Does this program make sense?

Help
1

Hi, I am new to fortran, and I’m playing with different compilers.
I find something I can’t explain, gfortran give different answers with intel’s compiler for the following program.
Does it make sense?

function f(str) result(alen)
character(*) :: str
character(len=len(str)), allocatable :: str2
integer :: alen
allocate(str2)
str2 = str
alen = len(str2)
end function

program aa
interface
function f(str) result(alen)
character(*) :: str
integer :: alen
end function
end interface

print *, f(“sdfsdf”)
pause
end program

2

I’m not sure, but I don’t think that the syntax

character(len=n), allocatable :: str
allocate(str)

is possible (unless n is a parameter that is known at compile-time). However:

character(len=n) :: str

or

character(len=:), allocatable :: str
allocate( character(len=n) :: str )

are possible for sure.

Anyway, whether your syntax is valid or not, it seems that gfortran has a bug here: it should either report a compilation error or give the right result.

PS: please format your code by enclosing it beetween triple backticks, i.e. (the </> button at the top of the editor will also do, after you select your code):

```
some code
more code
```

3

I believe that

character(len=n), allocatable :: str

is valid. It means that you declared an allocatable character with a fixed length n but it won’t access any memory until you explicitly allocate it.

4

character(len=len(str)), allocatable :: str2
is allowed because len(str) is a valid specification expression by virtue of the fact that str is a dummy argument that has neither the OPTIONAL nor the INTENT(OUT) attribute and LEN is an intrinsic inquiry function. It could also be written as str%len (a type-parameter-inquiry), if the compiler supports it.

5

Thank you. Will use that button next time.

6

Does that mean gfortran has a bug? I mean, I am totally new to fortran.

7

But you never said what the answers from the two compilers are. Can you enlighten us? (We may have different versions and so get different answers :slight_smile:).

8

I see no reason this program would be considered invalid, and it ought to behave as you probably expect. That said, it is an unusual example/style, and is mixing several features not commonly seen together in ways that are redundant, so not surprised by any bugs you find in compilers with it.

For example, you could remove the allocatable attribute and allocate statement, and the program would likely work fine. This makes str2 an “automatic” variable.

On the other hand you could change len(str) in the declaration to :, and either omit the allocate statement (to get automatic allocation on assignment) or change it to allocate(character(len=len(str)) :: str2). This has become the more common way since this feature was introduced.

9

Intel’s compiler gives the answer I expect, that is length of the string.
Different versions of gfortran give different answers.
Sometimes it gives random integers, sometimes the program backtraces.

10

Yes, I tried a few ways to rewrite the program and it works. I am just curious.

11

I am not sure why you need str2 at all, unless you are testing allocatable character strings ?

The following works with Gfortran ( after correcting printf as posted )
It could be better to have f in CONTAINS

  function f(str) result(alen)
    character(*) :: str
    character(len=len(str)), allocatable :: str2
    integer :: alen
    allocate(str2)
    str2 = str
    alen = len(str2)
  end function

  integer function slen (str)
    character(*) :: str
    slen = len(str)
  end function slen

 program aa
  interface
    function f(str) result(alen)
    character(*) :: str
    integer :: alen
    end function f

    integer function slen (str)
    character(*) :: str
    end function slen
  end interface

  print *, f ( 'sdfsdf' )
  print *, slen ( 'sdfsdf' )
!  pause
 end program

Also check your print statement, I changed: print *, f(“sdfsdf”)
to : print *, f ( ‘sdfsdf’ )
as your " may be another unicode character.