Subject: Re: vnconfig is broken (was: vnd is broken?)
To: Rui Paulo <rpaulo@netbsd-pt.org>
From: Greywolf <greywolf@starwolf.com>
List: current-users
Date: 05/10/2005 21:52:43
[Thus spake Rui Paulo ("RP: ") 10:29pm...]

RP: EXAMPLES
RP:            vnconfig vnd0 /tmp/diskimage
RP:      or
RP:            vnconfig /dev/rvnd0c /tmp/diskimage
RP:
RP:      Configures the vnode disk vnd0.  Please note that use of the second form
RP:      of the command is discouraged because it requires knowledge of the raw
RP:      partition which varies between architectures.
RP:
RP:      [...]
RP:
RP: I find useful to get the second argument for -c from the CWD, but not the
RP: first argument. If I really want it to choose a device not in /dev, I would
RP: use a full path. Another approach is: if 'vnd??' is specified for the first
RP: argument, check /dev first and if the file doesn't exists, check for it in
RP: the CWD.

Can anyone really think of a compelling reason not to look in /dev ONLY
for the vnd device?

Okay, that's done.  Now, can anyone think of a compelling reason that
a full pathname MUST be specified while configuring a device, i.e. why
the config program can't be bothered to prepend the CWD in the case of
a relative path?  Or do we already do this?  (last time I did a vnd,
it demanded an absolute pathname.)

				--*greywolf;
--