Date: Mon, 22 Dec 2008 10:12:03 +0100 From: Roman Divacky <rdivacky@FreeBSD.org> To: Ed Schouten <ed@80386.nl> Cc: FreeBSD Current <freebsd-current@FreeBSD.org>, Norikatsu Shigemura <nork@FreeBSD.org> Subject: Re: Change select(2) to kevent(2) on script(1)... Message-ID: <20081222091203.GA28920@freebsd.org> In-Reply-To: <20081221211949.GS1176@hoeg.nl> References: <20081221012752.cdc5cbfc.nork@FreeBSD.org> <20081221211949.GS1176@hoeg.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Dec 21, 2008 at 10:19:49PM +0100, Ed Schouten wrote: > * Norikatsu Shigemura <nork@FreeBSD.org> wrote: > > I inspired by your 'kqueue()-support to pseudo-terminal master > > devices' (svn commit: r185942). So I'm trying to use kevent(2) > > on script(1). (SEE ALSO following patch) > > > > But it doesn't work. Because 'master' file descriptor (from > > openpty(3)) doesn't return last 0 byte data readable when shell > > exit. In this case of select(2), 'master' file descriptor > > gets 0 byte data and exit script(1). > > > > Do you have any idea? > > I sure have. It turns out there is a small bug in the pts(4) driver. > Inside ptsdrv_close() (called when the TTY is being closed), we set a > flag to mark the pseudo-terminal as `finished' (which causes it to > return EOF's). > > A difference between select() and kqueue() is that the knote handlers > are executed syncrhonously, while select() calls ptsdev_poll() later on. > This should be fixed as of r186382. Thanks for reporting this issue! > > Are you planning to commit the patch for script(1) to SVN? I'm not sure > if we should port applications to kqueue(). Right now script(1) should > almost cleanly compile/work on other versions of FreeBSD and other > Unix-like operating systems. I believe other OSes should be able to do the porting effort when they want to use FreeBSD software. if kqueue-in-script makes any improvement go for it.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20081222091203.GA28920>