3eea619b0f
If /tmp is on /, then blowing away its contents, as appropriate, may get you enough space to do the installworld.
1140 lines
43 KiB
Plaintext
1140 lines
43 KiB
Plaintext
Updating Information for FreeBSD current users
|
|
|
|
This file is maintained and copyrighted by M. Warner Losh
|
|
<imp@village.org>. See end of file for further details. For commonly
|
|
done items, please see the COMMON ITEMS: section later in the file.
|
|
|
|
NOTE TO PEOPLE WHO THINK THAT 5.0-CURRENT IS SLOW:
|
|
FreeBSD 5.0-CURRENT has many debugging features turned on, in
|
|
both the kernel and userland. These features attempt to detect
|
|
incorrect use of system primitives, and encourage loud failure
|
|
through extra sanity checking and fail stop semantics. They
|
|
also substantially impact system performance. If you want to
|
|
do performance measurement, benchmarking, and optimization,
|
|
you'll want to turn them off. This includes various WITNESS-
|
|
related kernel options, INVARIANTS, malloc debugging flags
|
|
in userland, and various verbose features in the kernel. Many
|
|
developers choose to disable these features on build machines
|
|
to maximize performance.
|
|
|
|
20021024:
|
|
Old, compatibility slices have been removed in GEOM kernels. This
|
|
means that you will have to update your /etc/fstab to not use
|
|
disk devices of the form /dev/ad0a. Instead, you now must
|
|
specify /dev/ad0s1a, or whatever slice your FreeBSD partition
|
|
really is on. The old device names have gone away, so if you
|
|
use them anywhere else, you must also adjust those uses.
|
|
|
|
20021023:
|
|
Alphas with kernels from between 20020830 and 20021023 and/or
|
|
rtld (ld-elf.so.1) older than 20021023 may experience problems
|
|
with groff while doing a buildworld (kernel: "out of memory",
|
|
fixed in rev 1.129 of kern/imgact_elf.c; rtld: "too few PT_LOAD
|
|
segments", fixed in rev 1.8 of libexec/rtld-elf/map_object.c).
|
|
|
|
So, to successfully upgrade your Alpha, you must either
|
|
upgrade your kernel and rtld first (which might be a bit
|
|
tricky), or avoid running the bootstrapped groff during the
|
|
"transitional" buildworld. To avoid running groff during the
|
|
transitional upgrade run make buildworld with -DNOMAN,
|
|
-DNO_SHAREDOCS, and -DNO_LPR.
|
|
|
|
20020831:
|
|
gcc has been upgraded to 3.2. It is not all binary compatible
|
|
with earlier versions of gcc for c++ programs. All c++
|
|
programs and libraries need to be recompiled.
|
|
|
|
Also, if you encounter g++ issues, rm /usr/include/g++/* before
|
|
doing an installworld to make sure that stale files are removed.
|
|
|
|
20020827:
|
|
Our /etc/termcap now has all the entries from the XFree86 xterm
|
|
almost unchanged. This means xterm now supports color by default.
|
|
If you used TERM=xterm-color in the past you now should use
|
|
TERM=xterm. (xterm-color will lead to benign warnings).
|
|
|
|
20020815:
|
|
A "bug" in gcc(1) that was hiding warning in system headers was
|
|
fixed. It's probably time to add -DNO_WERROR to your make line
|
|
again.
|
|
|
|
20020729:
|
|
COPY is being deprecated. The 20010530 change was reverted, as
|
|
it causes far more pain than was expected, and to always compare
|
|
before installing, please use INSTALL="install -C" again. The
|
|
-C option is now silently ignored when used with the -d option.
|
|
|
|
20020702:
|
|
Problems with libc_r clients like KDE and GNOME have been resolved.
|
|
There are still some minor problems with some signals but the
|
|
system is stable enough for general use again. SMP is less so than UP
|
|
but each can successfully complete multiple buildworlds.
|
|
Libkvm needs to be recompiled due to KSE.
|
|
|
|
20020701:
|
|
Now would be a bad time to upgrade. Something in or near the
|
|
KSE commit totally broke programs using libc_r like KDE and
|
|
GNOME.
|
|
|
|
20020511:
|
|
The k5su utility installed as part of Kerberos 5 is no longer
|
|
installed with the set-user-ID bit set by default. Add
|
|
ENABLE_SUID_K5SU=yes to /etc/make.conf to have it installed
|
|
with the set-user-ID bit set.
|
|
|
|
20020510:
|
|
Gcc 3.1 debugging format (cc -g) has changed from STABS to DWARF2.
|
|
Unfortunately our native GDB (at version 4.18) does not understand
|
|
the DWARF2 debugging format. Thus you must use `gcc -gstabs+' to
|
|
generated debugging information for our native GDB.
|
|
|
|
20020510:
|
|
Due to the way CVS works, it may not properly update src/contrib/gcc
|
|
to the 3.1 sources. The easiest fix is to `rm -rf' src/contrib/gcc
|
|
and then do a cvs update.
|
|
|
|
20020421:
|
|
When exec'ing set[ug]id executables, the kernel now ensures that the
|
|
stdio file descriptors (0..2) are open. See FreeBSD-SA-02:23.stdio.
|
|
|
|
20020404:
|
|
New sendmail startup scripts have been installed to make it
|
|
easier to use alternative MTAs with FreeBSD. Setting the rc.conf
|
|
variable sendmail_enable to "NO" no longer prevents any sendmail
|
|
daemons from starting. Instead, either set sendmail_enable to
|
|
"NONE" or change mta_start_script to a script for starting
|
|
an alternative MTA. Setting mta_start_script to "" will
|
|
also prevent any MTA from being started at boot.
