Fix ldbl-128ibm fmodl handling of equal arguments with low part zero (bug 19602).

The ldbl-128ibm implementation of fmodl has logic to detect when the
first argument has absolute value less than or equal to the second.
This logic is only correct for nonzero low parts; if the high parts
are equal and the low parts are zero, then the signs of the low parts
(which have no semantic effect on the value of the long double number)
can result in equal values being wrongly treated as unequal, and an
incorrect result being returned from fmodl.  This patch fixes this by
checking for the case of zero low parts.

Although this does show up in tests from libm-test.inc (both tests of
fmodl, and, indirectly, of remainderl / dreml), the dependence on
non-semantic zero low parts means that test shouldn't be expected to
reproduce it reliably; thus, this patch adds a standalone test that
sets up affected values using unions.

Tested for powerpc.

	[BZ #19602]
	* sysdeps/ieee754/ldbl-128ibm/e_fmodl.c (__ieee754_fmodl): Handle
	equal high parts and both low parts zero specially.
	* sysdeps/ieee754/ldbl-128ibm/test-fmodl-ldbl-128ibm.c: New test.
	* sysdeps/ieee754/ldbl-128ibm/Makefile [$(subdir) = math] (tests):
	Add test-fmodl-ldbl-128ibm.

3 files changed, 88 insertions, 0 deletions

diff --git a/sysdeps/ieee754/ldbl-128ibm/Makefile b/sysdeps/ieee754/ldbl-128ibm/Makefile
index 5591814824..b633172721 100644
--- a/sysdeps/ieee754/ldbl-128ibm/Makefile
+++ b/sysdeps/ieee754/ldbl-128ibm/Makefile
@@ -8,3 +8,7 @@ ifeq ($(subdir),stdlib)
 tests += tst-strtold-ldbl-128ibm
 $(objpfx)tst-strtold-ldbl-128ibm: $(libm)
 endif
+
+ifeq ($(subdir),math)
+tests += test-fmodl-ldbl-128ibm
+endif
diff --git a/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c b/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c
index d756e3e2a6..5284fd0fd5 100644
--- a/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c
@@ -55,6 +55,13 @@ __ieee754_fmodl (long double x, long double y)
 	      return x;
 	    /* At this point the absolute value of the high doubles of
 	       x and y must be equal.  */
+	    if ((lx & 0x7fffffffffffffffLL) == 0
+		&& (ly & 0x7fffffffffffffffLL) == 0)
+	      /* Both low parts are zero.  The result should be an
+		 appropriately signed zero, but the subsequent logic
+		 could treat them as unequal, depending on the signs
+		 of the low parts.  */
+	      return Zero[(uint64_t) sx >> 63];
 	    /* If the low double of y is the same sign as the high
 	       double of y (ie. the low double increases |y|)...  */
 	    if (((ly ^ sy) & 0x8000000000000000LL) == 0
diff --git a/sysdeps/ieee754/ldbl-128ibm/test-fmodl-ldbl-128ibm.c b/sysdeps/ieee754/ldbl-128ibm/test-fmodl-ldbl-128ibm.c
new file mode 100644
index 0000000000..3717c90aba
--- /dev/null
+++ b/sysdeps/ieee754/ldbl-128ibm/test-fmodl-ldbl-128ibm.c
@@ -0,0 +1,77 @@
+/* Test for ldbl-128ibm fmodl handling of equal values (bug 19602).
+   Copyright (C) 2016 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <float.h>
+#include <math.h>
+#include <stdio.h>
+
+union u
+{
+  long double ld;
+  double d[2];
+};
+
+volatile union u p1 = { .d = { DBL_MIN, 0.0 } };
+volatile union u p2 = { .d = { DBL_MIN, -0.0 } };
+volatile union u m1 = { .d = { -DBL_MIN, 0.0 } };
+volatile union u m2 = { .d = { -DBL_MIN, -0.0 } };
+
+static int
+test_fmodl (const char *s, long double x, long double y, long double expected)
+{
+  volatile long double r;
+  r = fmodl (x, y);
+  if (r != expected || copysignl (1.0, r) != copysignl (1.0, expected))
+    {
+      printf ("FAIL: fmodl (%s)\n", s);
+      return 1;
+    }
+  else
+    {
+      printf ("PASS: fmodl (%s)\n", s);
+      return 0;
+    }
+}
+
+#define TEST_FMODL(a, b, e) test_fmodl (#a ", " #b, a, b, e)
+
+static int
+do_test (void)
+{
+  int result = 0;
+  result |= TEST_FMODL (p1.ld, p1.ld, 0.0L);
+  result |= TEST_FMODL (p1.ld, p2.ld, 0.0L);
+  result |= TEST_FMODL (p1.ld, m1.ld, 0.0L);
+  result |= TEST_FMODL (p1.ld, m2.ld, 0.0L);
+  result |= TEST_FMODL (p2.ld, p1.ld, 0.0L);
+  result |= TEST_FMODL (p2.ld, p2.ld, 0.0L);
+  result |= TEST_FMODL (p2.ld, m1.ld, 0.0L);
+  result |= TEST_FMODL (p2.ld, m2.ld, 0.0L);
+  result |= TEST_FMODL (m1.ld, p1.ld, -0.0L);
+  result |= TEST_FMODL (m1.ld, p2.ld, -0.0L);
+  result |= TEST_FMODL (m1.ld, m1.ld, -0.0L);
+  result |= TEST_FMODL (m1.ld, m2.ld, -0.0L);
+  result |= TEST_FMODL (m2.ld, p1.ld, -0.0L);
+  result |= TEST_FMODL (m2.ld, p2.ld, -0.0L);
+  result |= TEST_FMODL (m2.ld, m1.ld, -0.0L);
+  result |= TEST_FMODL (m2.ld, m2.ld, -0.0L);
+  return result;
+}
+
+#define TEST_FUNCTION do_test ()
+#include "../../../test-skeleton.c"