This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[PATCH] Fix PR84803


The following fixes PR84803 where we have a memory reference 
data-reference analysis thought it's not worth handling.
This properly guards the access.

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

Richard.

2018-03-12  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/84803
	* tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
	for refs DR analysis didn't process.

	* gcc.dg/torture/pr84803.c: New testcase.

Index: gcc/tree-if-conv.c
===================================================================
--- gcc/tree-if-conv.c	(revision 258445)
+++ gcc/tree-if-conv.c	(working copy)
@@ -864,6 +864,11 @@ base_object_writable (tree ref)
 static bool
 ifcvt_memrefs_wont_trap (gimple *stmt, vec<data_reference_p> drs)
 {
+  /* If DR didn't see a reference here we can't use it to tell
+     whether the ref traps or not.  */
+  if (gimple_uid (stmt) == 0)
+    return false;
+
   data_reference_p *master_dr, *base_master_dr;
   data_reference_p a = drs[gimple_uid (stmt) - 1];
 
Index: gcc/testsuite/gcc.dg/torture/pr84803.c
===================================================================
--- gcc/testsuite/gcc.dg/torture/pr84803.c	(nonexistent)
+++ gcc/testsuite/gcc.dg/torture/pr84803.c	(working copy)
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+
+long a;
+long *b;
+void c ();
+void d ();
+void
+e (long f)
+{
+  if (a)
+    *b = f;
+}
+void
+g ()
+{
+  c (g, e);
+}
+void
+c (int f, int h ())
+{
+  d (f, h, "");
+}
+void
+d (int f, int h (), char *i, char *k)
+{
+  int j;
+  d (f, h, i + 1, k);
+  while (--j)
+    h (*i);
+}


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]