obtained gcc-4.6.4.tar.bz2 from upstream website;upstream

verified gcc-4.6.4.tar.bz2.sig;
imported gcc-4.6.4 source tree from verified upstream tarball.

downloading a git-generated archive based on the 'upstream' tag
should provide you with a source tree that is binary identical
to the one extracted from the above tarball.

if you have obtained the source via the command 'git clone',
however, do note that line-endings of files in your working
directory might differ from line-endings of the respective
files in the upstream repository.

1 files changed, 579 insertions, 0 deletions

diff --git a/gcc/testsuite/gcc.c-torture/execute/builtins/memmove-chk.c b/gcc/testsuite/gcc.c-torture/execute/builtins/memmove-chk.c
new file mode 100644
index 000000000..73b35883e
--- /dev/null
+++ b/gcc/testsuite/gcc.c-torture/execute/builtins/memmove-chk.c
@@ -0,0 +1,579 @@
+/* Copyright (C) 2004, 2005  Free Software Foundation.
+
+   Ensure builtin __memcpy_chk performs correctly.  */
+
+extern void abort (void);
+typedef __SIZE_TYPE__ size_t;
+extern size_t strlen(const char *);
+extern void *memcpy (void *, const void *, size_t);
+extern void *memmove (void *, const void *, size_t);
+extern int memcmp (const void *, const void *, size_t);
+
+#include "chk.h"
+
+const char s1[] = "123"; 
+char p[32] = "";
+volatile char *s2 = "defg"; /* prevent constant propagation to happen when whole program assumptions are made.  */
+volatile char *s3 = "FGH"; /* prevent constant propagation to happen when whole program assumptions are made.  */
+volatile size_t l1 = 1; /* prevent constant propagation to happen when whole program assumptions are made.  */
+
+void
+__attribute__((noinline))
+test1 (void)
+{
+  int i;
+
+#if defined __i386__ || defined __x86_64__
+  /* The functions below might not be optimized into direct stores on all
+     arches.  It depends on how many instructions would be generated and
+     what limits the architecture chooses in STORE_BY_PIECES_P.  */
+  memmove_disallowed = 1;
+  memcpy_disallowed = 1;
+#endif
+
+  /* All the memmove calls in this routine except last have fixed length, so
+     object size checking should be done at compile time if optimizing.  */
+  chk_calls = 0;
+
+  if (memmove (p, "ABCDE", 6) != p || memcmp (p, "ABCDE", 6))
+    abort ();
+  if (memmove (p + 16, "VWX" + 1, 2) != p + 16
+      || memcmp (p + 16, "WX\0\0", 5))
+    abort ();
+  if (memmove (p + 1, "", 1) != p + 1 || memcmp (p, "A\0CDE", 6))
+    abort ();
+  if (memmove (p + 3, "FGHI", 4) != p + 3 || memcmp (p, "A\0CFGHI", 8))
+    abort ();
+
+  i = 8;
+  memmove (p + 20, "qrstu", 6);
+  memmove (p + 25, "QRSTU", 6);
+  if (memmove (p + 25 + 1, s1, 3) != p + 25 + 1
+      || memcmp (p + 25, "Q123U", 6))
+    abort ();
+
+  if (memmove (memmove (p, "abcdEFG", 4) + 4, "efg", 4) != p + 4
+      || memcmp (p, "abcdefg", 8))
+    abort();
+
+  /* Test at least one instance of the __builtin_ style.  We do this
+     to ensure that it works and that the prototype is correct.  */
+  if (__builtin_memmove (p, "ABCDE", 6) != p || memcmp (p, "ABCDE", 6))
+    abort ();
+
+  memmove (p + 5, s3, 1);
+  if (memcmp (p, "ABCDEFg", 8))
