Re: [PATCH v3] Add tests for the long double version of ecvt and fcvt

["Gabriel F. T. Gomes" <gabriel at inconstante dot eti dot br>]

On Wed, 12 Dec 2018, Florian Weimer wrote:

>* Gabriel F. T. Gomes:
>
>> On Tue, 11 Dec 2018, Florian Weimer wrote:
>>  
>>>Do you expect NAME to be customizable in the future?  
>>
>> No more than what is already implemented, i.e. when FLOAT is double,
>> errors print the function name as ecvt or fcvt, but when FLOAT is long
>> double, you get qecvt, qfcvt, etc.  For the long double transition on
>> powerpc64le, I do *not* plan to make this test print the internal name of
>> the functions, e.g. __qecvtieee128.  
>
>Can you move NAME into the template file, then?

Ah, yes, I can.  Does the following patch look better?

From dd8545e8b3a9da1977777591ec356fca52bebe99 Mon Sep 17 00:00:00 2001
From: "Gabriel F. T. Gomes" <gabriel@inconstante.eti.br>
Date: Tue, 30 Oct 2018 15:04:18 -0300
Subject: [PATCH v4] Add tests for the long double version of ecvt and fcvt

Changes since v3:

  - Added description for each of the macros used for templatization.
  - Moved NAME into the template file.

Changes since v2:

  - Remove the macro L and its uses.  Let the compiler handle the
    conversion.
  - Define EXTRA_ECVT_TESTS to contain the two additional tests for
    double and DBL_MANT_DIG == 53.  Otherwise, define it to nothing.

Changes since v1:

  - Rebased on top of the new patch that adapts the test to the new
    testing framework.

-- 8< --
The test case misc/tst-efgcvt.c only tests the double variants of the
Old-fashioned System V number-to-string functions: ecvt, fcvt, and their
re-entrant counterparts.  With a few macros, the code can be reused for
the long double versions of these functions.  A future patch will reuse
it again for IEEE long double on powerpc64le.

Tested for powerpc and powerpc64le.

	* misc/Makefile (tests): Remove tst-efgcvt.  Add tst-dbl-efgcvt
	and tst-ldbl-efgcvt.
	* misc/tst-efgcvt.c: Renamed to misc/tst-efgcvt-template.c.
	* misc/tst-efgcvt-template.c: Renamed from misc/tst-efgcvt.c.
	(struct testcase, efcvt_func, efcvt_r_func, ecvt_tests)
	(fcvt_tests, output_error, output_r_error, do_test): Use the
	macros defined in tst-dbl-efgcvt.c and tst-ldbl-efgcvt.c to:
	select the type of floating-point variables and arguments; to
	produce extra tests for double and conversion specifiers for
	printf; and to set the names of called functions.
	* misc/tst-dbl-efgcvt.c: New file that defines the macros used
	in tst-efgcvt-template.c.
	* misc/tst-ldbl-efgcvt.c: Likewise.
---
 misc/Makefile                                |  4 +-
 misc/tst-dbl-efgcvt.c                        | 36 ++++++++++++++
 misc/{tst-efgcvt.c => tst-efgcvt-template.c} | 72 ++++++++++++++++++----------
 misc/tst-ldbl-efgcvt.c                       | 28 +++++++++++
 4 files changed, 112 insertions(+), 28 deletions(-)
 create mode 100644 misc/tst-dbl-efgcvt.c
 rename misc/{tst-efgcvt.c => tst-efgcvt-template.c} (68%)
 create mode 100644 misc/tst-ldbl-efgcvt.c

diff --git a/misc/Makefile b/misc/Makefile
index c2c9994d17..cf3726b637 100644
--- a/misc/Makefile
+++ b/misc/Makefile
@@ -80,12 +80,12 @@ aux := init-misc
 install-lib := libg.a
 gpl2lgpl := error.c error.h
 
-tests := tst-dirname tst-tsearch tst-fdset tst-efgcvt tst-mntent tst-hsearch \
+tests := tst-dirname tst-tsearch tst-fdset tst-mntent tst-hsearch \
 	 tst-error1 tst-pselect tst-insremque tst-mntent2 bug-hsearch1 \
 	 tst-mntent-blank-corrupt tst-mntent-blank-passno bug18240 \
 	 tst-preadvwritev tst-preadvwritev64 tst-makedev tst-empty \
 	 tst-preadvwritev2 tst-preadvwritev64v2 tst-warn-wide \
-	 tst-ldbl-warn tst-ldbl-error
+	 tst-ldbl-warn tst-ldbl-error tst-dbl-efgcvt tst-ldbl-efgcvt
 