|
|
|
|
20020403:
|
|
UCONSOLE is no longer a valid kernel option.
|
|
|
|
20020315:
|
|
FreeBSD 5.0 DP-1 was basically branched today.
|
|
|
|
20020225:
|
|
Warnings are now errors in the kernel. Unless you are a developer,
|
|
you should add -DNO_WERROR to your make line.
|
|
|
|
20020217:
|
|
sendmail 8.12.2 has been imported. The sendmail binary is no
|
|
longer a set-user-ID root binary and the infrastructure to support
|
|
command line mail submission has changed. Be sure to run
|
|
mergemaster (especially for updating /etc/rc, /etc/defaults/rc.conf,
|
|
and /etc/mail) and read /etc/mail/README for more details.
|
|
|
|
Due to the import of sendmail 8.12.2, a new user and group are
|
|
required in order for sendmail to run as a set-group-ID
|
|
binary. A 'make installworld' will use the new user and group
|
|
to set the owner and group of /var/spool/clientmqueue and will
|
|
fail if the new user and group do not exist. The 'smmsp' user
|
|
and group must be merged from src/etc/group and
|
|
src/etc/master.passwd before using 'make installworld'.
|
|
'mergemaster -p' will do this. You may need to install
|
|
mergemaster before this will work if you are updating from a
|
|
very old version of current. The updating recipe has changed
|
|
as of this date.
|
|
|
|
20020112:
|
|
The preferred configuration method for PAM is now /etc/pam.d/
|
|
rather than /etc/pam.conf. If you have an unmodified
|
|
pam.conf, just delete it after your next mergemaster run. If
|
|
you have local modifications, you can use
|
|
/usr/src/etc/pam.d/convert.pl to incorporate them into your
|
|
/etc/pam.d.
|
|
|
|
Please see the following url for more details:
|
|
http://www.freebsd.org/cgi/mid.cgi?db=mid&id=<xzp6667fyoa.fsf@flood.ping.uio.no>
|
|
20011229:
|
|
If anyone here is already using the new rc.conf(5) variable
|
|
networkfs_types, please note that it has changed
|
|
http://www.freebsd.org/cgi/mid.cgi?db=mid&id=<9744.1009655556@axl.seasidesoftware.co.za>
|
|
|
|
20011220:
|
|
sys/i4b/driver/i4b_ispppsubr.c has been retired. This file
|
|
started out its life in the ISDN4BSD project as an offspring
|
|
from sys/net/if_spppsubr.c, which eventually got a life of its
|
|
own. All the accumulated features and bugfixes of the i4b
|
|
version have now been merged back into the base system's
|
|
version now. The only user-visible change resulting from this
|
|
is that i4b's sppp(4) interfaces are to be managed with
|
|
spppcontrol(8) again, since ispppcontrol(8) has been retired
|
|
as well. (There has never been rc file support for
|
|
ispppcontrol in -current, but only in -stable. That will be
|
|
reverted by the time the changes are MFCed.)
|
|
|
|
20011215:
|
|
The fdc(4) driver has been updated and now automatically
|
|
recognizes media in `standard' formats (like 1440 KB and
|
|
720 KB for a 3.5" high-density drive) when accessing the
|
|
default device node (e. g. /dev/fd0). The old variety of
|
|
floppy device nodes /dev/fd*.* is no longer present by
|
|
default, devices can be created (in DEVFS) on demand. They
|
|
will need to be customized then for `odd' densities using
|
|
fdcontrol(8).
|
|
|
|
20011209:
|
|
The bugs in procfs' debugging support code have been fixed,
|
|
and truss(1) now works again.
|
|
|
|
20011207:
|
|
Daily security checks have been split out to use the periodic(8)
|
|
scripts. Some change in configuration may be necessary. Please
|
|
see
|
|
http://www.freebsd.org/cgi/mid.cgi?db=mid&id=<20011207155805.R8975@blossom.cjclark.org>
|
|
for details.
|
|
|
|
20011204:
|
|
sos added VCD/SVCD support to ata driver and that needs the
|
|
kernel and burncd to be in sync.
|
|
|
|
20011203:
|
|
The procfs pseudo-filesystem has now been converted to use the
|
|
pseudofs framework. If you have 'options PROCFS' in your
|
|
kernel config, you'll need to add 'options PSEUDOFS' if it's
|
|
not there already.
|
|
|
|
This change temporarily breaks truss(1); use ktrace(1) instead
|
|
until the issue has been resolved.
|
|
|
|
20011202:
|
|
A security hole in OpenSSH involving `UseLogin yes' has been
|
|
patched.
|
|
|
|
20011126:
|
|
You need to remove /usr/obj/.../usr.bin/tip before rebuilding
|
|
after this date. You need to do this only once.
|
|
|
|
20011103:
|
|
Most of the awk issues have been resolved. Some rough
|
|
edges may be left, but for the most part things should be
|
|
back to "normal." For CURRENT's usual definition of "normal."
|
|
|
|
20011030:
|
|
Awk has been upgraded to the one true awk from bell labs. Expect
|
|
choppy waves in the upgrade process.
|
|
|
|
20011030:
|
|
The asr driver problem has been resolved.
|
|
|
|
20011027:
|
|
Due to changes in other parts of the system, the asr driver
|
|
now causes the system to panic on boot. Do not use it pending
|
|
correction. Comment it out of any kernel config file that you
|
|
try to use from this date forward.
