Date: Sat, 28 Oct 1995 13:30:32 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: davidg@root.com Cc: dufault@hda.com, terry@lambert.org, phk@freefall.freebsd.org, current@freebsd.org Subject: Re: fs layering pathces doesn't work. Message-ID: <199510282030.NAA02588@phaeton.artisoft.com> In-Reply-To: <199510281139.EAA00136@corbin.Root.COM> from "David Greenman" at Oct 28, 95 04:39:39 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > >> > >> There, dammit. > >> > >> Man, it is impossible to do diffs against -current! > >> > >> > >> The latest full patches which work are now in: > >> > >> freefall.cdrom.com:~terry/fs_layer_patch.gz > >> > >> These should fix all the problems in the ancillary file > >> systems and in the mkdir and link calls (your panic was > >> in mkdir). > > > >While it is still working why don't you get commit privs and commit them > >to -current? > > Because several people, me included, object to giving him commit privs. News to me; I've never asked. At first it would have been a bad idea because of my Novell connection, but later it never occurred to me that it would be interesting to tie myself to FreeBSD vs. NetBSD. I am an advocate of BSD, period. Having never asked, I didn't know there were personal problems. Go figure. > Furthermore, the changes haven't yet been reviewed by the people working on > the filesystem layering. Who would that be? Whoever they are, if I could find out what their future plans are, it could save me some work in cleaning up the 4.4BSD hacks to get the Heidemann code in in the first place. I'm sure we are all interested in the implementation matching the design document so that the original design goals can, infact, be met instead of being compromised (as they are in the 4.4 code). Currently, there is still layering manipulation which quite broken that prevents subclassing file systems. Specifically, the address reference for the per op descriptors should be replaces with a token and the instance registration mechanism needs to convert the token to the op descriptors in place. Or hasn't anyone (besides me) looked at the 4.4BSD and NetBSD unionfs (which works), and the original Ficus papers? I realize that my current patch is a big one, but I've given my full rationale, and no one objected to it. If I could ever get over chasing -current's tail for lack of a decent cvs diff/cvs merge facility that can be used remotely, it might even apply cleanly. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199510282030.NAA02588>