[Bug rtl-optimization/126412] [17 Regression] csmith: wrong code with -O3 -march=arm-32 -fno-strict-aliasing

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126412

--- Comment #10 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to Konstantinos Eleftheriou from comment #9)
> Hi David, thanks for reporting. I reproduced this and there seems to
> be an issue with our cancellation logic.

You are welcome. Please note that I had to use flags "-O3 -march=native" 
to find this bug and your testing seems to be an ordinary -g -O2 bootstrap.

If you have the resource, you might consider it an improvement
to your testing to use build target bootstrap-O3, but that
doesn't do -march=native.

For my bootstrap testing, I go further and use the following script:

CFLAGS="-g1 -O3 -march=native" \
CXXFLAGS="-g1 -O3 -march=native" \
../trunk/configure --prefix=$HOME/gcc/results \
        --disable-doc \
        --disable-multilib \
        --disable-werror \
        --enable-checking=yes \
        --enable-languages="c,c++,fortran" \
        --with-build-config="bootstrap-asan bootstrap-ubsan"

Just an idea, doesn't have to happen.