From mboxrd@z Thu Jan 1 00:00:00 1970 X-Msuck: nntp://news.gmane.org/gmane.linux.lib.musl.general/10419 Path: news.gmane.org!.POSTED!not-for-mail From: Szabolcs Nagy Newsgroups: gmane.linux.lib.musl.general Subject: [PATCH] fix strtod and strtof rounding with many trailing zeros Date: Sun, 4 Sep 2016 04:46:00 +0200 Message-ID: <20160904024600.GM1280@port70.net> Reply-To: musl@lists.openwall.com NNTP-Posting-Host: blaine.gmane.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Trace: blaine.gmane.org 1472957180 11776 195.159.176.226 (4 Sep 2016 02:46:20 GMT) X-Complaints-To: usenet@blaine.gmane.org NNTP-Posting-Date: Sun, 4 Sep 2016 02:46:20 +0000 (UTC) User-Agent: Mutt/1.6.0 (2016-04-01) To: musl@lists.openwall.com Original-X-From: musl-return-10432-gllmg-musl=m.gmane.org@lists.openwall.com Sun Sep 04 04:46:17 2016 Return-path: Envelope-to: gllmg-musl@m.gmane.org Original-Received: from mother.openwall.net ([195.42.179.200]) by blaine.gmane.org with smtp (Exim 4.84_2) (envelope-from ) id 1bgNRn-0002R6-Ox for gllmg-musl@m.gmane.org; Sun, 04 Sep 2016 04:46:15 +0200 Original-Received: (qmail 32005 invoked by uid 550); 4 Sep 2016 02:46:13 -0000 Mailing-List: contact musl-help@lists.openwall.com; run by ezmlm Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: List-ID: Original-Received: (qmail 31979 invoked from network); 4 Sep 2016 02:46:12 -0000 Mail-Followup-To: musl@lists.openwall.com Content-Disposition: inline Xref: news.gmane.org gmane.linux.lib.musl.general:10419 Archived-At: in certain cases excessive trailing zeros could cause incorrect rounding from long double to double or float in decfloat. e.g. in strtof("9444733528689243848704.000000", 0) the argument is 0x1.000001p+73, exactly halfway between two representible floats, this incorrectly got rounded to 0x1.000002p+73 instead of 0x1p+73, but with less trailing 0 the rounding was fine. the fix makes sure that the z index always points one past the last non-zero digit in the base 10^9 representation, this way trailing zeros don't affect the rounding logic. --- i think x[z-1] is fine in the loop here as z>0 and x[0]!=0 and later code does not turn x[z-1] into 0. the upscaling code path already removed trailing zeros but that was not always taken. only tested on x86, the bad rounding cases depend on long double representation. src/internal/floatscan.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/internal/floatscan.c b/src/internal/floatscan.c index 17d8e70..66d01ef 100644 --- a/src/internal/floatscan.c +++ b/src/internal/floatscan.c @@ -172,6 +172,9 @@ static long double decfloat(FILE *f, int c, int bits, int emin, int sign, int po return sign * (long double)x[0] * p10s[rp-10]; } + /* Drop trailing zeros */ + for (; !x[z-1]; z--); + /* Align radix point to B1B digit boundary */ if (rp % 9) { int rpm9 = rp>=0 ? rp%9 : rp%9+9; -- 2.9.3