+    abort ();
+
+  memmove_disallowed = 0;
+  memcpy_disallowed = 0;
+  if (chk_calls)
+    abort ();
+  chk_calls = 0;
+
+  memmove (p + 6, s1 + 1, l1);
+  if (memcmp (p, "ABCDEF2", 8))
+    abort ();
+
+  /* The above memmove copies into an object with known size, but
+     unknown length, so it should be a __memmove_chk call.  */
+  if (chk_calls != 1)
+    abort ();
+}
+
+long buf1[64];
+char *buf2 = (char *) (buf1 + 32);
+long buf5[20];
+char buf7[20];
+
+void
+__attribute__((noinline))
+test2_sub (long *buf3, char *buf4, char *buf6, int n)
+{
+  int i = 0;
+
+  /* All the memmove/__builtin_memmove/__builtin___memmove_chk
+     calls in this routine are either fixed length, or have
+     side-effects in __builtin_object_size arguments, or
+     dst doesn't point into a known object.  */
+  chk_calls = 0;
+
+  /* These should probably be handled by store_by_pieces on most arches.  */
+  if (memmove (buf1, "ABCDEFGHI", 9) != (char *) buf1
+      || memcmp (buf1, "ABCDEFGHI\0", 11))
+    abort ();
+
+  if (memmove (buf1, "abcdefghijklmnopq", 17) != (char *) buf1
+      || memcmp (buf1, "abcdefghijklmnopq\0", 19))
+    abort ();
+
+  if (__builtin_memmove (buf3, "ABCDEF", 6) != (char *) buf1
+      || memcmp (buf1, "ABCDEFghijklmnopq\0", 19))
+    abort ();
+
+  if (__builtin_memmove (buf3, "a", 1) != (char *) buf1
+      || memcmp (buf1, "aBCDEFghijklmnopq\0", 19))
+    abort ();
+
+  if (memmove ((char *) buf3 + 2, "bcd" + ++i, 2) != (char *) buf1 + 2
+      || memcmp (buf1, "aBcdEFghijklmnopq\0", 19)
+      || i != 1)
+    abort ();
+
+  /* These should probably be handled by move_by_pieces on most arches.  */
+  if (memmove ((char *) buf3 + 4, buf5, 6) != (char *) buf1 + 4
+      || memcmp (buf1, "aBcdRSTUVWklmnopq\0", 19))
+    abort ();
+
+  if (__builtin_memmove ((char *) buf1 + ++i + 8, (char *) buf5 + 1, 1)
+      != (char *) buf1 + 10
+      || memcmp (buf1, "aBcdRSTUVWSlmnopq\0", 19)
+      || i != 2)
+    abort ();
+
+  if (memmove ((char *) buf3 + 14, buf6, 2) != (char *) buf1 + 14
+      || memcmp (buf1, "aBcdRSTUVWSlmnrsq\0", 19))
+    abort ();
+
+  if (memmove (buf3, buf5, 8) != (char *) buf1
+      || memcmp (buf1, "RSTUVWXYVWSlmnrsq\0", 19))
+    abort ();
+
+  if (memmove (buf3, buf5, 17) != (char *) buf1
+      || memcmp (buf1, "RSTUVWXYZ01234567\0", 19))
+    abort ();
+
+  __builtin_memmove (buf3, "aBcdEFghijklmnopq\0", 19);
+
+  /* These should be handled either by movmemendM or memmove
+     call.  */
+
+  /* buf3 points to an unknown object, so __memmove_chk should not be done.  */
+  if (memmove ((char *) buf3 + 4, buf5, n + 6) != (char *) buf1 + 4
+      || memcmp (buf1, "aBcdRSTUVWklmnopq\0", 19))
+    abort ();
+
+  /* This call has side-effects in dst, therefore no checking.  */
+  if (__builtin___memmove_chk ((char *) buf1 + ++i + 8, (char *) buf5 + 1,
+			       n + 1, os ((char *) buf1 + ++i + 8))
+      != (char *) buf1 + 11
+      || memcmp (buf1, "aBcdRSTUVWkSmnopq\0", 19)
+      || i != 3)
+    abort ();
+
+  if (memmove ((char *) buf3 + 14, buf6, n + 2) != (char *) buf1 + 14
+      || memcmp (buf1, "aBcdRSTUVWkSmnrsq\0", 19))
+    abort ();
+
+  i = 1;
+
