This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[patch, fortran] Fix PR 66385, array constructors in FORALL
- From: Thomas Koenig <tkoenig at netcologne dot de>
- To: "fortran at gcc dot gnu dot org" <fortran at gcc dot gnu dot org>, gcc-patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 04 Jun 2015 11:46:24 +0200
- Subject: [patch, fortran] Fix PR 66385, array constructors in FORALL
- Authentication-results: sourceware.org; auth=none
Hello world,
front-end optimization and FORALL do not appear to mix well.
This patch fixes an ICE resulting from an attempt by front-end
optimization to use a BLOCK inside a FORALL statement.
I will commit this as obvious in a day or so unless somebody objects.
I will also backport (time to install the gcc 5 branch...)
There would be an alternative, in principle, to put the BLOCK around
the FORALL, but frankly, I don't think it is worth spending the
effort and complicating the code.
Regards
Thomas
2015-06-11 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/66385
* frontend-passes.c (combine_array_constructor): Return early if
inside a FORALL loop.
2015-06-11 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/66385
* gfortran.dg/forall_17.f90: New test.
! { dg-do compile }
! { dg-options "-ffrontend-optimize" }
! PR fortran/66385 - this used to ICE
! Original test case by Mianzhi Wang
program test
double precision::aa(30)
double precision::a(3,3),b
b=1d0
forall(i=1:3)
a(:,i)=b*[1d0,2d0,3d0]
end forall
forall(i=1:10)
aa(10*[0,1,2]+i)=1d0
end forall
end program
Index: frontend-passes.c
===================================================================
--- frontend-passes.c (Revision 223876)
+++ frontend-passes.c (Arbeitskopie)
@@ -1243,6 +1243,10 @@ combine_array_constructor (gfc_expr *e)
if (in_assoc_list)
return false;
+ /* With FORALL, the BLOCKS created by create_var will cause an ICE. */
+ if (forall_level > 0)
+ return false;
+
op1 = e->value.op.op1;
op2 = e->value.op.op2;