style: treat warnings as errors

[vil02]

Continuation of #18.

[Ramy-Badr-Ahmed]

Continuation of #18.

Thanks @vil02 ,

Now my local version has become in sync 🙂

I set it up like like this:
add_compile_options(-Wall -Wextra -Wpedantic -Wunused-variable -Werror)

Interested about other complile options?

[vil02]

I set it up like like this: add_compile_options(-Wall -Wextra -Wpedantic -Wunused-variable -Werror)

-Wunused-variable seams to be part of -Wall.

Interested about other complile options?

Sure, but I would be careful here: some of the warnings are already included into the used flags, some of the warnings are Fortran version specific etc.

[Ramy-Badr-Ahmed]

-Wunused-variable seams to be part of -Wall.

Right, I'll change my local gfortran version.

Sure, but I would be careful here: some of the warnings are already included into the used flags, some of the warnings are Fortran version specific etc.

I previously used these:

-Wconversion-extra     # Warn about implicit conversions that may cause issues
-fno-backtrace         # Enable backtrace information for runtime errors
-fcheck=all            # Enable all runtime checks (bounds, uninitialized, etc.)
-finit-real=nan        # Initialize all REAL variables to NaN.
-finit-integer=0       # Initializes all integer variables to 0.
-ffpe-trap=all         # Trap floating-point exceptions (zero,overflow,underflow)

[vil02]

@Ramy-Badr-Ahmed: I added -Wconversion-extra in fb742a5.

Regarding other flags, which you listed: they are not warnings and the discussion about them is out of scope of this PR. Few remarks for the future:

• I would not rely on -finit-real and -finit-integer but rather initialize all of the variable in the code,
• might make sense to use -fcheck=all and -ffpe-trap=all in the CI,
• -fno-backtrace is good for local debugging, in my opinion,
• we might want to introduce two types of builds: release and debug.

[Ramy-Badr-Ahmed]

Thanks for the feedback! I agree with your points, especially regarding introducing release and debug builds.

@Ramy-Badr-Ahmed: I added -Wconversion-extra in fb742a5.

Regarding other flags, which you listed: they are not warnings and the discussion about them is out of scope of this PR. Few remarks for the future:

Ok 👍

I would not rely on -finit-real and -finit-integer but rather initialize all of the variable in the code,

Yes, that's good practice. However, in the context of CI, these flags can help identify potential errors for contributors.

For example:

real :: x
x = x + 1.0  ! x is uninitialized

If compiled with -finit-real=nan, the program will crash or return NaN when it encounters x = NaN + 1.0.

might make sense to use -fcheck=all and -ffpe-trap=all in the CI,

-fno-backtrace is good for local debugging, in my opinion,

we might want to introduce two types of builds: release and debug.

For the builds, we could structure it like this:

• Debug build:

  • Include flags like -fcheck=all, -ffpe-trap=all, and keep -finit-real=nan, -finit-integer for uninitialized variable detection.
  • Keep backtrace enabled for detailed error reporting.

• Release build:

  • Strip runtime checks and use optimizations like -O2/-O3 for better performance.

[vil02]

I do not have enough knowledge about FORTRAN to judge relevance of the other warning flags.

@Ramy-Badr-Ahmed: all of this sounds like a good idea for the future work.
@SatinWukerORIG: please consider this PR as ready for a review. I will not push anymore changes here unless necessary.

[SatinWukerORIG]

This is a good way to eliminate potential conflicts and improve code quality for learners to understand. I am merging it 👍