Re: [gfortran] PATCH Fix PR 20058

[Tobias Schlüter <tobias dot schlueter at physik dot uni-muenchen dot de>]

Toon Moene wrote:
> It all started by the following remarks by our esteemed Editor of the 
> Fortran 2003 Standard, Richard Maine (in paper 02-102.txt for meeting 
> 160, February, 25 - March, 1, 2002):
> 
>      issue 351 - boz and REAL
> 
>       ....
> Hope this helps,

I understand that, and I had thought about much the same lines (esp. the
question "how do I initialize a REAL*10 to a certain bit pattern").
Fortunately, the exception Steve quoted is explicitly restricted to constants
in DATA statements.  In the other contexts where boz-literal-constants are
allowed by the standard they can still be 10 bytes wide, or whatever seems
appropriate.

If we want to support the full F2K semantics we'll have to write a more
general matcher and devise a way of converting bit patterns to REALs in the
frontend.

- Tobi