public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] fix ICE in generic_overlap (PR 84526)
@ 2018-02-23 19:57 Martin Sebor
  2018-02-23 20:49 ` Jakub Jelinek
  0 siblings, 1 reply; 20+ messages in thread
From: Martin Sebor @ 2018-02-23 19:57 UTC (permalink / raw)
  To: Gcc Patch List, Jakub Jelinek

[-- Attachment #1: Type: text/plain, Size: 898 bytes --]

r257860 introduced an unsafe assumption under some conditions
that the base object referenced by a memcpy strcpy necessarily
has a pointer or array type.  The attached patch removes this
assumption.

In addition, as discussed in the bug and in IRC, the patch
also removes a test for the result of get_inner_reference()
being non-null (a vestige of handling the result of a call
to get_addr_base_and_unit_offset() made previously here),
and a test for the bit offset computed by get_inner_reference()
not being convertible to HOST_WIDE_INT.  As far as we can tell
poly64_int is always convertible to HOST_WIDE_INT (i.e.,
poly64_int::is_constant() always returns true).

The conversion from poly64_int to HOST_WIDE_INT is necessary
even though the result is stored in an offset_int because there
is apparently no way to convert from the former to the latter
directly, even though it's wider.

Martin

[-- Attachment #2: gcc-84526.diff --]
[-- Type: text/x-patch, Size: 4721 bytes --]

PR tree-optimization/84526 - ICE in generic_overlap at gcc/gimple-ssa-warn-restrict.c:927 since r257860

gcc/ChangeLog:

	PR tree-optimization/84526
	* gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
	Remove dead code.
	(builtin_access::generic_overlap): Be prepared to handle non-array
	base objects.

gcc/testsuite/ChangeLog:

	PR tree-optimization/84526
	* gcc.dg/Wrestrict-10.c: New test.

Index: gcc/gimple-ssa-warn-restrict.c
===================================================================
--- gcc/gimple-ssa-warn-restrict.c	(revision 257933)
+++ gcc/gimple-ssa-warn-restrict.c	(working copy)
@@ -409,15 +409,15 @@ builtin_memref::set_base_and_offset (tree expr)
   base = get_inner_reference (expr, &bitsize, &bitpos, &var_off,
 			      &mode, &sign, &reverse, &vol);
 
+  /* get_inner_reference is not expected to return null.  */
+  gcc_assert (base != NULL);
+
   poly_int64 bytepos = exact_div (bitpos, BITS_PER_UNIT);
 
-  HOST_WIDE_INT const_off;
-  if (!base || !bytepos.is_constant (&const_off))
-    {
-      base = get_base_address (TREE_OPERAND (expr, 0));
-      return;
-    }
-
+  /* There is no conversion from poly_int64 to offset_int even
+     though the latter is wider, so go through HOST_WIDE_INT.
+     The offset is expected to always be constant.  */
+  HOST_WIDE_INT const_off = bytepos.to_constant ();
   offrange[0] += const_off;
   offrange[1] += const_off;
 