|
|
|
|
20011025:
|
|
When crossbuilding, use TARGET=xxx where you used to use
|
|
MACHINE=xxx. You don't need to set TARGET_ARCH and TARGET,
|
|
unless you are changing both of them. To cross build pc98 on
|
|
an alpha, for example, you need to set TARGET=pc98 and
|
|
TARGET_ARCH=i386.
|
|
|
|
20011001:
|
|
The kernel interface that burncd depends on has changed.
|
|
You must recompile both the kernel and userland applications
|
|
at the same time.
|
|
|
|
20010929:
|
|
When crossbuilding, please set TARGET_ARCH rather than
|
|
MACHINE_ARCH to indicate the target. In the future, one will
|
|
set TARGET_MACHINE where you set MACHINE now. At the moment,
|
|
setting MACHINE alone for same MACHINE_ARCH machines works
|
|
(eg, you can build pc98 on a i386 machine and vice versa).
|
|
|
|
20010927:
|
|
Some weird problems result from using ACPI on some machines.
|
|
To disable ACPI you can add
|
|
hint.acpi.0.disable="1"
|
|
to /boot/loader.conf (or by putting set X=Y at the boot
|
|
loader "ok" prompt).
|
|
|
|
Alternatively, you can remove it from /boot/kernel/acpi.ko
|
|
or use the MODULES_OVERRIDE function in your kernel config
|
|
file and not list acpi in that list.
|
|
|
|
20010924:
|
|
The buildworld has been fixed. You may need to install
|
|
the 4.x compatibility libraries for some old binaries
|
|
to work. Add COMPAT4X=true to your /etc/make.conf to
|
|
get them installed on every installworld, or execute the
|
|
following to get them installed only once:
|
|
cd src/lib/compat/compat4x.<arch>
|
|
make all install
|
|
You will see ``__stdoutp undefined'' until you do this.
|
|
|
|
20010919:
|
|
There's a bug in the world build process. The cross-tools
|
|
are build with the NEW headers, but the OLD libc.a. This
|
|
leads to all kinds of problems with the new libc. A temporary
|
|
workaround is to add
|
|
CFLAGS="-O -pipe -D_OLD_STDIO"
|
|
before building world when upgrading from 4.x to current. This
|
|
can be removed afterwards.
|
|
|
|
A proper fix to the buildworld target is needed.
|
|
|
|
20010918:
|
|
Peter has committed his new kthread nfs client/server code.
|
|
NFS may be unstable after this date.
|
|
|
|
20010912:
|
|
KSE has hit the tree. Lots of things are now different in
|
|
the kernel. While a few problems were introduced in the
|
|
initial commit, most of the major ones have been found and
|
|
corrected.
|
|
|
|
20010901:
|
|
In OLDCARD, CardBus bridges appear to be stable. The work
|
|
arounds described in the 20010604 entry are now no longer
|
|
necessary and will be ignored. Most insert/remove problems
|
|
have been rectified around this date.
|
|
|
|
20010823:
|
|
named now runs as user bind and group bind rather than as
|
|
root. If named_enable is set to YES in /etc/rc.conf, ensure
|
|
that user bind is available in /etc/passwd (using vipw(8))
|
|
and that group bind is available in /etc/group. Also make
|
|
sure that user or group bind has read (and not write)
|
|
permission for your name server configuration and that it
|
|
has read and write permission for your slave zone files and
|
|
directory.
|
|
|
|
If you wish to continue to run named as root (a less secure
|
|
alternative), add a line to /etc/rc.conf saying
|
|
|
|
named_flags=
|
|
|
|
20010709:
|
|
The PAM libraries have had an API upgrade that is beyond
|
|
the ability of the shared library major number to handle.
|
|
It is manifested by PAM-using ports dumping core. The
|
|
solution is to rebuild those ports.
|
|
|
|
20010628:
|
|
The kernel compile module has moved from src/sys/compile/FOO
|
|
to src/sys/${MACHINE}/compile/FOO.
|
|
|
|
20010625:
|
|
The pccard modem issue from 20010613 has been corrected.
|
|
OLDCARD support is still a little weak in -current. slot 1 is
|
|
known not to work on some TI based cardbus bridges. Some
|
|
cardbus bridges do not properly detect insert/removal events.
|
|
IRQ configuration needs more safety belts.
|
|
|
|
20010617:
|
|
Softupdates problems have been corrected.
|
|
|
|
20010614:
|
|
Peter ripped out the linkerset support. You must, as always,
|
|
rerun config after you cvsup if you are using the traditional
|
|
kernel building methods.
|
|
|
|
20010613:
|
|
pccard modems may not work with current after 20010604 date. Some
|
|
do, others result in panics. *MAKE*SURE* that you update your
|
|
config and /etc/rc.conf ala the 20010604 entry, or you will have
|
|
problems (this issue will be fixed, it just hasn't been yet).
|
|
|
|
20010613:
|
|
SOFTUPDATES seem to be broken since the middle of May or so. Do not
|
|
use them in current. You can disable softupdates on all mounted
|
|
partitions, or remove SOFTUPDATES the kernel config file.
|
|
|
|
20010612:
|
|
After Peter's commits to the hints code, people have been noticing
|
|
that certain devices are attached (or try to) twice. This is due
|
|
to having both static hints as well as a /boot/device.hints. To
|
|
work around this issue, please use only one or the other mechanism
|
|
until this bug is fixed.
|
|
|
|
Please note that a feature of config is that if you have config
|
|
file FOO and FOO.hints, it automatically adds FOO.hints to the
|
|
hints.c file, wheather you want it to or not.