 # Tests which need libdl.
 ifeq (yes,$(build-shared))
diff --git a/misc/tst-dbl-efgcvt.c b/misc/tst-dbl-efgcvt.c
new file mode 100644
index 0000000000..78accaddc7
--- /dev/null
+++ b/misc/tst-dbl-efgcvt.c
@@ -0,0 +1,36 @@
+/* Tests for *cvt function, double version.
+   Copyright (C) 1998-2018 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>
+
+#define ECVT ecvt
+#define FCVT fcvt
+#define ECVT_R ecvt_r
+#define FCVT_R fcvt_r
+#define FLOAT double
+#define PRINTF_CONVERSION "%f"
+
+#if DBL_MANT_DIG == 53
+# define EXTRA_ECVT_TESTS \
+  { 0x1p-1074, 3, -323, "494" }, \
+  { -0x1p-1074, 3, -323, "494" },
+#else
+# define EXTRA_ECVT_TESTS
+#endif
+
+#include <tst-efgcvt-template.c>
diff --git a/misc/tst-efgcvt.c b/misc/tst-efgcvt-template.c
similarity index 68%
rename from misc/tst-efgcvt.c
rename to misc/tst-efgcvt-template.c
index 5a598f49e3..cba964a842 100644
--- a/misc/tst-efgcvt.c
+++ b/misc/tst-efgcvt-template.c
@@ -15,11 +15,29 @@
    License along with the GNU C Library; if not, see
    <http://www.gnu.org/licenses/>.  */
 
+/* This template provides testing for the *cvt family of functions,
+   which deal with double or long double types.  In order to use the
+   template, the following macros must be defined before inclusion of
+   this template:
+
+     FLOAT: The floating-point type, i.e. double or long double.
+
+     ECVT: Appropriate *ecvt function for FLOAT, i.e. ecvt or qecvt.
+     FCVT: Likewise for *fcvt, i.e. fcvt or qfcvt.
+     ECVT_R: Likewise for *ecvt_r, i.e. ecvt_r or qecvt_r.
+     FCVT_R: Likewise for *fcvt_r, i.e. fcvt_r or qfcvt_r.
+
+     PRINTF_CONVERSION: The appropriate printf conversion specifier with
+     length modifier for FLOAT, i.e. "%f" or "%Lf".
+
+     EXTRA_ECVT_TESTS: Additional tests for the ecvt or qecvt function
+     that are only relevant to a particular floating-point type and
+     cannot be represented generically.  */
+
 #ifndef _GNU_SOURCE
 # define _GNU_SOURCE	1
 #endif
 
-#include <float.h>
 #include <math.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -27,17 +45,20 @@
 
 #include <support/check.h>
 
+#define NAME(x) NAMEX(x)
+#define NAMEX(x) #x
+
 typedef struct
 {
-  double value;
+  FLOAT value;
   int ndigit;
   int decpt;
   char result[30];
 } testcase;
 
-typedef char * ((*efcvt_func) (double, int, int *, int *));
+typedef char * ((*efcvt_func) (FLOAT, int, int *, int *));
 
-typedef int ((*efcvt_r_func) (double, int, int *, int *, char *, size_t));
+typedef int ((*efcvt_r_func) (FLOAT, int, int *, int *, char *, size_t));
 
 
 static testcase ecvt_tests[] =
@@ -59,10 +80,7 @@ static testcase ecvt_tests[] =
   { 123.01, -4, 3, "" },
   { 126.71, -4, 3, "" },
   { 0.0, 4, 1, "0000" },
-#if DBL_MANT_DIG == 53
-  { 0x1p-1074, 3, -323, "494" },
-  { -0x1p-1074, 3, -323, "494" },
-#endif
+  EXTRA_ECVT_TESTS
   /* -1.0 is end marker.  */
   { -1.0, 0, 0, "" }
 };
@@ -91,11 +109,12 @@ static testcase fcvt_tests[] =
 };
 
 static void
-output_error (const char *name, double value, int ndigit,
+output_error (const char *name, FLOAT value, int ndigit,
 	      const char *exp_p, int exp_decpt, int exp_sign,
 	      char *res_p, int res_decpt, int res_sign)
 {
-  printf ("%s returned wrong result for value: %f, ndigits: %d\n",
+  printf ("%s returned wrong result for value: " PRINTF_CONVERSION
+	  ", ndigits: %d\n",
 	  name, value, ndigit);
   printf ("Result was p: \"%s\", decpt: %d, sign: %d\n",
 	  res_p, res_decpt, res_sign);
@@ -106,11 +125,12 @@ output_error (const char *name, double value, int ndigit,
 
