[Bug c++/110075] New: Bogus -Wdangling-reference

[Bug c++/110075] New: Bogus -Wdangling-reference

[pilarlatiesa at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

            Bug ID: 110075
           Summary: Bogus -Wdangling-reference
           Product: gcc
           Version: 13.1.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pilarlatiesa at gmail dot com
  Target Milestone: ---

$ cat test.cpp
#include <map>
#include <string>

class T
{
  static inline std::map<std::string, int> const m = {{"foo", 0}};

public:
  static int const &Get(std::string const &s) { return m.at(s); }
};

int main()
{
  [[maybe_unused]] int const &i = T::Get("foo");
}


$ ./gcc-13/bin/g++ -Wall test.cpp
test.cpp: In function ‘int main()’:
test.cpp:14:31: warning: possibly dangling reference to a temporary
[-Wdangling-reference]
   14 |   [[maybe_unused]] int const &i = T::Get("foo");
      |                               ^
test.cpp:14:41: note: the temporary was destroyed at the end of the full
expression ‘T::Get(std::__cxx11::basic_string<char>(((const char*)"foo"),
std::allocator<char>()))’
   14 |   [[maybe_unused]] int const &i = T::Get("foo");
      |                                   ~~~~~~^~~~~~~

[Bug c++/110075] Bogus -Wdangling-reference

[pilarlatiesa at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

--- Comment #1 from Pilar Latiesa <pilarlatiesa at gmail dot com> ---
I forgot to paste:

$ ./gcc-13/bin/g++ -v
Usando especificaciones internas.
COLLECT_GCC=./gcc-13/bin/g++
COLLECT_LTO_WRAPPER=/home/pililatiesa/gcc-13/libexec/gcc/x86_64-pc-linux-gnu/13.1.0/lto-wrapper
Objetivo: x86_64-pc-linux-gnu
Configurado con: ../gcc-13.1.0/configure --enable-languages=c,c++
--disable-multilib --prefix=/home/pililatiesa/gcc-13/
Modelo de hilos: posix
Algoritmos de compresión LTO admitidos: zlib
gcc versión 13.1.0 (GCC)

[Bug c++/110075] Bogus -Wdangling-reference

[roystgnr at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

Roy Stogner <roystgnr at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |roystgnr at gmail dot com

--- Comment #2 from Roy Stogner <roystgnr at gmail dot com> ---
This case looks most similar to the one that my code triggered, which I
simplified to the following:

---

#include <map>

const double &
find_wrapper(const std::map<int, double> & map,
             const int & key)
{
  auto it = map.find(key);
  return it->second;
}


int main(void)
{
  std::map<int, double> testmap{{1,0.0}};
  const double & d = find_wrapper(testmap, 1);
  return int(d);
}

---

$ g++ -Wall -Wextra -Werror -o test.x test.C
test.C: In function ‘int main()’:
test.C:15:18: error: possibly dangling reference to a temporary
[-Werror=dangling-reference]
   15 |   const double & d = find_wrapper(testmap, 1);
      |                  ^
test.C:15:34: note: the temporary was destroyed at the end of the full
expression ‘find_wrapper(testmap, 1)’
   15 |   const double & d = find_wrapper(testmap, 1);
      |                      ~~~~~~~~~~~~^~~~~~~~~~~~
cc1plus: all warnings being treated as errors

---

If I change the parameter type to `const int key`, then the warning is not
triggered.  If I leave the parameter type as a reference but I pass in a
non-temporary variable "one" instead of "1", then the warning is not triggered.


Either or both of these cases might be a dup of bug 109642

[Bug c++/110075] Bogus -Wdangling-reference

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
                 CC|                            |mpolacek at gcc dot gnu.org
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-01-19

--- Comment #3 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Confirmed, thanks for reporting.  I'm not certain I can devise a fix, but maybe
a new attribute to suppress the warning would help.  Like in bug 109642 as
mentioned.

[Bug c++/110075] Bogus -Wdangling-reference

[ostash at ostash dot kiev.ua]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

Viktor Ostashevskyi <ostash at ostash dot kiev.ua> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ostash at ostash dot kiev.ua

--- Comment #4 from Viktor Ostashevskyi <ostash at ostash dot kiev.ua> ---
I have even simpler reproducer:

---
class X{};
const X x1;
const X x2;

const X& get(const int& i)
{
   return i == 0 ? x1 : x2;
}

void foo() {
    [[maybe_unused]] const X& x = get(10);
}
---

<source>: In function 'void foo()':
<source>:23:31: warning: possibly dangling reference to a temporary
[-Wdangling-reference]
   23 |     [[maybe_unused]] const X& x = get(10);
      |                               ^
<source>:23:38: note: the temporary was destroyed at the end of the full
expression 'get(10)'
   23 |     [[maybe_unused]] const X& x = get(10);
      |

[Bug c++/110075] Bogus -Wdangling-reference

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

--- Comment #5 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
Yes, because we'd have to analyze the body of the function to see that it does
not return one of the parameters, which often we can't do.

There will be an attribute soon to suppress that warning.

[Bug c++/110075] Bogus -Wdangling-reference

[ostash at ostash dot kiev.ua]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

--- Comment #6 from Viktor Ostashevskyi <ostash at ostash dot kiev.ua> ---
(In reply to Marek Polacek from comment #5)
> Yes, because we'd have to analyze the body of the function to see that it
> does not return one of the parameters, which often we can't do.
> 
> There will be an attribute soon to suppress that warning.

I'm wondering whether this diagnostics should be in -Wall or -Wextra at all.
It is assumed that diagnostics in -Wall or -Wextra generally "works".

Here, diagnostics clearly fails on simplest case and I don't think marking huge
codebase with some attribute is a proper solution...

[Bug c++/110075] Bogus -Wdangling-reference

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

--- Comment #7 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
It's hard to decide.  It seems that once we've covered std::span-like classes,
in practice the warning points out real issues.  I would hope that code like
you posted is actually fairly rare.  But it's difficult to be sure.

[Bug c++/110075] Bogus -Wdangling-reference

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110075

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #8 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
In r14-9263-gc7607c4cf18986 I added [[gnu::no_dangling]] which can be used to
suppress the warning in cases where the compiler can't detect that the
reference doesn't actually dangle.

Hopefully that'll work.