Date: Mon, 9 May 2011 18:29:01 -0400 From: David Schultz <das@FreeBSD.ORG> To: Chris Rees <utisoft@gmail.com> Cc: FreeBSD Standards <freebsd-standards@FreeBSD.ORG>, Maxim Konovalov <maxim.konovalov@gmail.com>, Oliver Fromme <olli@lurza.secnetix.de> Subject: Re: dd dies on SIGUSR1 Message-ID: <20110509222901.GA18428@zim.MIT.EDU> In-Reply-To: <BANLkTinZiFWLe-Xj=Y9Awe2SM9R_d7%2BToQ@mail.gmail.com> References: <AANLkTikoZNpmM83%2BU-0AWhO43K67gKNq1dZ4UnL2UAPo@mail.gmail.com> <201103221457.p2MEvJub035858@lurza.secnetix.de> <AANLkTinzhKi-sfW-kz9W6EkA0WtB5-nO0gpyCLRyyHCn@mail.gmail.com> <20110322181604.GA47588@zim.MIT.EDU> <AANLkTi=PE6beTB1wmC8v41PqAWWSqq%2B6z-Be44uePYtZ@mail.gmail.com> <alpine.BSF.2.00.1103222136510.17256@qvfongpu.qngnvk.ybpny> <AANLkTikhgk3YRuFoGjBf725b%2B421qDXCWBMSn3PrA5t5@mail.gmail.com> <AANLkTimya2k%2B9mNzFnVCL1jjqj%2BQ9xDBYO2VO5d-AQyY@mail.gmail.com> <20110325033736.GA64512@zim.MIT.EDU> <BANLkTinZiFWLe-Xj=Y9Awe2SM9R_d7%2BToQ@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, May 09, 2011, Chris Rees wrote: > On 25 March 2011 03:37, David Schultz <das@freebsd.org> wrote: > > On Wed, Mar 23, 2011, Eitan Adler wrote: > >> > We are talking about a design decision taken decades ago, which quite > >> > possibly was a mistake. > >> > >> Historical reasons are not be discounted, but in this case because the > >> behavior is already non-portable, and already not be relied upon, so > >> there is no reason that changing the default is harmful. > >> > >> > Again, how many people rely on USR1 to terminate a process? > >> > >> Hopefully none. Even if there are people who do rely on such behavior > >> that reliance could be said to be a mistake or otherwise broken. > > > > Please see my previous message. The historical behavior of SIGUSR1 > > terminating a process by default is standard, even on Linux. > > > > I believe one of the original uses of the signal was to allow > > daemons and their children to signal each other. In this use > > case, if the notification can't be delivered because the recipient > > is unprepared to accept it, termination is appropriate for a > > fail-fast design. > > Since the consensus seems to be for leaving as-is, perhaps someone > could please close bin/155034? > > You can state that I've abandoned it! Looking into the solution originally proposed is still on my todo list... In researching it further, I noticed that even Linux doesn't support this convention consistently: Most utilities die on receipt of SIGUSR1, or fail to do anything useful. dd appears to be the sole exception.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20110509222901.GA18428>