Argument mismatch errors in C code with GCC-15

[nncarlson]

I'm trying to build scalapack-2.2.2 using the new gcc-15 compiler and am encountering this error, and many others like it, in C code:

/opt/src/scalapack/scalapack-2.2.2/TOOLS/reshape.c:32:4: error: too many arguments to function ‘Cblacs_gridinfo’; expected 0, have 5
   32 |    Cblacs_gridinfo( context_in, &nprow_in, &npcol_in, &myrow_in, &mycol_in );
      |    ^~~~~~~~~~~~~~~  ~~~~~~~~~~
/opt/src/scalapack/scalapack-2.2.2/TOOLS/reshape.c:15:9: note: declared here
   15 |    void Cblacs_gridinfo();
      |         ^~~~~~~~~~~~~~~

It seems gcc has gotten stricter since version 12, which compiles this without error. Is there a gcc option that downgrades this to a warning (like exists for Fortran)? I'm not seeing anything that jumps out at me.

[nncarlson]

GCC-15 upped the default for -std= to gnu23 (more or less c23) from gnu18, which resulted in this error. The solution I found was to use -std=gnu90. This had the additional benefit of eliminating the need to use -Wno-implicit-function-declarations, which I had been using all along. (Are there any plans to update the source to conform to contemporary standards?)

Feel free to close.

[bangerth]

May I advocate for actually fixing this so that the code compiles based on current standards? C has moved on from implicit declarations where you can pass whatever you want as arguments, to a type-safe system in which the compiler can check whether the passed arguments actually have the right type and number. Most people think that that is progress. Relying on old K&R where declarations weren't necessary is just asking for hard-to-find bugs, and it's definitely not in the spirit of 2025 any more.

[mgates3]

@bangerth Updating the code to modern C practices such as prototypes — which has been a best practice since C89 and, IMO, should have been followed since the start in 1995 — would be great. But it's a lot of work and not much funding, so don't hold your breath. Any volunteers?