[Bug c++/12668] New: Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[Bug c++/12668] New: Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[rmerkert at alphatech dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668

           Summary: Strange warnings with -O1 -Wunreachable-code in
                    3.3.1,3.3.2,3.4
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rmerkert at alphatech dot com
                CC: gcc-bugs at gcc dot gnu dot org

The following code produces a strange warning when compiled with
-Wunrechable-code and optimization:

cat > test.c <<EOF
inline void foo (int x)
{
   if (x!=0)
     x = x + 1;
}

int main ()
{
  foo(0);
/* if (0) { foo(1); } */  
  return 0;
}

EOF

gcc -O1 test.c -Wunreachable-code

The error message I get with both 3.3.1, 3.3.2, 3.4-20031001 is this:
test.c: In function `main':
test.c:4: warning: will never be executed

The message is somewhat misleading. I think it should either say something like:
test.c: In function 'foo':
test.c:4 warning: will never be executed

or 

test.c: In function 'main':
test.8: warning: will never be executed


By the way, if compiled with g++, there will be an extra empty line the output:
test.c: In function `main':

test.c:4: warning: will never be executed

and if you uncomment line 10 you'll get 4 warnings with 3.4 and 3 warnings with
3.3 and they are not helpful (gcc-3.3 doesn't even tell me about line 10).

I'm actually not sure I want the warning in this case - I really want the
compiler to tell me that a section of code cannot possibly be reached such as
"while (false) {}"

[Bug c++/12668] Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |diagnostic
   Last reconfirmed|0000-00-00 00:00:00         |2003-10-18 04:25:22
               date|                            |


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-18 04:25 -------
-O1 enables inlining so that foo is inlined and the warning is right when x is always 0 because that 
line is never executed in the line version.
What is wrong is when you enable line 10, it says that line 2 is never executed twice. It should 
never tell you that line 2 is not executed at all (because there is no code assocated with it).

[Bug c++/12668] Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[rmerkert at alphatech dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668



------- Additional Comments From rmerkert at alphatech dot com  2003-10-18 13:24 -------
I still think that the warning is not very helpful. Consider this program instead:
cat > test.c <<EOF
inline void foo (int x)
{
   if (x!=0)
     x = x + 1;
}

int main ()
{
  foo(-1);
  foo(0);
  foo(1);
  foo(2);

  return 0;
}

EOF

gcc -O1 test.c -Wunreachable-code

This will give exactly the same warning as before:

main.c: In function `main':
main.c:4: warning: will never be executed

The warning is completely useless. The warning is not even pointing me to the
root cause which is the call to foo(0). 
I think that the warning should be something like this:
main.c: In function `main':
main.c:10: warning: call to foo has been optimized away

[Bug c++/12668] Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[rmerkert at alphatech dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668



------- Additional Comments From rmerkert at alphatech dot com  2003-10-18 13:51 -------
One more example, where I don't really want to see warning:
cat > test.cpp <<EOF
#include <iostream>
struct Foo {

  inline Foo (int x) : foo(0) {init(x); }

  private:
   inline void init (int x) {
     if (x!=0)
       foo = x + 1;
   }
  int foo;
};


struct Z {
  Z (int x) : foo(0),z(x) {}
  Foo foo;
  int z;
};
int main (int argc, int argv)
{
  Z z(argc);
  ::std::cerr << "Z: " << z.z << ::std::endl;
  return 0;
}


EOF
g++ -O1 -Wunreachable-code test.cpp


The warning in this case is:
test.cpp: In function `int main(int, int)':
test.cpp:9: warning: will never be executed

There is no unreachable code in the example - it's just that 
by choosing a certain parameter, a path in the program does not
need to be executed. That's not the same as being unreachable, that's
optimizing away things.
I've been getting lots of these warnings when I compile with O1 because of this
kind of situation. I think things are just propagated too far down the line.
What if I had 7 nested inlined functions that turn out to be "unreachable".

[Bug c++/12668] Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668



------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-18 18:58 -------
That is the reason why -Wunreachable-code is not turned on by default or included in -W 
(-Wextra) or -Wall.

[Bug optimization/12668] Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-06 09:44 -------
Invalid as -Wunreachable-code is right no matter which way you look at it (do not use 
-Wunreachable-code unless you want to know that placeswhere are unreachable in some cases).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
          Component|c++                         |optimization
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668

[Bug optimization/12668] Strange warnings with -O1 -Wunreachable-code in 3.3.1,3.3.2,3.4

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-04-08 13:00 -------
*** Bug 14889 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jmarronn at rockwellcollins
                   |                            |dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12668