[PATCH] Fix PR87917

[Richard Biener]

The following side-steps a possible issue with the 
evolution_function_is_affine_multivariate_p predicate by guarding
the call to analyze_subscript_affine_affine in analyze_miv_subscript
in the same way as the call from analyze_siv_subscript.

Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.

Richard.

2018-11-15  Richard Biener  <rguent...@suse.de>

        PR middle-end/87917
        * tree-data-ref.c (analyze_miv_subscript): Guard calls to
        analyze_subscript_affine_affine properly.

        * gcc.dg/tree-ssa/pr87917.c: New testcase.

Index: gcc/tree-data-ref.c
===================================================================
--- gcc/tree-data-ref.c (revision 266145)
+++ gcc/tree-data-ref.c (working copy)
@@ -3994,9 +3993,9 @@ analyze_miv_subscript (tree chrec_a,
       dependence_stats.num_miv_independent++;
     }
 
-  else if (evolution_function_is_affine_multivariate_p (chrec_a, 
loop_nest->num)
+  else if (evolution_function_is_affine_in_loop (chrec_a, loop_nest->num)
           && !chrec_contains_symbols (chrec_a)
-          && evolution_function_is_affine_multivariate_p (chrec_b, 
loop_nest->num)
+          && evolution_function_is_affine_in_loop (chrec_b, loop_nest->num)
           && !chrec_contains_symbols (chrec_b))
     {
       /* testsuite/.../ssa-chrec-35.c

Index: gcc/testsuite/gcc.dg/tree-ssa/pr87917.c
===================================================================
--- gcc/testsuite/gcc.dg/tree-ssa/pr87917.c     (nonexistent)
+++ gcc/testsuite/gcc.dg/tree-ssa/pr87917.c     (working copy)
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-loop-distribute-patterns -ftrapv -fno-tree-fre" } 
*/
+
+void foo(int x[])
+{
+ int i, j;
+
+ for (i = 0; i < 2; i++)
+   for (j = 0; j < 2; j++)
+   {
+     x[i] = x[i*j];
+     x[i] = x[i*j];
+   }
+}