Assumed Rank Character Variable Fails in Select Rank Statement

[CarlB]

Hello everyone,

This function compiles with all data types using gfortran. ifort, however, fails if I try to define a behaviour for rank(0). Higher ranks do work, including rank default.

Minimal example:

subroutine sub(c)
character(*), intent(in) :: c(..)

select rank(c)
rank(0)
   print*, 0
end select
end subroutine sub

 Result:

$ ifort -c ./src/lib.f90 -warn all -check all -error-limit 1 -O0 -g -assume byterecl -traceback -module build/ifort_5C58216731706F11/testingrange -Ibuild/ifort_5C58216731706F11/testingrange  -o build/ifort_5C58216731706F11/testingrange/src_lib.f90.o
./src/lib.f90(10): error #5529: CHARACTER variable 'C' has no length argument in routines with C or STDCALL attribute
select rank(c)
------------^

 Compiler provided by my company:

ifort (IFORT) 19.1.2.254 20200623

Is this a bug or a feature? At least the error message isn't that clear.

I assume it has something to do with the additional len parameter of the character type. But I'm curious because it does compile and work with gfortran.

Best,
Carl

[Steve_Lionel]

Bug.

[Barbara_P_Intel · Accepted solution]

I got the same wrong error message with the current compiler release, 2021.3.0. I filed a bug report, CMPLRIL0-34096, on your behalf. I'll keep you posted on its progress to a fix.

[CarlB]

Thank you for clarifying and for the bug report!

[Devorah_H_Intel]

The fix will be available in the next oneAPI release. 

[MarkBuehner]

I'm having the same error with inteloneapi-2022.1.2. Is it fixed in a later version? If so, which one?

Thanks!

[Steve_Lionel]

The example compiles successfully in the current ifx version (2025.1.0)

[MarkBuehner]

Thanks so much!

[ethanmusk89]

You're spot on — the issue arises due to how ifort handles character arguments with assumed-rank and missing length metadata, especially for rank(0). Unlike gfortran, ifort expects a length parameter even if the rank is zero, and that’s why the compiler throws that error. It’s not exactly a bug, but more of a strict interpretation of Fortran standards in Intel’s implementation.

I ran into a similar issue while testing assumed-rank character handling across compilers and shared a few insights and workarounds here :executordelta.com — might be useful if you're aiming for cross-compiler compatibility.