Re: [Patch, fortran] PR24174 real(10) and complex(10) array IO broken.

[Janne Blomqvist <jblomqvi at cc dot hut dot fi>]

On Tue, Oct 25, 2005 at 12:49:34PM +0200, FX Coudert wrote:
> >PING**2
> 
> It looks OK. Only one comment...
> 
>      gfc_build_library_function_decl (get_identifier
>  				     (PREFIX("transfer_array")),
> -				     void_type_node, 2, pvoid_type_node,
> +				     void_type_node, 3, pvoid_type_node,
> +				     gfc_int4_type_node,
>  				     gfc_charlen_type_node);
> 
> contradicts
> 
>  void
> -transfer_array (gfc_array_char *desc, gfc_charlen_type charlen)
> +transfer_array (gfc_array_char *desc, int kind, gfc_charlen_type charlen)
> 
> since a gfc_int4_type_node might not always be an int. Could you use 
> gfc_c_int_type_node?

Ok, here is an updated patch, fixing the above issue. While I was at
it, I also made formatted complex(10) input work in addition to output
(i.e. fixing PR 24305 as well). I also made a couple of utility
functions for determining the size from the kind, and put them in
their own file.

However, there is still some kind of issue with formatted input of
very big real(10) numbers. But AFAICS that has to do with parsing and
not with the padding issues that this patch tries to address, so I'll
file a separate PR for that.



-- 
Janne Blomqvist

Attachment: ChangeLog-3
Description: Text document

Attachment: pr24174-3.diff
Description: Text document

Attachment: size_from_kind.c
Description: Text document

Attachment: kind10_formatted_io.f90
Description: Text document