On Thu, Jan 14, 1999 at 03:55:50PM -0500, Roland McGrath wrote: > > You may be right, but a package that installed both /usr/lib/foo and > > /lib/foo would have to be perverse, in my opinion. Both are searched > > by default, so one is as good as the other, except when /usr is not > > mounted. Because /lib comes first (I think), and / is always mounted, > > installing /usr/lib/libfoo.so.1 when /lib/libfoo.so.1 exists would > > have no effect. > > Well, then we think alike. A package that installed both /usr/bin/foo and > /bin/foo would have to be perverse, in my opinion. Both are searched by > default, so one is as good as the other, except when /usr is not mounted.
Well, the point is that usr/lib -> /lib will cause less to none problems than /usr -> /, because we map much less back tot he root filesystem, and conflicts in lib directories are less expected than problems in bin directories. However, this is all moot if we can add support for /usr/lib to the linker easily, which we should do in the Debian packages, IMHO. > Recall now that all of this was prompted by a package that did just that. > So go figure. This is not correct. I know at least one packages which does install the same directories in lib and /usr/lib (the libc5-compat package, which fortunately doesn't apply to the Hurd), but I don't know of any files. Which package do you mean? The original report was about /usr/lib not being searched. Marcus -- "Rhubarb is no Egyptian god." Debian GNU/Linux finger brinkmd@ Marcus Brinkmann http://www.debian.org master.debian.org [EMAIL PROTECTED] for public PGP Key http://homepage.ruhr-uni-bochum.de/Marcus.Brinkmann/ PGP Key ID 36E7CD09
