[Bug tree-optimization/84480] New: Unexpected -Wstringop-truncation on strncpy with string literal at -O2

["romain.geissler at amadeus dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84480

            Bug ID: 84480
           Summary: Unexpected -Wstringop-truncation on strncpy with
                    string literal at -O2
           Product: gcc
           Version: 8.0.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: romain.geissler at amadeus dot com
  Target Milestone: ---

Hi,

As required in PR 84474, here is a more real life example of unexpected
-Wstringop-truncation. Let's consider a special string implementation that
would use the fact that the string length can be bounded to a given limit. This
would be a templated class (so visible to other translation units, and thus
optimizable through inlining and constant propagation).

<<EOF
#include <string.h>

template <size_t N> class String 
{
    public:
        String(const char* iString, size_t iStringLength = N)
        { 
            if (iStringLength > N)
            {
                iStringLength = N;
            }

            strncpy(_cstring, iString, iStringLength);
            _cstring[iStringLength] = '\0'; 
        }

    private:
        char _cstring[N + 1]; 
};

void f(const char* iUserProvidedString)
{
    String<3> aString1("12"); // No warning
    String<3> aString2("123"); // Warning
    String<3> aString3(iUserProvidedString); // Warning here if g is
uncommented.
}

void g()
{
    f("456");
}

compiled with: -Wall -Werror -Wextra -std=gnu++17 -O2

I get: error: ‘char* strncpy(char*, const char*, size_t)’ output truncated
before terminating nul copying 3 bytes from a string of the same length
[-Werror=stringop-truncation]
for string "123" and "456".

I don't know if you can do anything in that case. Actually I still haven't
understood what this particular warning would prevent me from doing. Does gcc
fears that I don't null terminate _cstring ? Shouldn't the fact that the next
statement actually null terminates _cstring be enough to silence the warning ?

Cheers,
Romain