|
|
|
|
20010610:
|
|
Locale names have changed to match other systems better.
|
|
|
|
20010604:
|
|
pccard support for pci cards has been committed. You must change
|
|
your /etc/pccard.conf irq lines. It must match the irq used by
|
|
pcic device. Interrupt storms may result if you fail to do this.
|
|
Interrupt storms look a lot like a hang.
|
|
|
|
You must also install a new pccardd, otherwise you will get an
|
|
interrupt storm at card reset time (just after it tells you what
|
|
it is).
|
|
|
|
pccardd_flags="-I" is necessary for the time being. It tells pccardd
|
|
not to ask the kernel if the interrupt is really free or not before
|
|
using it. You can either change the /etc/pccard.conf irq lines to
|
|
match pcic, or add "-i X" to the pccardd_flags.
|
|
|
|
20010530:
|
|
INSTALL=install -C is being deprecated. If you want to do this,
|
|
use COPY=-C instead. The former method will be supported for only
|
|
a limited time. If you see
|
|
|
|
install: warning: the -d and -C options may not be specified together
|
|
|
|
in your makeworld, then you need to migrate towards using
|
|
COPY=-C.
|
|
|
|
20010525:
|
|
It appears that vm is now stable enough to use again. However,
|
|
there may be other problems, so caution is still urged. alpha
|
|
definitely is in bad shape.
|
|
|
|
20010521:
|
|
Minor repo damange has happened. This may cause problems
|
|
with cvsup of ports. If you get errors, please see
|
|
http://www.FreeBSD.org/cgi/query-pr.cgi?pr=27495
|
|
at the bottom for details on a workaround. The error message
|
|
is
|
|
Updater failed: Cannot delete "/usr/ports/www/jakarta-tomcat/files": Directory not empty
|
|
|
|
20010520:
|
|
Vm and/or swapping are busted on -current. Please be patient.
|
|
|
|
20010519:
|
|
pccard has had much reorganizational work done to it over
|
|
the past few days. Everything should still work, but if
|
|
not, please contact imp@freebsd.org.
|
|
|
|
20010517:
|
|
ata ioctl changed. Make sure to recompile both kernel and
|
|
userland at the same time.
|
|
|
|
20010517:
|
|
New ncurses imported.
|
|
|
|
20010512:
|
|
DEVFS is now opt out, not opt in. Barring major problems, this
|
|
will be the only way to go starting July 1.
|
|
|
|
20010504:
|
|
OpenSSH has been updated to 2.9. Some defaults are different,
|
|
including RhostsRSAAuthentication, which changes from yes to no.
|
|
|
|
20010502:
|
|
Perl breakage in 20010501 was corrected at 14:18:33 PDT.
|
|
|
|
20010501:
|
|
Building perl was broken at 02:25:25 PDT.
|
|
|
|
20010430:
|
|
The bug in 20010429 was corrected at 07:35:37 PDT. It is safe to
|
|
go back in the water.
|
|
|
|
20010429:
|
|
A bad bug was committed at 04:48:42 PDT. Don't use kernels after
|
|
this date, but before the correction date.
|
|
|
|
20010423:
|
|
old fsck and new kernel interactions appear to have been fixed.
|
|
|
|
20010411:
|
|
fsck and the kernel were changed to handle some optimizations
|
|
to directory layout. This breaks backward compatibility.
|
|
Update only if you understand that you must not use the old
|
|
fsck with the new kernel ever.
|
|
|
|
20010330:
|
|
fsck has changed the meaning of the pass column in /etc/fstab.
|
|
Please see the cvs commit to fsck.8 or the fsck.8 man page for
|
|
details. It is unclear if changes to /etc/fstab are necessary.
|
|
|
|
20010319:
|
|
portmap had changed name to rpcbind for maximum POLA in your
|
|
current world. /etc/hosts.{allow,deny} needs changes. nfs and
|
|
other rpc based programs that rely on portmapper will not work
|
|
without updates to /etc/hosts.{allow,deny} and /etc/netconfig.
|
|
|
|
20010315:
|
|
ata subsystem changes. ATA_ENABLE_ATAPI_DMA, ATA_ENABLE_WC
|
|
and ATA_ENABEL_TAGS are no longer kernel options. They have
|
|
been replaced by tunables. See ata.4 for details.
|
|
|
|
20010312:
|
|
The fxp driver was converted to use miibus. If you compile
|
|
fxp into your kernel statically, you will need to add miibus.
|
|
|
|
20010312:
|
|
The wi device now defaults to BSS (infrastructure) mode
|
|
instead of ad-hoc.
|
|
|
|
20010310:
|
|
/dev/urandom should be a symbolic link to /dev/random now.
|
|
Users of current not using DEVFS need to run MAKEDEV std.
|
|
ssh might not work if you don't.
|
|
|
|
20010303:
|
|
The ed driver has been updated. It now allows mii attachments,
|
|
which means that you must include the miibus in your kernel if
|
|
you use the ed driver.
|
|
|
|
20010220:
|
|
The problems with libc have been corrected. It is now mostly
|
|
safe to go back into the water.
|
|
|
|
20010211:
|
|
The size of FILE was changed. This breaks upgrading. If
|
|
you must upgrade, be prepared for pain. It also breaks almost
|
|
all binaries that you've compiled on -current. You are warned
|
|
that before upgrading would be a good time to do a level 0
|
|
dump of your system. No, really, I mean it this time.
|
|
|
|
To get to the new system, you'll need to use the following
|
|
workaround. Hopefully this can be sorted out so that we
|
|
don't have to move this to the updating section.