+  /* These might be handled by store_by_pieces.  */
+  if (memmove (buf2, "ABCDEFGHI", 9) != buf2
+      || memcmp (buf2, "ABCDEFGHI\0", 11))
+    abort ();
+
+  if (memmove (buf2, "abcdefghijklmnopq", 17) != buf2
+      || memcmp (buf2, "abcdefghijklmnopq\0", 19))
+    abort ();
+
+  if (__builtin_memmove (buf4, "ABCDEF", 6) != buf2
+      || memcmp (buf2, "ABCDEFghijklmnopq\0", 19))
+    abort ();
+
+  if (__builtin_memmove (buf4, "a", 1) != buf2
+      || memcmp (buf2, "aBCDEFghijklmnopq\0", 19))
+    abort ();
+
+  if (memmove (buf4 + 2, "bcd" + i++, 2) != buf2 + 2
+      || memcmp (buf2, "aBcdEFghijklmnopq\0", 19)
+      || i != 2)
+    abort ();
+
+  /* These might be handled by move_by_pieces.  */
+  if (memmove (buf4 + 4, buf7, 6) != buf2 + 4
+      || memcmp (buf2, "aBcdRSTUVWklmnopq\0", 19))
+    abort ();
+
+  /* Side effect.  */
+  if (__builtin___memmove_chk (buf2 + i++ + 8, buf7 + 1, 1,
+			       os (buf2 + i++ + 8))
+      != buf2 + 10
+      || memcmp (buf2, "aBcdRSTUVWSlmnopq\0", 19)
+      || i != 3)
+    abort ();
+
+  if (memmove (buf4 + 14, buf6, 2) != buf2 + 14
+      || memcmp (buf2, "aBcdRSTUVWSlmnrsq\0", 19))
+    abort ();
+
+  __builtin_memmove (buf4, "aBcdEFghijklmnopq\0", 19);
+
+  /* These should be handled either by movmemendM or memmove
+     call.  */
+  if (memmove (buf4 + 4, buf7, n + 6) != buf2 + 4
+      || memcmp (buf2, "aBcdRSTUVWklmnopq\0", 19))
+    abort ();
+
+  /* Side effect.  */
+  if (__builtin___memmove_chk (buf2 + i++ + 8, buf7 + 1, n + 1,
+			       os (buf2 + i++ + 8))
+      != buf2 + 11
+      || memcmp (buf2, "aBcdRSTUVWkSmnopq\0", 19)
+      || i != 4)
+    abort ();
+
+  if (memmove (buf4 + 14, buf6, n + 2) != buf2 + 14
+      || memcmp (buf2, "aBcdRSTUVWkSmnrsq\0", 19))
+    abort ();
+
+  if (chk_calls)
+    abort ();
+}
+
+void
+__attribute__((noinline))
+test2 (void)
+{
+  long *x;
+  char *y;
+  int z;
+  __builtin_memmove (buf5, "RSTUVWXYZ0123456789", 20);
+  __builtin_memmove (buf7, "RSTUVWXYZ0123456789", 20);
+ __asm ("" : "=r" (x) : "0" (buf1));
+ __asm ("" : "=r" (y) : "0" (buf2));
+ __asm ("" : "=r" (z) : "0" (0));
+  test2_sub (x, y, "rstuvwxyz", z);
+}
+
+static const struct foo
+{
+  char *s;
+  double d;
+  long l;
+} foo[] =
+{
+  { "hello world1", 3.14159, 101L },
+  { "hello world2", 3.14159, 102L },
+  { "hello world3", 3.14159, 103L },
+  { "hello world4", 3.14159, 104L },
+  { "hello world5", 3.14159, 105L },
+  { "hello world6", 3.14159, 106L }
+};
+
+static const struct bar
+{
+  char *s;
+  const struct foo f[3];
+} bar[] =
+{
+  {
+    "hello world10",
+    {
+      { "hello1", 3.14159, 201L },
+      { "hello2", 3.14159, 202L },
+      { "hello3", 3.14159, 203L },
+    }
+  },
+  {
+    "hello world11",
+    {
+      { "hello4", 3.14159, 204L },
+      { "hello5", 3.14159, 205L },
+      { "hello6", 3.14159, 206L },
+    }
+  }
+};
+
+static const int baz[] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 0 };
+
+void
+__attribute__((noinline))
+test3 (void)
+{
+  const char *s;
+  struct foo f1[sizeof foo/sizeof*foo];
+  struct bar b1[sizeof bar/sizeof*bar];
+  int bz[sizeof baz/sizeof*baz];
+