@@ -923,7 +923,11 @@ builtin_access::generic_overlap ()
       /* There's no way to distinguish an access to the same member
 	 of a structure from one to two distinct members of the same
 	 structure.  Give up to avoid excessive false positives.  */
-      tree basetype = TREE_TYPE (TREE_TYPE (dstref->base));
+      tree basetype = TREE_TYPE (dstref->base);
+      if (POINTER_TYPE_P (basetype)
+	  || TREE_CODE (basetype) == ARRAY_TYPE)
+	basetype = TREE_TYPE (basetype);
+
       if (RECORD_OR_UNION_TYPE_P (basetype))
 	return false;
     }
Index: gcc/testsuite/gcc.dg/Wrestrict-10.c
===================================================================
--- gcc/testsuite/gcc.dg/Wrestrict-10.c	(nonexistent)
+++ gcc/testsuite/gcc.dg/Wrestrict-10.c	(working copy)
@@ -0,0 +1,121 @@
+/* PR tree-optimization/84526 - ICE in generic_overlap
+   { dg-do compile }
+   { dg-options "-O2 -Wrestrict" } */
+
+typedef __SIZE_TYPE__ size_t;
+
+extern void* memcpy (void* restrict, const void* restrict, size_t);
+extern char* strcat (char* restrict, const char* restrict);
+extern char* strcpy (char* restrict, const char* restrict);
+extern char* strncat (char* restrict, const char* restrict, size_t);
+extern char* strncpy (char* restrict, const char* restrict, size_t);
+
+struct
+{
+  char a[1];
+} b;
+
+int i;
+size_t n;
+
+void __attribute__ ((noclone, noinline))
+test_arr_memcpy_1 (void)
+{
+  memcpy (&b.a[i], b.a, n);
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_memcpy_2 (void)
+{
+  memcpy (b.a, &b.a[i], n);
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_strcat_1 (void)
+{
+  strcat (&b.a[i], b.a);            /* { dg-warning "\\\[-Wrestrict" } */
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_strcat_2 (void)
+{
+  /* This probably deserves a warning.  */
+  strcpy (b.a, &b.a[i]);
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_strncat_1 (void)
+{
+  strncat (&b.a[i], b.a, n);        /* { dg-warning "\\\[-Wrestrict" } */
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_strncat_2 (void)
+{
+  strncat (b.a, &b.a[i], n);        /* { dg-warning "\\\[-Wrestrict" } */
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_strcpy_1 (void)
+{
+  strcpy (&b.a[i], b.a);
+}
+
+void __attribute__ ((noclone, noinline))
+test_arr_strcpy_2 (void)
+{
+  strcpy (b.a, &b.a[i]);
+}
+
+
+struct S {
+  int a;
+  char b[10];
+} d;
+
+void __attribute__ ((noclone, noinline))
+test_obj_memcpy_1 (void)
+{
+  memcpy (d.b, (char *) &d, n);
+}
+
+void __attribute__ ((noclone, noinline))
+test_obj_memcpy_2 (void)
+{
+  memcpy ((char *) &d, d.b, n);
+}
+
+void __attribute__ ((noclone, noinline))
+test_obj_strcpy_1 (void)
+{
+  strcpy (d.b, (char *) &d);
+}
+
+void __attribute__ ((noclone, noinline))
+test_obj_strcpy_2 (void)
+{
+  strcpy ((char *) &d, d.b);
+}
+
+void __attribute__ ((noclone, noinline))
+test_obj_strncat_1 (void)
+{
+  strncat (d.b, (char *) &d, n);    /* { dg-warning "\\\[-Wrestrict" } */
+}
+
+void __attribute__ ((noclone, noinline))
+test_obj_strncat_2 (void)
+{
+  strncat ((char *) &d, d.b, n);    /* { dg-warning "\\\[-Wrestrict" } */
+}
+
+void __attribute__ ((noclone, noinline))
+test_obj_strncpy_1 (void)
+{
+  strncpy (d.b, (char *) &d, n);
+}
+
+void test_obj_strncpy_2 (void)
+{
+  strncpy ((char *) &d, d.b, n);
+}

^ permalink raw reply	[flat|nested] 20+ messages in thread
end of thread, other threads:[~2018-03-09 19:07 UTC | newest]

Thread overview: 20+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-02-23 19:57 [PATCH] fix ICE in generic_overlap (PR 84526) Martin Sebor
2018-02-23 20:49 ` Jakub Jelinek
2018-02-23 21:46   ` Martin Sebor
2018-02-24  9:32     ` Richard Sandiford
2018-02-24 14:45       ` Martin Sebor
2018-02-25 10:03         ` Richard Sandiford
2018-02-24 12:49     ` Jakub Jelinek
2018-02-23 23:25       ` Martin Sebor
2018-02-26 17:32     ` Martin Sebor
2018-02-26 20:49       ` Jakub Jelinek
2018-02-26 18:13         ` Martin Sebor
2018-03-07 15:17           ` Martin Sebor
2018-02-27 12:33       ` Richard Sandiford
2018-03-07 19:37       ` Jeff Law
2018-03-07 22:52         ` Martin Sebor
2018-03-07 23:04           ` Richard Sandiford
2018-03-07 23:59             ` Martin Sebor
2018-03-09 19:07           ` Jeff Law
2018-03-07 19:18     ` Jeff Law
2018-03-07 19:14   ` Jeff Law

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).