Date: Sun, 23 Apr 2017 16:10:56 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Jilles Tjoelker <jilles@stack.nl> Cc: Ngie Cooper <ngie@FreeBSD.org>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r317315 - head/lib/libc/gen Message-ID: <20170423131056.GA1622@kib.kiev.ua> In-Reply-To: <20170423130736.GA53007@stack.nl> References: <201704230317.v3N3HWtd040005@repo.freebsd.org> <20170423095345.GC57674@kib.kiev.ua> <20170423130736.GA53007@stack.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Apr 23, 2017 at 03:07:36PM +0200, Jilles Tjoelker wrote: > On Sun, Apr 23, 2017 at 12:53:45PM +0300, Konstantin Belousov wrote: > > On Sun, Apr 23, 2017 at 03:17:32AM +0000, Ngie Cooper wrote: > > > Author: ngie > > > Date: Sun Apr 23 03:17:32 2017 > > > New Revision: 317315 > > > URL: https://svnweb.freebsd.org/changeset/base/317315 > > > > Log: > > > Note that getpagesize(3) can return -1 on failure > > > getpagesize() cannot fail. This change only confuses the readers. > > Although I agree that getpagesize() cannot fail and that it is > inappropriate to check for it, the code in lib/libc/gen/getpagesize.c > has a path where it returns -1 (if sysctl CTL_HW HW_PAGESIZE fails). It > probably makes more sense to abort() here to avoid confusing both humans > and static analyzers, and allow easier debugging if a bug causes it to > fail (instead of calculating bogus sizes from the -1 return value). If the sysctl() call fails there, I think it does not make any sense to bother with abort(), since we do not run on a FreeBSD(-like) kernel. Might be, it would be cleaner to change return (-1) to return (PAGE_SIZE) in the 'error case'.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20170423131056.GA1622>