Subject: Re: [2.0.2/i386] Install without sysinst : single-user mode? sysinst?
To: Dan LaBell <email@example.com>
From: David Laight <firstname.lastname@example.org>
Date: 06/27/2005 21:43:13
On Mon, Jun 27, 2005 at 04:06:40PM -0400, Dan LaBell wrote:
> On Jun 27, 2005, at 7:33 AM, Sylvain Briole wrote:
> >I am also looking for informations about sysinst, in order to
> >automatize my
> >installations, e.g. how to track the commands given by sysinst during
> >installation or what are the modified/created files (e.g. to select a
> >encryption scheme, disklabels definitions, ...).
How long does it take to answer the questions sysinst asks?
Not long - especially when you've done it a few times.
So why bother!
OTOH most of systint is an interactive fdisk + disklabel + newfs,
followed by running pax to extract the sets.
A few fields in files like rc.conf are changed, but not much.
> Run sysinst, then "Utilities" , then "Logging functions", then
> "Logging" on, then
> "Scripting" On", this will produce a files sysinst.log and sysinst.sh
> I don't know how useful sysinst.sh is -- if it's complete, as some of
> the shell looks wrong to me. Looks likes it's trying multiple test
> against $? (2nd reference to $? would in-fact be test ( [ ) -- so,
> judging from the file, so I'm assuming it's not complete. However some
> files (disktab, fstab,rc.conf ) are in sysinst.sh as "here text", so
> could be useful just for that -- you'll need to copy them off ramdisk
> manually, or they're gone.
The 'scripting' feature is a pain, I've sometimes thought of removing it.
Maybe adding some extra logging instead.
The problem is that it forces sysinst to execute programs instead of
making system calls - so it does system("rm xxx") instead of unlink("xxx").
This increases the complexity and makes error reporting/recovery more
painful - not to mention problems when rmdir(1) is absent from the media.
Note that sysinst dosn't run fdisk, but does run disklabel - even though
the system call would work better!
David Laight: email@example.com