Nullptr dereference false positive with variadic function?

[Mark Bourgeault]

I'm on cppcheck 2.3. I apologize in advance if this is already fixed...it's not trivial for me to upgrade my environment.

Test program, a.cpp:

#include <cstdlib>

template <typename... Args>
void fatalError(int code, const char* fmt, Args&&... args) { exit(code); }

int main(int argc, char* argv[]) {
    int* ptr = argc == 2 ? new int : nullptr;
    if (!ptr) {
        fatalError(1, "%s", argv[0]);
        /* fatalError(1, "%d", argc); // uncomment line to bypass false +ve */
    }
    *ptr;
    return 0;
}

cppcheck --enable=warning a.cpp results in a nullPointerRedundantCheck warning. Uncommenting the line results in no warnings.

At this point, I'm pretty confident that it's not my error. Thanks in advance for any help.

[Daniel Marjamäki]

I can reproduce with cppcheck-2.3 and cppcheck-2.4. but it's fixed in cppcheck-2.5

I'm on cppcheck 2.3. I apologize in advance if this is already fixed...it's not trivial for me to upgrade my environment.

no problem, thanks for reporting.

[Mark Bourgeault]

There is a variation involving namespaces that results in a false positive on 2.5.

#include <cstdlib>

namespace ns {
template <typename... Args>
void fatalError(int code, const char* fmt, Args&&... args) { exit(code); }
}

int main(int argc, char* argv[]) {
int* ptr = argc == 2 ? new int : nullptr;
if (!ptr) {
    using namespace ns;
    fatalError(1, "%s", argv[0]);
    /* ns::fatalError(1, "%s", argv[0]); // uncomment line to bypass false +ve */
}
*ptr;
return 0;
}