 
 static void
-output_r_error (const char *name, double value, int ndigit,
+output_r_error (const char *name, FLOAT value, int ndigit,
 		const char *exp_p, int exp_decpt, int exp_sign, int exp_return,
 		char *res_p, int res_decpt, int res_sign, int res_return)
 {
-  printf ("%s returned wrong result for value: %f, ndigits: %d\n",
+  printf ("%s returned wrong result for value: " PRINTF_CONVERSION
+	  ", ndigits: %d\n",
 	  name, value, ndigit);
   printf ("Result was buf: \"%s\", decpt: %d, sign: %d return value: %d\n",
 	  res_p, res_decpt, res_sign, res_return);
@@ -171,30 +191,30 @@ special (void)
   char *p;
   char buf [1024];
 
-  p = ecvt (NAN, 10, &decpt, &sign);
+  p = ECVT (NAN, 10, &decpt, &sign);
   if (sign != 0 || strcmp (p, "nan") != 0)
-    output_error ("ecvt", NAN, 10, "nan", 0, 0, p, decpt, sign);
+    output_error (NAME (ECVT), NAN, 10, "nan", 0, 0, p, decpt, sign);
 
-  p = ecvt (INFINITY, 10, &decpt, &sign);
+  p = ECVT (INFINITY, 10, &decpt, &sign);
   if (sign != 0 || strcmp (p, "inf") != 0)
-    output_error ("ecvt", INFINITY, 10, "inf", 0, 0, p, decpt, sign);
+    output_error (NAME (ECVT), INFINITY, 10, "inf", 0, 0, p, decpt, sign);
 
   /* Simply make sure these calls with large NDIGITs don't crash.  */
-  (void) ecvt (123.456, 10000, &decpt, &sign);
-  (void) fcvt (123.456, 10000, &decpt, &sign);
+  (void) ECVT (123.456, 10000, &decpt, &sign);
+  (void) FCVT (123.456, 10000, &decpt, &sign);
 
   /* Some tests for the reentrant functions.  */
   /* Use a too small buffer.  */
-  res = ecvt_r (123.456, 10, &decpt, &sign, buf, 1);
+  res = ECVT_R (123.456, 10, &decpt, &sign, buf, 1);
   if (res == 0)
     {
-      printf ("ecvt_r with a too small buffer was succesful.\n");
+      printf (NAME (ECVT_R) " with a too small buffer was succesful.\n");
       support_record_failure ();
     }
-  res = fcvt_r (123.456, 10, &decpt, &sign, buf, 1);
+  res = FCVT_R (123.456, 10, &decpt, &sign, buf, 1);
   if (res == 0)
     {
-      printf ("fcvt_r with a too small buffer was succesful.\n");
+      printf (NAME (FCVT_R) " with a too small buffer was succesful.\n");
       support_record_failure ();
     }
 }
@@ -203,10 +223,10 @@ special (void)
 static int
 do_test (void)
 {
-  test (ecvt_tests, ecvt, "ecvt");
-  test (fcvt_tests, fcvt, "fcvt");
-  test_r (ecvt_tests, ecvt_r, "ecvt_r");
-  test_r (fcvt_tests, fcvt_r, "fcvt_r");
+  test (ecvt_tests, ECVT, NAME (ECVT));
+  test (fcvt_tests, FCVT, NAME (FCVT));
+  test_r (ecvt_tests, ECVT_R, NAME (ECVT_R));
+  test_r (fcvt_tests, FCVT_R, NAME (FCVT_R));
   special ();
 
   return 0;
diff --git a/misc/tst-ldbl-efgcvt.c b/misc/tst-ldbl-efgcvt.c
new file mode 100644
index 0000000000..35d24ae3ad
--- /dev/null
+++ b/misc/tst-ldbl-efgcvt.c
@@ -0,0 +1,28 @@
+/* Tests for *cvt function, long double version.
+   Copyright (C) 1998-2018 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/>.  */
+
+#define ECVT qecvt
+#define FCVT qfcvt
+#define ECVT_R qecvt_r
+#define FCVT_R qfcvt_r
+#define FLOAT long double
+#define PRINTF_CONVERSION "%Lf"
+
+#define EXTRA_ECVT_TESTS
+
+#include <tst-efgcvt-template.c>
-- 
2.14.5