Avoid false positive warnings turned into an error #791
Conversation
|
Since you are here, could you also add my other comment #783 (comment) ? |
Thank you, Alfio! Will do. The comment did not arrive on my end yet because of the GitHub user name (hpf). Ahh, High Performance Fortran for sure ;-) |
Oh dear... sorry for that... I will try to fix the macos error too (seems now we have gcc v13). |
hfp
force-pushed
the
upstream
branch
2 times, most recently
from
876c7f1
to
7b5a4d5
Compare
|
GCC/latest test is now passing. |
hfp
force-pushed
the
upstream
branch
2 times, most recently
from
4f331b0
to
f2c634b
Compare
cmake/CompilerConfiguration.cmake
Outdated
| set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -fallow-argument-mismatch") # requires for 10+ for the MPI wrap module | ||
| set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -fallow-argument-mismatch") # required for 10+ (MPI wrap) | ||
| endif () | ||
| if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 13) # comparison against CXX version rather than GFortran version |
There was a problem hiding this comment.
It seems we have to extend the condition to cover both cases, i.e.:
if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 13) # false positive (allocatable array)
set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -Werror=uninitialized")
else ()
set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -Wno-error=uninitialized")
endif ()
Then, the warning seems real, due to #694 (likely the new GCC is more pedantic).
|
In the meantime, I've merged #792 to fix ROCm and macOS CI tests |
For instance, https://github.com/cp2k/dbcsr/actions/runs/9078189846/job/24944690222