Help with the front-end

[Paul Brook]

On Tuesday 28 June 2005 16:06, François-Xavier Coudert wrote:
> Hi all,
>
> With my current front-end patch for large real kinds, I have a small
> problem with large complex kinds. The following code:
>
>   complex(10) :: w
>   w = sin (w)
>
> gives the correct parse tree: "ASSIGN w __sin_c8[[((w))]]" but an

You mean __sin_c10, right?
The name generated by the frontend resolution routines is often ignored. Many 
of the code generation routines just look at the type of the expression.

> incorrect tree-original: "w = csin (w);" (instead of csinl).
>
> I don't really know where to look. Where do this transformation from
> the __sin_c8 form to the libm function occurs? Grepping didn't help me
> on this...

trans-intrinsic.c:gfc_conv_intrinsic_lib_function

Paul