|
|
|
|
To get around the installworld problem, do:
|
|
# cd /usr/src/usr.bin/sed
|
|
# make install
|
|
# cd /usr/src
|
|
# make installworld
|
|
If that doesn't work, then try:
|
|
# make -k installworld
|
|
# make installworld
|
|
|
|
20010207:
|
|
DEVFS is now the default. If you use vinum, make sure that you
|
|
do not include devfs in your kernel as problems result.
|
|
|
|
20010205:
|
|
FFS_ROOT and CD9660_ROOT have been removed or deprecated.
|
|
Remove them from your config.
|
|
|
|
20010122:
|
|
****************************** WARNING ******************************
|
|
buildkernel has been changed slightly
|
|
****************************** WARNING ******************************
|
|
KERNCONF replaces the variable KERNEL for buildkernel. You
|
|
should update your scripts and make.conf accordingly.
|
|
|
|
20010119:
|
|
config has changed to allow DEV_FOO as a replacement for NFOO.
|
|
This requires a new config to build correctly.
|
|
|
|
20010116:
|
|
The kerrnel option I386_CPU is now mutually exclusive with the
|
|
other cpu types. If you have an i386 system, be sure that it
|
|
only had this line. Remove it for all other configurations.
|
|
|
|
20010110:
|
|
Changes to the kernel require it and burncd be in sync.
|
|
|
|
20010102:
|
|
Everyone who has hw.sndunit set to something in
|
|
/etc/sysctl.conf, it is now hw.snd.unit.
|
|
|
|
20010101:
|
|
ex and vi were broken by some changes to sys/queue.h. If you
|
|
have a bad vi, you will see make buildworld fail with a core
|
|
dump while building termcap. You can work around this problem
|
|
by adding -k to your make buildworld. This will cause the
|
|
build to complete and install a new vi. Once that's done, you
|
|
can rebuild again without the -k to pick up anything that
|
|
might have been ignored by the -k option.
|
|
|
|
Others have suggested that you can just rebuild libc if your
|
|
vi/ex is dynamically linked, but I've not received any reports
|
|
of this working.
|
|
|
|
20001228:
|
|
There have been some changes to libcrypt in -current. The
|
|
libscrypt/libdescrypt symlink sillyness is gone and the installed
|
|
libcrypt is fully functional. Be aware of this.
|
|
|
|
20001218:
|
|
Linksys Fast Ethernet PCCARD cards supported by the ed driver
|
|
now require the addition of flag 0x80000 to their config line
|
|
in pccard.conf(5). This flag is not optional. These Linksys
|
|
cards will not be recognized without it.
|
|
|
|
20001205:
|
|
Important new FreeBSD-version stuff: PAM support has been worked
|
|
in, partially from the "Unix" OpenSSH version. This requires
|
|
adding the following in pam.conf:
|
|
|
|
sshd auth sufficient pam_skey.so
|
|
sshd auth required pam_unix.so try_first_pass
|
|
sshd session required pam_permit.so
|
|
|
|
20001031:
|
|
cvs updated to 1.11.
|
|
|
|
20001020:
|
|
The random device needs more entropy, so you need to make sure
|
|
that you've run mergemaster to get a /etc/rc which will seed
|
|
/dev/random. If you don't and the system hangs after ldconfig,
|
|
then banging on the keyboard randomly until it unhangs is one
|
|
workaround.
|
|
|
|
20001010:
|
|
****************************** WARNING ******************************
|
|
Sendmail has been updated.
|
|
****************************** WARNING ******************************
|
|
o mail.local(8) is no longer installed as a set-user-id binary.
|
|
o sendmail(8) is now built with STARTTLS support unless NO_OPENSSL
|
|
is set.
|
|
o The default /etc/mail/sendmail.cf disables the SMTP EXPN and VRFY
|
|
commands.
|
|
o Now using sendmail's version of vacation(1).
|
|
o The sendmail cf building tools (contrib/sendmail/cf) are installed
|
|
in /usr/share/sendmail/cf.
|
|
o sendmail.cw changed to local-host-names
|
|
|
|
More details can be found at
|
|
http://people.freebsd.org/~imp/UPDATING/sendmail-20001010
|
|
|
|
20001009:
|
|
The ports tree's new layout is in place. Be sure to update
|
|
your entire ports tree, or you will have problems.
|
|
|
|
20001006:
|
|
The perl build procedure no longer installs miniperl, nor uses
|
|
the installed miniperl. It is recommended that you delete
|
|
/usr/bin/miniperl.
|
|
|
|
20001005:
|
|
This weekend the ports tree will be updated to a new layout.
|
|
It will be in an inconsistent state until noted in the UPDATING
|
|
file, or with asami-san's message to the relevant mailing
|
|
lists. With this new layout, you'll need to update the whole
|
|
tree for anything to work.
|
|
|
|
20000928:
|
|
There was a change in the passwd format. Need more information.
|
|
|
|
20000916:
|
|
/boot/kernel/kernel.ko -> /boot/kernel/kernel change has taken
|
|
place. Please update boot loader (not the boot blocks) at the
|
|
same time as your kernel.
|
|
|
|
20000914:
|
|
The new pmtimer device is necessary for laptops. Failure to
|
|
include the device will cause suspended laptops losing time
|
|
when they resume. Include
|
|
device pmtimer
|
|
in your config file and
|
|
hint.pmtimer.0.at="isa"
|
|
to your /boot/device.hints file.
