[Bug fortran/32732] [Bind C] Character scalars are passed as arrays

[Christopher D. Rickett]

i think you're right in that we do have to do more than just change the 
type.  the type does need to change, to promote character scalar args to 
ints just as the C frontend does.  however, since the fortran frontend 
will expect them as strings of length one, it seems they'll have to be 
converted to that upon entering a bind(c) routine.

thoughts?
Chris

On Mon, 30 Jul 2007, Steve Ellcey wrote:

>> thanks for trying it out.  i'm quite at a loss now as to why it's failing;
>> i thought that fixing the array_type still being seen in
>> gfc_conv_function_call would do it.  now i'm not even sure where to start
>> looking...  the only thing i could think to do now is to run cc1 through
>> gdb and figure out what the C frontend is doing with characters that isn't
>> happening on the fortran side.  it would be good to see the debug_tree
>> output for the tree created by cc1, but that would be a bit more work for
>> you.  i'll see if i can figure out where to put the call to debug_tree and
>> maybe pass it on to you.
>>
>> thanks again for all of the help.
>> Chris
>
> I wonder if we have to do more than change the type.  If I understand
> the code correctly we have converted the type of the formal parameter in
> the call from an array of char to just a char.  But the actual argument
> to the call (the variable or constant being passed) is still an array of
> char.  Do we need code to change the actual argument so we wind up
> passing "array[1]" instead of "array", i.e.  to pass in just the first
> element of the array instead of the entire array.
>
> In C, if I have
>
> 	char s[10];
> 	call foo(s);
>
> That is different then
>
> 	char s[10];
> 	call foo(s[0]);
>
> Regardless of what prototype foo has and what type we claim the argument
> to foo is.
>
> Steve Ellcey
> sje@cup.hp.com
>