+  /* All the memmove/__builtin_memmove calls in this routine have fixed
+     length.  */
+  chk_calls = 0;
+
+  /* All the *memmove calls below have src in read-only memory, so all
+     of them should be optimized into memcpy.  */
+  memmove_disallowed = 1;
+  if (memmove (f1, foo, sizeof (foo)) != f1 || memcmp (f1, foo, sizeof (foo)))
+    abort ();
+  if (memmove (b1, bar, sizeof (bar)) != b1 || memcmp (b1, bar, sizeof (bar)))
+    abort ();
+  memmove (bz, baz, sizeof (baz));
+  if (memcmp (bz, baz, sizeof (baz)))
+    abort ();
+
+  if (memmove (p, "abcde", 6) != p || memcmp (p, "abcde", 6))
+    abort ();
+  s = s1;
+  if (memmove (p + 2, ++s, 0) != p + 2 || memcmp (p, "abcde", 6) || s != s1 + 1)
+    abort ();
+  if (__builtin_memmove (p + 3, "", 1) != p + 3 || memcmp (p, "abc\0e", 6))
+    abort ();
+  memmove (p + 2, "fghijk", 4);
+  if (memcmp (p, "abfghi", 7))
+    abort ();
+  s = s1 + 1;
+  memmove (p + 1, s++, 0);
+  if (memcmp (p, "abfghi", 7) || s != s1 + 2)
+    abort ();
+  __builtin_memmove (p + 4, "ABCDE", 1);
+  if (memcmp (p, "abfgAi", 7))
+    abort ();
+
+  /* memmove with length 1 can be optimized into memcpy if it can be
+     expanded inline.  */
+  if (memmove (p + 2, p + 3, 1) != p + 2)
+    abort ();
+  if (memcmp (p, "abggAi", 7))
+    abort ();
+
+  if (chk_calls)
+    abort ();
+  memmove_disallowed = 0;
+}
+
+/* Test whether compile time checking is done where it should
+   and so is runtime object size checking.  */
+void
+__attribute__((noinline))
+test4 (void)
+{
+  struct A { char buf1[10]; char buf2[10]; } a;
+  char *r = l1 == 1 ? &a.buf1[5] : &a.buf2[4];
+  char buf3[20];
+  int i;
+  size_t l;
+
+  /* The following calls should do runtime checking
+     - length is not known, but destination is.  */
+  chk_calls = 0;
+  memmove (a.buf1 + 2, s3, l1);
+  memmove (r, s3, l1 + 1);
+  r = l1 == 1 ? __builtin_alloca (4) : &a.buf2[7];
+  memmove (r, s2, l1 + 2);
+  memmove (r + 2, s3, l1);
+  r = buf3;
+  for (i = 0; i < 4; ++i)
+    {
+      if (i == l1 - 1)
+	r = &a.buf1[1];
+      else if (i == l1)
+	r = &a.buf2[7];
+      else if (i == l1 + 1)
+	r = &buf3[5];
+      else if (i == l1 + 2)
+	r = &a.buf1[9];
+    }
+  memmove (r, s2, l1);
+  if (chk_calls != 5)
+    abort ();
+
+  /* Following have known destination and known length,
+     so if optimizing certainly shouldn't result in the checking
+     variants.  */
+  chk_calls = 0;
+  memmove (a.buf1 + 2, s3, 1);
+  memmove (r, s3, 2);
+  r = l1 == 1 ? __builtin_alloca (4) : &a.buf2[7];
+  memmove (r, s2, 3);
+  r = buf3;
+  l = 4;
+  for (i = 0; i < 4; ++i)
+    {
+      if (i == l1 - 1)
+	r = &a.buf1[1], l = 2;
+      else if (i == l1)
+	r = &a.buf2[7], l = 3;
+      else if (i == l1 + 1)
+	r = &buf3[5], l = 4;
+      else if (i == l1 + 2)
+	r = &a.buf1[9], l = 1;
+    }
+  memmove (r, s2, 1);
+  /* Here, l is known to be at most 4 and __builtin_object_size (&buf3[16], 0)
+     is 4, so this doesn't need runtime checking.  */
+  memmove (&buf3[16], s2, l);
+  if (chk_calls)
+    abort ();
+  chk_calls = 0;
+}
+
+/* Test whether runtime and/or compile time checking catches
+   buffer overflows.  */
+void
+__attribute__((noinline))
