| Age | Commit message (Collapse) | Author |
|---|
|
that use the miniroot install scripts.
During an install, donetconfig() in miniroot/install.sub now
creates a minimal /etc/protocols and /etc/services.
During an upgrade, enable_network() now copies the existing
protocols and services files into /etc.
Saves approximately 13K on floppy/cd images.
|
|
out. Save about 287 bytes. Based on diff from deraadt@.
Update Copyright (ok deraadt@).
|
|
trailing spaces during install.
Do not make the previously entered first password value the default
value if passwords do not match the first time.
Update Copyright (ok deraadt@).
|
|
When a system is rebooted rather than power cycled it is possible
for duplicate entries for a disk or cd to appear in the dmesg
output. One from the previous boot and one from the current
reboot.
bsort() correctly filtered out the duplicates, but could emit a
trailing space.
This trailing space caused problems when, for instance, the root
disk was being determined. With the trailing space present, the
install or upgrade script would not fill in the default root disk,
even if the system only had one disk.
|
|
default response ($1) are preserved in $resp.
Fix handling of _ftp_server_login, _ftp_server_password and
_url_base so any blanks present are preserved.
Problem with blanks pointed out by drahn@.
|
|
section 3.1. Allows for use of '@', ':' and '/' in usercodes
and passwords without confusion.
ok deraadt@
|
|
eat your heart out krw
|
|
comments that, being on the end of code lines, cannot be stripped
automatically.
Just delete unhelpful ones, and move the rest onto separate lines
where they can be stripped.
Also remove a duplicated initialization of ROOTDISK.
In total, reduces script size on floppy by 1117 bytes, making room
for coming improvements.
|
|
a) Allow user to enter empty FQDN and assume DHCP will supply one.
b) Silently check for any DHCP supplied FQDN after network
interfaces are initialized. *NOTE* DHCP information will
now override any user supplied FQDN.
c) Apply FQDN information to hosts file only after DHCP has had a
chance to supply information.
d) Don't modify hosts file if FQDN is empty.
e) Don't preserve hosts file across install script restarts.
f) Don't preserve hostname.* files across install script restarts.
g) Don't preserve decision to use resolver during install across
install script restarts.
h) Don't allow DHCP to force use of resolver during install.
i) Ensure DHCP created resolv.conf file is saved as /etc/resolv.conf.
j) Add IPv6 localhost ('::1') line to hosts file. (Closes PR#2155)
k) Make IPv4 and IPv6 localhost lines in hosts file look like
'X localhost.domain.name localhost'
Bonus: install.sub on the floppy image is 51 bytes smaller.
ok millert@
|
|
Obtaining the FQDN *after* calling configurenetwork(), so that
DHCP info can be used, turns out to not be a good idea because
/etc/hosts is created in configurenetwork().
As a result the /etc/hosts created by the 3.0 install script
contains bogus entries with a '.' instead of '.<domain name>'.
So, put configurenetwork() back where it was pre-3.0, i.e. after
FQDN is determined.
Noticed by David Krause (PR#2155).
Other /etc/hosts issues (including David's ipv6 suggestion) under
investigation.
|
|
current /etc/netstart, leaving out all code relevant to IPv6 and
multicasting.
In addition, ignore any '!' lines in hostname.if files and ignore
any bridge.if files.
This change enables dhcp configured interfaces to work during the
upgrade process, amoung other improvements.
Also rework FQDN handling to ensure that it is set *after* any
dhcp activity and is set both during installs and upgrades. FQDN
is now set from the last DOMAIN or SEARCH statement in
resolv.conf.
Architecture specific install.sub's (macppc, mvme68k, mvme88k,
mvmeppc, sparc64) are not being updated for 3.0.
ok deraadt@
|
|
|
|
promising this would not happen.
Historically this was being done to preserve any ufs -> ffs
changes the upgrade script had done, but the script has not done
such changes for some time.
ok millert@ deraadt@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
install.md at build time. remove some crud. and the icing on the cake:
skip testing all this until next week
|
|
Remove chunk of commented out, unused code.
Add _didnet=1 to enable_network() so that ftp upgrade
does not ask for network information it already has.
Tell a user that selects the http installation method
that the files will be downloaded via http.
Suppress all 'softdep' options in fstab_shadow so
there are no warnings when the filesystems are
mounted.
Suppress all non-ffs (including nfs, ufs) filesystems
from fstab_shadow. Ufs was never used to install
OpenBSD.
Suppress all 'noauto' filesystems from fstab_shadow.
Reword some comments and messages, esp. the one
just before the offer to edit fstab. Punch up the
fact that the fstab will not be saved, mention
that non-ffs and noauto filesystems will be
ignored.
Move
/usr/X11R6/lib/X11/{rstart,twm,xkb,xsm,xinit}
directories to
/usr/X11R6/lib/X11/XF3{rstart,twm,xkb,xsm,xinit}
and replace with links. Links can be overwritten
during upgrade. Directories can't. This enables the
upgrade from XFree86 3.x to XFree86 4.x.
ok deraadt@ millert@
|
|
only print a 'Done.' for fsck'ing if fsck'ing is actually
attempted.
ok deraadt@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
don't eval ${_ftp_server_password}. This should allow FTP passwords
that contain shell meta characters.
ok millert@
|
|
me, bytes saved by millert, OKed by mickey, millert.
|
|
|
|
|
|
i think i'll fix sparc next week to use it.
|
|
|
|
|
|
|
|
|
|
|
|
For ISO9660 CDs we can take advantage of the label spoofing and not
ask the user for partition/fstype. Also shorten a few variable names.
This makes CD installs go more smoothly.
|
|
Don't show gre interfaces when prompting user for an ether device
|
|
/etc/mail/sendmail.cf, move /etc/sendmail.cf -> /etc/mail/sendmail.cf
and edit /etc/rc if needed. That way the sendmail binary will be
able to find its .cf file on reboot.
|
|
are upgrading not installing) then execute it chroot'd to /mnt.
This allows people to customize the install procedure in a simple
yet flexible manner.
|
|
|
|
|
|
|
|
right before unmounting the new system.
Beats having mtree files around every package...
|
|
|
|
|
|
|
