Re: [Patch, fortran] PR18769, PR25049 and PR25050 - transformational functions in initialization expressions.

[Paul Thomas <paulthomas2 at wanadoo dot fr>]

FX,

> I'm probably mistaken here, but I thought there was a difference 
> between what F95 allowed as initialization expr. and what F2003 
> allows. For example, the way I read and understood both standards, the 
> following code:
>
>   INTEGER, PARAMETER :: K(2)=1
>   SELECT CASE(I)
>   CASE(MAXLOC(K,1))
>   END SELECT
>   END
>
> (from PR 25104) is invalid in F95 but valid according to F2003.
>
> Am I right?

That is correct.  However, in the absence of a simplifier, these 
intrinsics all do some variant on what your example does:
test2003.f90:1: internal compiler error: in gfc_conv_constant_to_tree, 
at fortran/trans-const.c:276

If somebody has the intestinal fortitude, they could write 
gfc_simplify_maxminloc and friends, changing the error in check.c to 
being standard dependent.  At the moment, however, let's aim to make 
gfortran F95 compliant and get rid of these nasty ICE's.

In file test2003.f90:3

 CASE(MAXLOC(K,1))
     1
Error: transformational intrinsic 'maxloc' at (1) is not permitted in an 
initialization expression

seems friendlier, doesn't it?

Talking of intestinal fortitude, what constitutes a "strong breakfast"? 
.... a la danoise, with schnaps and raw fish?

Paul