|
|
|
|
20000911:
|
|
The random device has been turned into a (pseudo-)device,
|
|
rather than an option. The supplied kernel config files have
|
|
been updated. You will need to do something similar in your
|
|
own kernel config file.
|
|
Remove:
|
|
options RANDOMDEV
|
|
Add:
|
|
device random
|
|
If you prefer to load the loadable module, you need to do
|
|
nothing.
|
|
|
|
20000909:
|
|
The random device module has been renamed from randomdev.ko to
|
|
random.ko. You will need to edit your /boot/loader.conf to
|
|
reflect this if you load this module at boot time.
|
|
The line should read:
|
|
random_load="YES"
|
|
|
|
20000907:
|
|
The SMPNG commit has happened. It should work, but if it
|
|
doesn't, fallback to the PRE_SMPNG CVS tag. There are likely
|
|
to be a variety of minor issues. Please see 20000905 to make
|
|
sure you don't have model loading problems which might at
|
|
first blush appear related to SMP.
|
|
|
|
20000906:
|
|
nsswitch has been imported from NetBSD. Among other things,
|
|
this means that /etc/host.conf is no longer used. See
|
|
nsswitch.conf(5) instead. Note that at boot time rc.network
|
|
will attempt to produce a new nsswitch.conf file for you if you
|
|
don't have one, and you have host.conf.
|
|
|
|
20000905:
|
|
The ucred structure changed size. This breaks the interface
|
|
that mountd uses. Trying to use an older mountd with a newer
|
|
kernel guarantees a panic. This means that you need to use
|
|
kernels newer than today only with matching mountd, but you
|
|
needed to do that anyway with the boot loader changes.
|
|
|
|
20000905:
|
|
The boot loader has been updated. The new default kernel is
|
|
now /boot/kernel/kernel.ko. The new default module location
|
|
is /boot/kernel.
|
|
|
|
You *MUST* upgrade your boot loader and kernel at the same time.
|
|
The easiest way to do this is to do the buildworld/buildkernel/
|
|
installkernel/installworld dance.
|
|
|
|
Furthermore, you are urged to delete your old /modules directory
|
|
before booting the new kernel, since kldload will find stale
|
|
modules in that directory instead of finding them in the correct
|
|
path, /boot/kernel. The most common complaint that this cures
|
|
is that the linux module crashes your machine after the update.
|
|
|
|
if [ ! -d /boot/kernel.old ]; then
|
|
mv /modules.old /boot/kernel.old
|
|
chflags noschg /kernel.old
|
|
mv /kernel.old /boot/kernel.old/kernel.ko
|
|
chflags schg /boot/kernel.old/kernel.ko
|
|
fi
|
|
|
|
20000904:
|
|
A new issue with the sendmail upgrade has come to light.
|
|
/etc/aliases has moved to /etc/mail/aliases. Mergemaster will
|
|
incorrectly install the default aliases in /etc/mail rather than
|
|
move the old one from /etc. So you'll need to manually move the
|
|
file, create a symbolic link, remove the old /etc/aliases.db and
|
|
run newaliases. For safety sake, you should stop sendmail
|
|
while doing this and run the upgrade when locally sourced email
|
|
is not likely to be generated.
|
|
|
|
20000825:
|
|
/boot/device.hints is now required for installkernel to
|
|
succeed. You should copy GENERIC.hints for your architecture
|
|
into /boot/device.hints. If and only if you compile hints
|
|
into your kernel, then this file may be empty. Please note,
|
|
if you have an empty or missing /boot/device.hints file and
|
|
you neglected to compile hints into your kernel, no boot
|
|
messages will appear after the boot loader tries to start the
|
|
kernel.
|
|
|
|
20000821:
|
|
If you do NOT have ``options RANDOMDEV'' in your kernel and
|
|
you DO want the random device then add randomdev_load="YES" to
|
|
/boot/loader.conf.
|
|
|
|
20000812:
|
|
suidperl is now always built and installed on the system, but
|
|
with permissions of 511. If you have applications that use
|
|
this program, you are now required to add ENABLE_SUIDPERL=true
|
|
to /etc/make.conf. If you forget to do this,
|
|
chmod 4511 /usr/bin/suidperl
|
|
will fix this until the next build.
|
|
|
|
20000812:
|
|
sendmail has been updated from 8.9.3 to 8.11.0. Some of the more
|
|
visible changes that may immediately affect your configuration
|
|
include:
|
|
- New default file locations from src/contrib/sendmail/cf/README
|
|
- newaliases limited to root and trusted users
|
|
- MSA port (587) turned on by default
|
|
- New queue file naming system so can't go from 8.11 -> 8.9
|
|
- FEATURE(`rbl') renamed to FEATURE(`dnsbl')
|
|
- FEATURE(`nullclient') is more full featured
|
|
- FEATURE(`nouucp') requires an argument: `reject' or `nospecial'
|
|
- mail.local FreeBSD-only -b option changed to -B
|
|
- See src/contrib/sendmail/RELEASE_NOTES for more info
|
|
|
|
20000810:
|
|
suidperl (aka sperl) is no longer build by default. You must
|
|
specifically define BUILD_SUIDPERL to "true" for it to be build.
|
|
Furthermore, we recommend that you remove /usr/bin/sperl* and
|
|
/usr/bin/suidperl files from your system unless you have a
|
|
specific use for it.
