[Bug c++/108747] New: Optimization makes user mistake more dangerous when optimization is disabled by flag -O0

[Bug c++/108747] New: Optimization makes user mistake more dangerous when optimization is disabled by flag -O0

[alireza at hexosys dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108747

            Bug ID: 108747
           Summary: Optimization makes user mistake more dangerous when
                    optimization is disabled by flag -O0
           Product: gcc
           Version: og11 (devel/omp/gcc-11)
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: alireza at hexosys dot com
  Target Milestone: ---

As we can see below link:
https://godbolt.org/z/ze3nhszxx

although the user had a mistake and create undefined behavior maybe he expected
it because of the -O0 flag, the compiler won't do anything smart and use div
for %. so he thinks his code will work correctly.

I think -the O0 flag means no smarty thing for the compiler.

[Bug c++/108747] Dangerous optimization when optimization is disabled by flag -O0

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108747

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|UNCONFIRMED                 |RESOLVED
            Version|og11 (devel/omp/gcc-11)     |13.0

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
>        and     eax, 7


What are you asking about? Why GCC decided to still use %8 here (and do &7)
even with the static_cast?

The reasoning is because constexpr causes a const integer expression here so
the front-end will do the replacement.

There is nothing to fix here really because the user asked the compiler to do
the casting away the const via const_cast and everything.

[Bug c++/108747] Dangerous optimization when optimization is disabled by flag -O0

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108747

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 54446
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=54446&action=edit
original testcase

Next time also please don't just link godbolt but either put the testcase
inline or attach it. Without that context (in the bug report) it was even hard
to figure out what you were asking about.

[Bug c++/108747] Dangerous optimization when optimization is disabled by flag -O0

[alireza at hexosys dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108747

--- Comment #3 from alireza at hexosys dot com ---
I Agree with you that the user has made mistake.

but why does the compiler does this optimization that can change the behavior
of the code, when we explicitly asked her to don't do any optimization when we
set the -O0?

[Bug c++/108747] Dangerous optimization when optimization is disabled by flag -O0

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108747

--- Comment #4 from Jonathan Wakely <redi at gcc dot gnu.org> ---
-O0 does not mean "let me write invalid code"