc:74:33: in constexpr expansion of 's<4>("bla")'
prog.cc:74:33: internal compiler error: in cxx_eval_store_expression, at
cp/constexpr.c:2808
auto constexpr s1 = s( "bla" );
^
0x6f249e cxx_eval_store_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:2808
0x6f03bb cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3312
0x6efc5f cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3570
0x6f0267 cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3349
0x6f0267 cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3349
0x6f2b21 cxx_eval_statement_list
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3028
0x6f0563 cxx_eval_loop_expr
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3055
0x6f0563 cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3680
0x6f2b21 cxx_eval_statement_list
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3028
0x6f044c cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3614
0x6efd2f cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3669
0x6f2b21 cxx_eval_statement_list
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3028
0x6f044c cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3614
0x6efd2f cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3669
0x6ef1df cxx_eval_call_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:1386
0x6efd0b cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3238
0x6f0c7a cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3296
0x6efedb cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3378
0x6efc5f cxx_eval_constant_expression
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3570
0x6f2cf8 cxx_eval_outermost_constant_expr
/home/heads/gcc/gcc-source/gcc/cp/constexpr.c:3773
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.
1
Finish
[Bug c++/69263] internal compiler error: in cxx_eval_store_expression
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69263 --- Comment #1 from Jens Auer --- The source code is same as in bug #69261. On gcc 5.2, it compiles but produces a different output than a program compiled with clang. I have opened two separate bug reports because the effects are different and it affects two different versions, but feel free to merge them if it makes sense.
[Bug c++/69261] Copying char arrays during constexpr evaluation does not work reliably
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69261 --- Comment #2 from Jens Auer --- I have submitted a second bug report for the crash: Bug 69263. I don't know if that was appropriate, but the bug has different effects on different versions.
[Bug c++/67951] New: Wshadow for type inferenced (auto) lambda parameters
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67951
Bug ID: 67951
Summary: Wshadow for type inferenced (auto) lambda parameters
Product: gcc
Version: 5.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: jens.auer at cgi dot com
Target Milestone: ---
int f()
{
auto l = [](int i) {};
int i=0;
l(i);
return i;
}
void h()
{
auto l = [](auto i) {};
int i=0;
l(i);
}
If compiled with "g++ -std=c++1y -Wshadow", I get a warning in h:
shadow.cpp:15:23: warning: declaration of ‘i’ shadows a previous local
[-Wshadow]
auto l = [](auto i) {};
^
shadow.cpp:17:9: note: shadowed declaration is here
int i=0;
I think this is wrong. If it is not wrong, it is inconsistent because there is
no warning in f().
[Bug c++/67952] New: Compilation error with boost::signals2 and generic lambda: cannot call member function without object
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67952
Bug ID: 67952
Summary: Compilation error with boost::signals2 and generic
lambda: cannot call member function without object
Product: gcc
Version: 5.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: jens.auer at cgi dot com
Target Milestone: ---
Created attachment 36495
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36495=edit
Preprocessed source code
The following code doesn't compile with boost 1.59 and gcc 5.1, but with clang
3.6:
#include
class A
{
public:
A()
{
// these two work
sig2.connect( [this]() {f();} );
sig.connect( [this](auto x) {this->g(x);} );
// this does not work
sig.connect( [this](auto x) {g(x);} );
}
private:
void f();
void g(int);
boost::signals2::signal sig;
boost::signals2::signal sig2;
};
The error message is
shadow.cpp: In instantiation of ‘A::A()::<lambda(auto:2)> [with auto:2 = int]’:
/lhome1/auerj/opt/include/boost/function/function_template.hpp:153:11:
required from ‘static void
boost::detail::function::void_function_obj_invoker1<FunctionObj, R,
T0>::invoke(boost::detail::function::function_buffer&, T0) [with FunctionObj =
A::A()::<lambda(auto:2)>; R = void; T0 = int]’
/lhome1/auerj/opt/include/boost/function/function_template.hpp:934:38:
required from ‘void boost::function1<R, T1>::assign_to(Functor) [with Functor =
A::A()::<lambda(auto:2)>; R = void; T0 = int]’
/lhome1/auerj/opt/include/boost/function/function_template.hpp:722:7:
required from ‘boost::function1<R, T1>::function1(Functor, typename
boost::enable_if_c<(boost::type_traits::ice_not<(boost::is_integral::value)>::value),
int>::type) [with Functor = A::A()::<lambda(auto:2)>; R = void; T0 = int;
typename
boost::enable_if_c<(boost::type_traits::ice_not<(boost::is_integral::value)>::value),
int>::type = int]’
/lhome1/auerj/opt/include/boost/function/function_template.hpp:1071:16:
required from ‘boost::function<R(T0)>::function(Functor, typename
boost::enable_if_c<(boost::type_traits::ice_not<(boost::is_integral::value)>::value),
int>::type) [with Functor = A::A()::<lambda(auto:2)>; R = void; T0 = int;
typename
boost::enable_if_c<(boost::type_traits::ice_not<(boost::is_integral::value)>::value),
int>::type = int]’
/lhome1/auerj/opt/include/boost/function/function_template.hpp:1126:5:
required from ‘typename
boost::enable_if_c<(boost::type_traits::ice_not<(boost::is_integral::value)>::value),
boost::function<R(T0)>&>::type boost::function<R(T0)>::operator=(Functor) [with
Functor = A::A()::<lambda(auto:2)>; R = void; T0 = int; typename
boost::enable_if_c<(boost::type_traits::ice_not<(boost::is_integral::value)>::value),
boost::function<R(T0)>&>::type = boost::function<void(int)>&]’
/lhome1/auerj/opt/include/boost/signals2/detail/slot_template.hpp:160:24:
required from ‘void boost::signals2::slot<R(Args ...),
SlotFunction>::init_slot_function(const F&) [with F = A::A()::<lambda(auto:2)>;
SlotFunction = boost::function<void(int)>; R = void; Args = {int}]’
/lhome1/auerj/opt/include/boost/signals2/detail/slot_template.hpp:85:27:
required from ‘boost::signals2::slot<R(Args ...), SlotFunction>::slot(const F&)
[with F = A::A()::<lambda(auto:2)>; SlotFunction = boost::function<void(int)>;
R = void; Args = {int}]’
shadow.cpp:13:45: required from here
shadow.cpp:13:38: error: cannot call member function ‘void A::g(int)’ without
object
sig.connect( [this](auto x) {g(x);} );
I think it is a compiler bug because adding the implicit this-> solves the
problem (see second connect).
My gcc version:
g++ -v
Using built-in specs.
COLLECT_GCC=g++
COLLECT_LTO_WRAPPER=/lhome1/auerj/opt/libexec/gcc/x86_64-unknown-linux-gnu/5.1.0/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-5.1.0/configure --prefix=/lhome1/auerj/opt
--enable-languages=c,c++
Thread model: posix
gcc version 5.1.0 (GCC)
I have attached the preprocessed output so it should be easy to reproduce the
issue.
It is probably a duplicate of 61636, but I such that it can be tracked as such.
I guess boost::signals2 is a popular library, so it may be worth to have it in
the database explicitly.
[Bug middle-end/60725] [-Wreturn-type] false positive in trivial switch
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60725 Jens Auer changed: What|Removed |Added CC||jens.auer at cgi dot com --- Comment #2 from Jens Auer --- This warnings becomes annoying when you also compile with -Wswitch-enum which forces you to cover all enum values. When they are all covered, I do not need a default statement, but to prevent Wreturn-type, I have to add a pseudo-default.
