[Patch, fortran] PR35932 - [4.3/4.4 Regression] ICE: CHAR with array arg and also a KIND arg

[Paul Richard Thomas]

Fixed on 4.4/4.3 as 'obvious' after bootstrapping and regtesting on
x86_ia64/FC8.

The ICE is a giveaway.  The conversion of the rhs must be incomplete
for this to occur.  Looking at trans-intrinsic.c
(gfc_conv_intrinsic_char) it was immediately obvious that the KIND
argument was not being converted.  This dates back to the patch that
brought in gfc_conv_intrinsic_function_args  - hence the regression.

I have not had time to go through all the intrinsics with KIND or very
infrequently used arguments.  However, I'll bet that there are more
like this.

Paul

2008-04-16  Paul Thomas  <pault@gcc.gnu.org>

	PR fortran/35932
	* trans-intrinsic.c (gfc_conv_intrinsic_char): Even though KIND
	is not used, the argument must be converted.

2008-04-16  Paul Thomas  <pault@gcc.gnu.org>

	PR fortran/35932
	* gfortran.dg/intrinsic_char_1.f90: New test.



-- 
The knack of flying is learning how to throw yourself at the ground and miss.
       --Hitchhikers Guide to the Galaxy