Date: Thu, 2 Nov 1995 19:53:30 -0600 From: rkw@dataplex.net (Richard Wackerbarth) To: "Garrett A. Wollman" <wollman@lcs.mit.edu> Cc: "Jordan K. Hubbard" <jkh@time.cdrom.com>, hackers@freebsd.org Subject: Re: An Outsider's View Message-ID: <v02130502acbf24028c7d@[199.183.109.242]>
next in thread | raw e-mail | index | archive | help
At 6:44 PM 11/2/95, Garrett A. Wollman wrote: ><<On Thu, 02 Nov 1995 16:10:20 -0800, "Jordan K. Hubbard" ><jkh@time.cdrom.com> said: > >> Nobody really likes the model anyway since it requires a read/write >> /usr/src to make the symlinks, and that bites. You won't get much >> (if any) pushback on rearchitecting that. > >Actually, I am rather fond of it myself, principally because it keeps >a lot of cruft out of my way when I'm working on something, but >doesn't require funny filesystems or other sorts of things that, while >they would be nice, I'm not willing to trust on an experimental machine. > >As for the read/write question, well, `man lndir'. I agree, it's not >the best approach, but it's what we've got. (I would frankly be far >more comfortable with a system that kept obj links but made use of a >funky filesystem to cover this case rather than getting rid of obj >links and using a funky filesystem to hide the object cruft.) The scheme that I have in mind would create a separate object tree just as is now created in /usr/obj. The only reason that I see for the links is to have a more direct access path to the intermediate directory of object modules. For many users, this is unnecessary. Those object modules could be kept in some invisible location which is only accessed through the makefile. ---- Richard Wackerbarth rkw@dataplex.net
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?v02130502acbf24028c7d>