--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,10 @@
 2012-03-06  Bruno Haible  <bruno@clisp.org>
 
+	Tests for module 'expm1'.
+	* modules/expm1-tests: New file.
+	* tests/test-expm1.c: New file.
+	* tests/test-expm1.h: New file.
+
 	New module 'expm1'.
 	* lib/math.in.h (expm1): New declaration.
 	* lib/expm1.c: New file.

new file mode 100644
--- /dev/null
+++ b/modules/expm1-tests
@@ -0,0 +1,16 @@
+Files:
+tests/test-expm1.c
+tests/test-expm1.h
+tests/signature.h
+tests/macros.h
+tests/randomd.c
+
+Depends-on:
+
+configure.ac:
+
+Makefile.am:
+TESTS += test-expm1
+check_PROGRAMS += test-expm1
+test_expm1_SOURCES = test-expm1.c randomd.c
+test_expm1_LDADD = $(LDADD) @EXPM1_LIBM@

new file mode 100644
--- /dev/null
+++ b/tests/test-expm1.c
@@ -0,0 +1,49 @@
+/* Test of expm1() function.
+   Copyright (C) 2010-2012 Free Software Foundation, Inc.
+
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program 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 General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+/* Written by Bruno Haible <bruno@clisp.org>, 2010.  */
+
+#include <config.h>
+
+#include <math.h>
+
+#include "signature.h"
+SIGNATURE_CHECK (expm1, double, (double));
+
+#include <float.h>
+
+#include "macros.h"
+
+#define DOUBLE double
+#define L_(literal) literal
+#define MANT_DIG DBL_MANT_DIG
+#define MIN_EXP DBL_MIN_EXP
+#define EXPM1 expm1
+#define RANDOM randomd
+#include "test-expm1.h"
+
+int
+main ()
+{
+  /* A particular value.  */
+  x = 0.6;
+  y = expm1 (x);
+  ASSERT (y >= 0.822118800 && y <= 0.822118801);
+
+  test_function ();
+
+  return 0;
+}

new file mode 100644
--- /dev/null
+++ b/tests/test-expm1.h
@@ -0,0 +1,115 @@
+/* Test of expm1*() function family.
+   Copyright (C) 2012 Free Software Foundation, Inc.
+
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program 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 General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+static void
+test_function (void)
+{
+  int i;
+  int j;
+  const DOUBLE TWO_MANT_DIG =
+    /* Assume MANT_DIG <= 5 * 31.
+       Use the identity
+         n = floor(n/5) + floor((n+1)/5) + ... + floor((n+4)/5).  */
+    (DOUBLE) (1U << ((MANT_DIG - 1) / 5))
+    * (DOUBLE) (1U << ((MANT_DIG - 1 + 1) / 5))
+    * (DOUBLE) (1U << ((MANT_DIG - 1 + 2) / 5))
+    * (DOUBLE) (1U << ((MANT_DIG - 1 + 3) / 5))
+    * (DOUBLE) (1U << ((MANT_DIG - 1 + 4) / 5));
+
+  /* Arguments near zero.  */
+  {
+    DOUBLE x;
+
+    for (i = -1, x = L_(0.5); i >= MIN_EXP; i--, x *= L_(0.5))
+      {
+        DOUBLE y;
+
+        y = EXPM1 (x);
+#ifdef __MINGW32__ /* mingw returns a value that is slightly too small.  */
+        ASSERT (y >= x * (L_(1.0) - L_(1.0) / TWO_MANT_DIG));
+#else
+        ASSERT (y >= x);
+#endif
+        ASSERT (y < L_(1.5) * x);
+
+        y = EXPM1 (- x);
+        ASSERT (y >= - x);
+        ASSERT (y < - L_(0.5) * x);
+      }
+  }
+
+  /* Randomized tests.  */
+  {
+    /* Error bound, in ulps.  */
+    const DOUBLE err_bound =
+      (sizeof (DOUBLE) == sizeof (long double) ?
+#if defined __i386__ && defined __FreeBSD__
+       /* On FreeBSD/x86 6.4, the 'long double' type really has only 53 bits of
+          precision in the compiler but 64 bits of precision at runtime.  See
+          <http://lists.gnu.org/archive/html/bug-gnulib/2008-07/msg00063.html>.
+          The compiler has truncated all 'long double' literals in expl.c to
+          53 bits of precision.  */
+       L_(1024.0)
+#else
+       L_(5.0)
+#endif
+       : L_(5.0));
+
+    for (i = 0; i < SIZEOF (RANDOM); i++)
+      {
+        DOUBLE x = L_(2.0) * RANDOM[i]; /* 0.0 <= x <= 2.0 */
+        DOUBLE y = EXPM1 (x);
+        DOUBLE z = EXPM1 (- x);
+        DOUBLE err = y + (1 + y) * z;
+        ASSERT (y >= L_(0.0));
+        ASSERT (z <= L_(0.0));
+        ASSERT (err > - err_bound / TWO_MANT_DIG
+                && err < err_bound / TWO_MANT_DIG);
+      }
+  }
+
+  {
+    /* Error bound, in ulps.  */
+    const DOUBLE err_bound =
+      (sizeof (DOUBLE) == sizeof (long double) ?
+#if defined __i386__ && defined __FreeBSD__
+       L_(1536.0)
+#else
+       L_(11.0)
+#endif
+       : L_(9.0));
+
+    for (i = 0; i < SIZEOF (RANDOM) / 5; i++)
+      for (j = 0; j < SIZEOF (RANDOM) / 5; j++)
+        {
+          DOUBLE x = L_(2.0) * RANDOM[i] - L_(1.0); /* -1.0 <= x <= 1.0 */
+          DOUBLE y = L_(2.0) * RANDOM[j] - L_(1.0); /* -1.0 <= y <= 1.0 */
+          DOUBLE z = - x - y;
+          /* Approximately  x + y + z = 0.  */
+          {
+            DOUBLE a = EXPM1 (x);
+            DOUBLE b = EXPM1 (y);
+            DOUBLE c = EXPM1 (z);
+            DOUBLE err = a + (1 + a) * (b + (1 + b) * c);
+            ASSERT (err > - err_bound / TWO_MANT_DIG
+                    && err < err_bound / TWO_MANT_DIG);
+          }
+        }
+  }
+}
+
+volatile DOUBLE x;
+DOUBLE y;