Date: Fri, 26 Oct 2001 18:23:23 -0700 (PDT) From: John Baldwin <jhb@FreeBSD.org> To: John Baldwin <jhb@FreeBSD.org> Cc: Dag-Erling Smorgrav <des@ofug.org>, Mike Smith <msmith@FreeBSD.ORG>, Bakul Shah <bakul@bitblocks.com>, Poul-Henning Kamp <phk@critter.freebsd.dk>, Peter Wemm <peter@wemm.org>, arch@FreeBSD.ORG, Matthew Dillon <dillon@apollo.backplane.com> Subject: Re: 64 bit times revisited.. Message-ID: <XFMail.011026182323.jhb@FreeBSD.org> In-Reply-To: <XFMail.011026181705.jhb@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 27-Oct-01 John Baldwin wrote: > > On 27-Oct-01 Matthew Dillon wrote: >> >>: >>:Matthew Dillon <dillon@apollo.backplane.com> writes: >>:> Is anyone game for this project? >>: >>:I'd volunteer, but I have too many of my own patches to worry about >>:right now. How about mid-November, after BSDCon Europe? >>: >>:DES >>:-- >>:Dag-Erling Smorgrav - des@ofug.org >> >> With the vnode and sync scaleability stuff *almost* out of the way I've >> started working on the Giant lock unwinding stuff, so I don't have time >> this moment either, but I would certainly have time available >> mid-november to help out! >> >> The project could be done and stabilized in a week with three or more >> people helping out. There is good functional separation: >> >> >> * type changes (stat, timespec, timeval, timex, time_t) >> * syscall number rolls & compatibility code (sorry BSDI, it's more then >> 10) >> * kernel side audit to handle new time_t & structures >> * libc audit - all time related functions >> * userland audit to handle new time_t & structures >> >> >> I think everyone has agreed on time_t going to 64 bits, and of course >> it must be seconds. We have to decide in regards to timeval, stat, and >> timespec. It looks like we may not have to mess with timex, which is >> good. > > You did read the e-mail from Garrett where either SUS or POSIX one requires > time_t to fit in a long? I.e. sizeof(time_t) <= sizeof(long). This means > that if you bump time_t to 64 on i386, you have to bump long to 64, which is > decidely something many people don't want to do. I would suggest if you > insist > on working on this, you first convert time_t to a long so that platforms with > 64-bit longs will have a 64-bit time_t, and then once you've cleaned up the > messes that makes, you will still have time to decide if you want ppc and > i386 > to go form ILP32 to IP32L64 (or however you specify that) which will probably > involve backwards compatible syscalls, etc. One step a time. You don't have > to do it all at once, and after doing the first step, you may find that that > is > enough. My bad. C90 requires that time_t fit into a long according to Garrett. POSIX requires it to be either an integer or floating point with the fractional part zero according to his mail as well. -- John Baldwin <jhb@FreeBSD.org> -- http://www.FreeBSD.org/~jhb/ PGP Key: http://www.baldwin.cx/~john/pgpkey.asc "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.011026182323.jhb>