[Bug target/106827] New: operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[Bug target/106827] New: operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[bugzilla@little-bat.de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

            Bug ID: 106827
           Summary: operator++ doesn't work for enum  -O2
                    -mcpu=cortex-m0plus
           Product: gcc
           Version: 11.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: bugzilla@little-bat.de
  Target Milestone: ---

Created attachment 53536
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53536&action=edit
the example on godbolt.org

gcc "optimizes" away prefix and postfix operator++ for enums.

see https://godbolt.org/z/1E65Ej7aE

--> three loops, only version 3 works (presumably).

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[bugzilla@little-bat.de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

--- Comment #1 from Kio <bugzilla@little-bat.de> ---
updated link at godbolt:

https://godbolt.org/z/5sf7cr9c3

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[rearnsha at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

Richard Earnshaw <rearnsha at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2022-09-05
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |WAITING

--- Comment #2 from Richard Earnshaw <rearnsha at gcc dot gnu.org> ---
Please put testcases in the bug report, godbolt links are not stable.

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW
                 CC|                            |marxin at gcc dot gnu.org

--- Comment #3 from Martin Liška <marxin at gcc dot gnu.org> ---
Copied test-case:

enum Foo{a,b,c,d,e,f};

Foo operator+(Foo n,int m) { return Foo(int(n)+m); }
Foo operator++(Foo n,int) { return Foo(int(n)+1); }
Foo operator++(Foo n) { return Foo(int(n)+1); }

void test1()
{
    for (Foo i=a; i<=f; i++) 
    {
        volatile Foo z=i;
    }
}

void test2()
{
    for (Foo i=a; i<=f; ++i) 
    {
        volatile Foo z=i;
    }
}

void test3()
{
    for (Foo i=a; i<=f; i=i+1) 
    {
        volatile Foo z=i;
    }
}

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[bugzilla@little-bat.de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

--- Comment #4 from Kio <bugzilla@little-bat.de> ---
(In reply to Richard Earnshaw from comment #2)
> Please put testcases in the bug report, godbolt links are not stable.

test case was also attached as a file with the bug report.

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[schwab@linux-m68k.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

Andreas Schwab <schwab@linux-m68k.org> changed:

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

--- Comment #5 from Andreas Schwab <schwab@linux-m68k.org> ---
The operator++ function is expected to modify the argument in place, and needs
to take a reference to it.

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think C++ doesn't explicitly say that the argument has to be a reference:
https://eel.is/c++draft/over.oper#over.inc
but obviously if it isn't and it isn't a non-static method operator++,
then it can only compute the return value and can't actually modify in place.
For ++i just operator++(i) is called and for i++ operator++(i, 0).
So, with your overloads you'd actually need to do i = ++i
The other overload is just unusable because it does the wrong thing completely.

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[bugzilla@little-bat.de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

--- Comment #7 from Kio <bugzilla@little-bat.de> ---
(In reply to Andreas Schwab from comment #5)
> The operator++ function is expected to modify the argument in place, and
> needs to take a reference to it.

Yes, that's my error! after modifying the operators to:

Foo operator++(Foo& n,int) { Foo v=n; n=Foo(int(n)+1); return v; }
Foo& operator++(Foo& n)    { return n=Foo(int(n)+1); }

then the compiled code is as expected.

Obviously arg#1 must be a reference, because the argument must be modified.
Actually i think this code should not compile or at least produce a warning (i
compile with almost all warnings on, but that went silent). When passing a
value the operator can't do what is expected.

[Bug target/106827] operator++ doesn't work for enum -O2 -mcpu=cortex-m0plus

[rearnsha at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106827

--- Comment #8 from Richard Earnshaw <rearnsha at gcc dot gnu.org> ---
That's C++ for you, it permits you to overload operators in completely
meaningless ways.