+test5 (void)
+{
+  struct A { char buf1[10]; char buf2[10]; } a;
+  char buf3[20];
+
+  chk_fail_allowed = 1;
+  /* Runtime checks.  */
+  if (__builtin_setjmp (chk_fail_buf) == 0)
+    {
+      memmove (&a.buf2[9], s2, l1 + 1);
+      abort ();
+    }
+  if (__builtin_setjmp (chk_fail_buf) == 0)
+    {
+      memmove (&a.buf2[7], s3, strlen (s3) + 1);
+      abort ();
+    }
+  /* This should be detectable at compile time already.  */
+  if (__builtin_setjmp (chk_fail_buf) == 0)
+    {
+      memmove (&buf3[19], "ab", 2);
+      abort ();
+    }
+  chk_fail_allowed = 0;
+}
+
+#ifndef MAX_OFFSET
+#define MAX_OFFSET (sizeof (long long))
+#endif
+
+#ifndef MAX_COPY
+#define MAX_COPY (10 * sizeof (long long))
+#endif
+
+#ifndef MAX_EXTRA
+#define MAX_EXTRA (sizeof (long long))
+#endif
+
+#define MAX_LENGTH (MAX_OFFSET + MAX_COPY + MAX_EXTRA)
+
+/* Use a sequence length that is not divisible by two, to make it more
+   likely to detect when words are mixed up.  */
+#define SEQUENCE_LENGTH 31
+
+static union {
+  char buf[MAX_LENGTH];
+  long long align_int;
+  long double align_fp;
+} u1, u2;
+
+void
+__attribute__((noinline))
+test6 (void)
+{
+  int off1, off2, len, i;
+  char *p, *q, c;
+
+  for (off1 = 0; off1 < MAX_OFFSET; off1++)
+    for (off2 = 0; off2 < MAX_OFFSET; off2++)
+      for (len = 1; len < MAX_COPY; len++)
+	{
+	  for (i = 0, c = 'A'; i < MAX_LENGTH; i++, c++)
+	    {
+	      u1.buf[i] = 'a';
+	      if (c >= 'A' + SEQUENCE_LENGTH)
+		c = 'A';
+	      u2.buf[i] = c;
+	    }
+
+	  p = memmove (u1.buf + off1, u2.buf + off2, len);
+	  if (p != u1.buf + off1)
+	    abort ();
+
+	  q = u1.buf;
+	  for (i = 0; i < off1; i++, q++)
+	    if (*q != 'a')
+	      abort ();
+
+	  for (i = 0, c = 'A' + off2; i < len; i++, q++, c++)
+	    {
+	      if (c >= 'A' + SEQUENCE_LENGTH)
+		c = 'A';
+	      if (*q != c)
+		abort ();
+	    }
+
+	  for (i = 0; i < MAX_EXTRA; i++, q++)
+	    if (*q != 'a')
+	      abort ();
+	}
+}
+
+#define TESTSIZE 80
+
+char srcb[TESTSIZE] __attribute__ ((aligned));
+char dstb[TESTSIZE] __attribute__ ((aligned));
+
+void
+__attribute__((noinline))
+check (char *test, char *match, int n)
+{
+  if (memcmp (test, match, n))
+    abort ();
+}
+
+#define TN(n) \
+{ memset (dstb, 0, n); memmove (dstb, srcb, n); check (dstb, srcb, n); }
+#define T(n) \
+TN (n) \
+TN ((n) + 1) \
+TN ((n) + 2) \
+TN ((n) + 3)
+
+void
+__attribute__((noinline))
+test7 (void)
+{
+  int i;
+
+  chk_calls = 0;
+
+  for (i = 0; i < sizeof (srcb); ++i)
+      srcb[i] = 'a' + i % 26;
+
+  T (0);
+  T (4);
+  T (8);
+  T (12);
+  T (16);
+  T (20);
+  T (24);
+  T (28);
+  T (32);
+  T (36);
+  T (40);
+  T (44);
+  T (48);
+  T (52);
+  T (56);
+  T (60);
+  T (64);
+  T (68);
+  T (72);
+  T (76);
+
+  /* All memmove calls in this routine have constant arguments.  */
+  if (chk_calls)
+    abort ();
+}
+
+void
+main_test (void)
+{
+#ifndef __OPTIMIZE__
+  /* Object size checking is only intended for -O[s123].  */
+  return;
+#endif
+  __asm ("" : "=r" (l1) : "0" (l1));
+  test1 ();
+  test2 ();
+  __builtin_memset (p, '\0', sizeof (p));
+  test3 ();
+  test4 ();
+  test5 ();
+  test6 ();
+  test7 ();
+}