No longer catching memcpy on class in some cases?

[Steve Hopkins]

Just upgraded from cppcheck v1.33 to v1.78. The code below was formerly flagged asUsing 'memcpy' on class, but is ignored with v1.78.
test26.h:
class foo_struct;
namespace this_stuff
{
class FUBAR_struct
{...}
...
}
test26.cpp:
bool FUBAR_struct::getnext(foo_struct inp)
{
memcpy(inp, some_const, sizeof(foo_struct));

[Daniel Marjamäki]

we still have the warning as far as I know.

not sure.. we might have some new heuristic. maybe the class must have non-POD data or something.

[Daniel Marjamäki]

Could you provide a better example code? When I test this code then latest cppcheck writes a warning:

class Fred {
  std::string data;
};

void f(Fred *fred1, Fred *fred2) {
  memcpy(fred1, fred2, sizeof(Fred));
}

[Steve Hopkins]

OK. I reduced my offending file down to this, and I still get a 'memcpy' on class error with v1.33 but not with v1.78.
test26.h:
class ItemStruct;
class CurrentStruct
{
bool getNext(ItemStruct *val);
};

test26.cpp:
#include "test26.h"
bool CurrentStruct::getNext(ItemStruct val)
{
ItemStruct bi = &(*next);
memcpy(val, bi, sizeof(ItemStruct));
return true;
}

[Paul Fultz]

CurrentStruct is still considered a trivial(or POD-type) type and it can be default-constructed and copied. The following program compiles for me without an error:

#include <type_traits>

class ItemStruct;
class CurrentStruct
{
    bool getNext(ItemStruct *val);
};

int main() {
    static_assert(std::is_trivial<CurrentStruct>{}, "Its not trivial");
    static_assert(std::is_pod<CurrentStruct>{}, "Its not pod");
    // Default construct the object
    auto s1 = CurrentStruct{};
    // Copy the object
    auto s2 = s1;
}

So calling memcpy and memset on CurrentStruct is valid. Although, we could provide another diagnostic here. It could be suspicious to call memcpy or memset on an empty object. Within a generic template it may not be suspicious at all, so we need to be carefult to avoid FPs.