From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 2134) id 34E843854807; Wed, 7 Jul 2021 17:22:10 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 34E843854807 Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit From: Jeff Johnston To: newlib-cvs@sourceware.org Subject: [newlib-cygwin] stdlib: Make strtod/strtof set ERANGE consistently for underflow. X-Act-Checkin: newlib-cygwin X-Git-Author: Keith Packard X-Git-Refname: refs/heads/master X-Git-Oldrev: 1290301b893189b50569ac9b43a62afd65e0e064 X-Git-Newrev: fb01286fab9b370c86323f84a46285cfbebfe4ff Message-Id: <20210707172210.34E843854807@sourceware.org> Date: Wed, 7 Jul 2021 17:22:10 +0000 (GMT) X-BeenThere: newlib-cvs@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Newlib GIT logs List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 07 Jul 2021 17:22:10 -0000 https://sourceware.org/git/gitweb.cgi?p=newlib-cygwin.git;h=fb01286fab9b370c86323f84a46285cfbebfe4ff commit fb01286fab9b370c86323f84a46285cfbebfe4ff Author: Keith Packard Date: Tue Jun 22 10:26:26 2021 -0700 stdlib: Make strtod/strtof set ERANGE consistently for underflow. The C standard says that errno may acquire the value ERANGE if the result from strtod underflows. According to IEEE 754, underflow occurs whenever the value cannot be represented in normalized form. Newlib is inconsistent in this, setting errno to ERANGE only if the value underflows to zero, but not for denorm values, and never for hex format floats. This patch attempts to consistently set errno to ERANGE for all 'underflow' conditions, which is to say all values which are not exactly zero and which cannot be represented in normalized form. This matches glibc behavior, as well as the Linux, Mac OS X, OpenBSD, FreeBSD and SunOS strtod man pages. Signed-off-by: Keith Packard Diff: --- newlib/libc/stdlib/strtod.c | 31 +++++++++++++++++++++++++++++-- 1 file changed, 29 insertions(+), 2 deletions(-) diff --git a/newlib/libc/stdlib/strtod.c b/newlib/libc/stdlib/strtod.c index 8bb75ef0a..019416ca7 100644 --- a/newlib/libc/stdlib/strtod.c +++ b/newlib/libc/stdlib/strtod.c @@ -326,6 +326,11 @@ _strtod_l (struct _reent *ptr, const char *__restrict s00, char **__restrict se, Bfree(ptr,bb); } ULtod(rv.i, bits, exp, i); +#ifndef NO_ERRNO + /* try to avoid the bug of testing an 8087 register value */ + if ((dword0(rv)&Exp_mask) == 0) + errno = ERANGE; +#endif }} goto ret; } @@ -1238,7 +1243,7 @@ _strtod_l (struct _reent *ptr, const char *__restrict s00, char **__restrict se, dval(rv) *= dval(rv0); #ifndef NO_ERRNO /* try to avoid the bug of testing an 8087 register value */ - if (dword0(rv) == 0 && dword1(rv) == 0) + if ((dword0(rv) & Exp_mask) == 0) ptr->_errno = ERANGE; #endif } @@ -1298,6 +1303,28 @@ strtof_l (const char *__restrict s00, char **__restrict se, locale_t loc) return retval; } +/* + * These two functions are not quite correct as they return true for + * zero, however they are 'good enough' for the test in strtof below + * as we only need to know whether the double test is false when + * the float test is true. + */ +static inline int +isdenorm(double d) +{ + U u; + dval(u) = d; + return (dword0(u) & Exp_mask) == 0; +} + +static inline int +isdenormf(float f) +{ + union { float f; __uint32_t i; } u; + u.f = f; + return (u.i & 0x7f800000) == 0; +} + float strtof (const char *__restrict s00, char **__restrict se) @@ -1307,7 +1334,7 @@ strtof (const char *__restrict s00, return signbit (val) ? -nanf ("") : nanf (""); float retval = (float) val; #ifndef NO_ERRNO - if (isinf (retval) && !isinf (val)) + if ((isinf (retval) && !isinf (val)) || (isdenormf(retval) && !isdenorm(val))) _REENT->_errno = ERANGE; #endif return retval;