|
|
|
|
20000729:
|
|
Networking defaults have been tightened. Anybody upgrading
|
|
/etc/defaults/rc.conf needs to add the following lines to
|
|
/etc/rc.conf if they want to have the same setup
|
|
afterwards (unless the variables already are set, of course):
|
|
# Enable network daemons for user convenience.
|
|
inetd_enable="YES"
|
|
portmap_enable="YES"
|
|
sendmail_enable="YES"
|
|
|
|
20000728:
|
|
If you have null_load="YES" in your /boot/loader.conf, you
|
|
will need to change that to nullfs_load="YES".
|
|
|
|
20000728:
|
|
The "installkernel" target has changed slightly. Now even if
|
|
you override KERNEL e.g. 'make installkernel KERNEL=MYKERNEL'
|
|
it will install the MYKERNEL file (built with the buildkernel
|
|
target) as /kernel rather than /MYKERNEL. Those who have
|
|
updated their /boot/loader.conf files to point to /MYKERNEL
|
|
should remove that entry or perform manual rename of /kernel
|
|
to /MYKERNEL.
|
|
|
|
20000711:
|
|
If you use CVSUP or CTM to get CVS trees, AND you used to get
|
|
the old crypto files from internat.freebsd.org AND you check
|
|
out files from the CVS tree with the cvs command, please read
|
|
http://people.freebsd.org/~imp/internat.txt
|
|
for details on potential problems that you might have and how
|
|
to get around them.
|
|
|
|
If you are merely a mirror, or don't answer yes to each of the
|
|
clauses above, you needn't worry.
|
|
|
|
20000711:
|
|
/etc/security has been updated to print the inode number of
|
|
setuid programs that have changed. You will see a large spike
|
|
in the number of changed programs the first time when you run
|
|
mergemaster to get a new /etc/security.
|
|
|
|
20000710:
|
|
/dev/random now has good entropy collection (from the keyboard
|
|
and sysmouse drivers). Please ensure that either `options
|
|
RANDOMDEV' is present in your kernel config file or that
|
|
`randomdev_load="YES"' is in your /boot/loader.conf. If you do
|
|
not have the /dev/random driver, OpenSSL (and consequently
|
|
lots of crypto tools (like SSH)) will fail with strange
|
|
errors. (see below, 20000624).
|
|
|
|
FreeBSD-current is safe again to run Crypto.
|
|
|
|
20000709:
|
|
phk made the malloc default options AJ. This may slow things
|
|
down and uncover other latent bugs in the code. If you need to
|
|
run at full speed, you can disable this by doing the following:
|
|
ln -s aj /etc/malloc.conf
|
|
|
|
20000706:
|
|
libftpio's version was accidentally bumped a few days ago. This
|
|
has been corrected. You may need to remove /usr/lib/libftpio.so.6
|
|
before doing your next buildworld/installworld pair. It certainly
|
|
won't hurt to remove it before the update procedure. It will
|
|
break fetch until a new one is built, but ftp can be used in the
|
|
interim if needed.
|
|
|
|
20000705:
|
|
The crypto packages have changed for the cvsup. This has been done
|
|
in a backward compatible way, but the old packages will go away at
|
|
some point in the future. Look at /usr/share/examples/cvsup for
|
|
details.
|
|
|
|
20000704:
|
|
With the new sys/modules/sound/drivers/*, you will need to
|
|
set SYSDIR until you do an installworld after July 7th.
|
|
|
|
20000704:
|
|
rc.shutdown and rc will now call the rc.d scripts with start
|
|
or stop. This may cause some harmless warnings from older
|
|
rc.d scripts that haven't been updated.
|
|
|
|
20000630:
|
|
The libfetch based version of fetch has gone into the tree.
|
|
Minor problems may result on some of the less popular sites,
|
|
which should be reported to des@freebsd.org.
|
|
|
|
20000625:
|
|
From approximately this date forward, one must have the crypto
|
|
system installed in order to build the system and kernel.
|
|
While not technically strictly true, one should treat it as
|
|
required and grab the crypto bits. If you are grabbing CVS
|
|
trees, src-all and cvs-crypto should be treated as if they
|
|
were required. You should check with the latest collections
|
|
to make sure that these haven't changed.
|
|
|
|
20000624:
|
|
Mark Murray just committed the first parts of a cleanup of
|
|
/dev/zero, et al. This is also cleaning up /dev/random.
|
|
The entropy is disconnected, so DO NOT USE VERSIONS OF FREEBSD
|
|
-CURRENT FROM THIS POINT to 2000710 for cryptographic services
|
|
until Mark can merge in the fixes to this work in progress.
|
|
openssh and openssl should not be used to generate keys from this
|
|
date to the completion of the work.
|
|
|
|
If you must operate at this reduced level of security, add '
|
|
options RANDOMDEV' to your kernel or modload the randomdev
|
|
module. You may also need to copy a new MAKEDEV to /dev and
|
|
recreate the random and urandom devices.
|
|
|
|
20000622:
|
|
The license on the softupdates is now a standard 2 clause
|
|
BSD license. You may need to remove your symbolic links
|
|
that used to be required when updating.
|
|
|
|
20000621:
|
|
Scott Flatman <sf@aracnet.com> sent in a decent write-up on
|
|
the config file update procedure.
|
|
http://people.freebsd.org/~imp/config-upd.html
|
|
NOTE: LINT is gone. It has been replaced with NOTES. NOTES
|
|
isn't buildable. However, you can generate a LINT file.
|
|
|
|
20000620:
|
|
Binutils 2.10 have hit the tree, or will shortly. As soon
|
|
as they do, the problem noted in 20000522 will be resolved and
|
|
that workaround will no longer be required.
|
|
|
|
20000615:
|
|
phk removed the compatibility creation of wd devices in the
|
|
ad driver. If you haven't done so already, you must update
|
|
your fstab, etc to use the ad devices instead of the wd
|
|
devices.
