From patchwork Sat Oct 13 21:16:41 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Albert ARIBAUD X-Patchwork-Id: 29734 Received: (qmail 76589 invoked by alias); 13 Oct 2018 21:16:49 -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 76566 invoked by uid 89); 13 Oct 2018 21:16:48 -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, RCVD_IN_DNSWL_NONE autolearn=ham version=3.3.2 spammy=Wrap, pairs, Y2038, 64-bit-time X-HELO: smtp3-g21.free.fr From: "Albert ARIBAUD (3ADEV)" To: libc-alpha@sourceware.org Cc: "Albert ARIBAUD (3ADEV)" Subject: [PATCH v3] Y2038: make __difftime compatible with 64-bit time Date: Sat, 13 Oct 2018 23:16:41 +0200 Message-Id: <20181013211641.5215-1-albert.aribaud@3adev.fr> Provide a 64-bit-time version of __difftime (but do not assume __time64_t is a signed int so that Gnulib can reuse the code) and make the 32-bit version a wrapper of it. Current difftime expects two time_t arguments and returns a double. To preserve source-code compatibility, its 64-bit-time equivalent expects two __time64_t arguments but still returns a double. This patch was tested by running 'make check' on branch master then applying this patch and its two predecessors and running 'make check' again, and checking that both 'make check' yield identical results. This was done on x86_64-linux-gnu and i686-linux-gnu. This patch was also functionally tested with an ad hoc userland C program which checks the result of difftime for various pairs of 32-bit and, for 64-bit builds, of 64-bit time_t values too. The program was built and run against a glibc with and without the patch, and the results compared to ensure the patch does not change the behavior of difftime. * include/time.h (__difftime64): Add. * time/difftime.c (subtract): convert to 64-bit time. * time/difftime.c (__difftime64): Add. * time/difftime.c (__difftime): Wrap around __difftime64. --- include/time.h | 6 ++++++ time/difftime.c | 31 ++++++++++++++++++++++--------- 2 files changed, 28 insertions(+), 9 deletions(-) diff --git a/include/time.h b/include/time.h index 2710b4521a..7b435c00c2 100644 --- a/include/time.h +++ b/include/time.h @@ -129,6 +129,12 @@ extern char * __strptime_internal (const char *rp, const char *fmt, struct tm *tm, void *statep, locale_t locparam) attribute_hidden; +#if __TIMESIZE == 64 +# define __difftime64 __difftime +#else +extern double __difftime64 (__time64_t time1, __time64_t time0); +#endif + extern double __difftime (time_t time1, time_t time0); /* Use in the clock_* functions. Size of the field representing the diff --git a/time/difftime.c b/time/difftime.c index 7c5dd9898b..e3a4e57b44 100644 --- a/time/difftime.c +++ b/time/difftime.c @@ -31,9 +31,9 @@ time_t is known to be an integer type. */ static double -subtract (time_t time1, time_t time0) +subtract (__time64_t time1, __time64_t time0) { - if (! TYPE_SIGNED (time_t)) + if (! TYPE_SIGNED (__time64_t)) return time1 - time0; else { @@ -76,9 +76,9 @@ subtract (time_t time1, time_t time0) 1 is unsigned in C, so it need not be compared to zero. */ uintmax_t hdt = dt / 2; - time_t ht1 = time1 / 2; - time_t ht0 = time0 / 2; - time_t dht = ht1 - ht0; + __time64_t ht1 = time1 / 2; + __time64_t ht0 = time0 / 2; + __time64_t dht = ht1 - ht0; if (2 < dht - hdt + 1) { @@ -99,18 +99,18 @@ subtract (time_t time1, time_t time0) /* Return the difference between TIME1 and TIME0. */ double -__difftime (time_t time1, time_t time0) +__difftime64 (__time64_t time1, __time64_t time0) { /* Convert to double and then subtract if no double-rounding error could result. */ - if (TYPE_BITS (time_t) <= DBL_MANT_DIG - || (TYPE_FLOATING (time_t) && sizeof (time_t) < sizeof (long double))) + if (TYPE_BITS (__time64_t) <= DBL_MANT_DIG + || (TYPE_FLOATING (__time64_t) && sizeof (__time64_t) < sizeof (long double))) return (double) time1 - (double) time0; /* Likewise for long double. */ - if (TYPE_BITS (time_t) <= LDBL_MANT_DIG || TYPE_FLOATING (time_t)) + if (TYPE_BITS (__time64_t) <= LDBL_MANT_DIG || TYPE_FLOATING (__time64_t)) return (long double) time1 - (long double) time0; /* Subtract the smaller integer from the larger, convert the difference to @@ -118,4 +118,17 @@ __difftime (time_t time1, time_t time0) return time1 < time0 ? - subtract (time0, time1) : subtract (time1, time0); } + +/* Provide a 32-bit wrapper if needed */ + +#if __TIMESIZE != 64 + +double +__difftime (time_t time1, time_t time0) +{ + return __difftime64 (time1, time0); +} + +#endif + strong_alias (__difftime, difftime)