From patchwork Wed Dec 12 16:49:03 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Gabriel F. T. Gomes" X-Patchwork-Id: 30650 X-Patchwork-Delegate: fweimer@redhat.com Received: (qmail 32035 invoked by alias); 12 Dec 2018 16:49:43 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 32015 invoked by uid 89); 12 Dec 2018 16:49:42 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-25.9 required=5.0 tests=BAYES_00, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3, KAM_LAZY_DOMAIN_SECURITY, KAM_SHORT, RCVD_IN_DNSWL_NONE autolearn=ham version=3.3.2 spammy=H*r:172.16.2 X-HELO: mo20.mail-out.ovh.net Date: Wed, 12 Dec 2018 14:49:03 -0200 From: "Gabriel F. T. Gomes" To: Florian Weimer CC: Subject: Re: [PATCH v3] Add tests for the long double version of ecvt and fcvt Message-ID: <20181212144903.6526bd46@tereshkova> In-Reply-To: <87imzyri8q.fsf@oldenburg2.str.redhat.com> References: <20181108222916.27414-1-gabriel@inconstante.eti.br> <87bm5sulmt.fsf@oldenburg2.str.redhat.com> <20181212110741.40805913@tereshkova> <87imzyri8q.fsf@oldenburg2.str.redhat.com> MIME-Version: 1.0 X-Ovh-Tracer-Id: 4331618420543770260 X-VR-SPAMSTATE: OK X-VR-SPAMSCORE: -100 X-VR-SPAMCAUSE: gggruggvucftvghtrhhoucdtuddrgedtkedrudehtddgtdduucetufdoteggodetrfdotffvucfrrhhofhhilhgvmecuqfggjfdpvefjgfevmfevgfenuceurghilhhouhhtmecuhedttdenucesvcftvggtihhpihgvnhhtshculddquddttddm 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" 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 + . */ + +#include + +#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 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 . */ +/* 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 #include #include #include @@ -27,17 +45,20 @@ #include +#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 + . */ + +#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