|
|
|
|
In addition, you'll need to update your boot blocks to a
|
|
more modern version, if you haven't already done so. Modern
|
|
here means 4.0 release or newer (although older releases
|
|
may work).
|
|
|
|
20000612:
|
|
Peter took an axe to config(8). Be sure that you read his mail
|
|
on the topic before even thinking about updating. You will
|
|
need to create a /boot/device.hints or add a hints directive
|
|
to your config file to compile them in statically. The format
|
|
of the config file has changed as well. Please see GENERIC or
|
|
NEWCARD for examples of the new format.
|
|
|
|
Indirectly, this also breaks USERCONFIG. Unless a newer entry
|
|
says that it has been fixed, assume that must use the hints mechanism
|
|
in the loader if you need to use a machine with very old ISA cards
|
|
in it.
|
|
|
|
20000522:
|
|
A new set of binutils went into the tree today. Anybody
|
|
building a kernel after this point is advised that they need
|
|
to rebuild their binutils (or better yet do a
|
|
buildworld/installworld) before building a new kernel.
|
|
|
|
Due to bugs in binutils, using malloc options (eg /etc/malloc.conf
|
|
or MALLOC_OPTIONS env var) J will cause ld to dump core. It
|
|
is recommended that you don't set this option until the problem
|
|
is resolved.
|
|
|
|
20000513:
|
|
The ethernet drivers were all updated to clean up the BPF handling.
|
|
|
|
20000510:
|
|
The problems with boot blocks on the alphas have been corrected.
|
|
This will require some care in updating alphas. A new libstand
|
|
is requires for the boot blocks to build properly.
|
|
|
|
20000503:
|
|
Recompile all kld modules. Proper version dependency info
|
|
is now available.
|
|
|
|
20000502:
|
|
Modules have been disconnected from the buildworld tree and
|
|
connected to the kernel building instead.
|
|
|
|
20000427:
|
|
You may need to build gperf
|
|
cd /usr/src/gnu/usr.bin/gperf && make depend all install
|
|
when upgrading from 4.0 -> current. The build system now uses
|
|
an option only in -current.
|
|
|
|
20000417:
|
|
The method that we brand ELF binaries has changed to be more
|
|
acceptable to the binutils maintainers. You will need to
|
|
rebrand your ELF binaries that aren't native. One problem
|
|
binary is the Linux ldconfig. After your make world, but
|
|
before you reboot, you'll need to issue:
|
|
brandelf -t Linux /compat/linux/sbin/ldconfig
|
|
if you have Linux compatibility enabled on your machine.
|
|
|
|
20000320:
|
|
If you have really bad/marginal IDE drives, you may find they
|
|
don't work well. Use pio mode instead. The easiest way to
|
|
cope if you have a problem combination is to add:
|
|
/sbin/sysctl hw.ata.ata_dma=0
|
|
to the start of /etc/rc.conf.
|
|
|
|
20000319:
|
|
The ISA and PCI compatibility shims have been connected to the
|
|
options COMPAT_OLDISA and COMPAT_OLDPCI. If you are using old
|
|
style PCI or ISA drivers (i.e. tx, voxware, etc.) you must
|
|
include the appropriate option in your kernel config. Drivers
|
|
using the shims should be updated or they won't ship with
|
|
5.0-RELEASE, targeted for 2001.
|
|
|
|
20000318:
|
|
We've entered the traditional post release dumping party.
|
|
Large kernel changes are being committed and are in the
|
|
works. It is important to keep the systems' klds and kernel
|
|
in sync as kernel interfaces and structures are changing.
|
|
Before reporting kernel panics, make sure that all modules
|
|
that you are loading are up to date.
|
|
|
|
20000315:
|
|
If you are upgrading from an older version of FreeBSD, you
|
|
need to update your boot blocks as well. 'disklabel -B ad0'
|
|
will do the trick. This isn't critical until you remove your
|
|
wd device entries in /dev, at which point your system will not
|
|
boot.
|
|
|
|
20000315:
|
|
4.0 RELEASE shipped. Please see the 4.0 UPDATING file for how
|
|
to upgrade to 4.0 from 3.x.
|
|
|
|
COMMON ITEMS:
|
|
|
|
General Notes
|
|
-------------
|
|
Avoid using make -j when upgrading. From time to time in the
|
|
past there have been problems using -j with buildworld and/or
|
|
installworld. This is especially true when upgrading between
|
|
"distant" versions (eg one that cross a major release boundary
|
|
or several minor releases, or when several months have passed
|
|
on the -current branch).
|
|
|
|
Sometimes, obscure build problems are the result of environment
|
|
poisoning. This can happen because the make utility reads its
|
|
environment when searching for values for global variables.
|
|
To run your build attempts in an "environmental clean room",
|
|
prefix all make commands with 'env -i '. See the env(1) manual
|
|
page for more details.
|
|
|
|
To build a kernel
|
|
-----------------
|
|
If you are updating from a prior version of FreeBSD (even one just
|
|
a few days old), you should follow this procedure. With a
|
|
/usr/obj tree with a fresh buildworld,
|
|
make buildkernel KERNCONF=YOUR_KERNEL_HERE
|
|
make installkernel KERNCONF=YOUR_KERNEL_HERE
|
|
|
|
To just build a kernel when you know that it won't mess you up
|
|
--------------------------------------------------------------
|
|
cd src/sys/{i386,alpha}/conf
|
|
config KERNEL_NAME_HERE
|
|
cd ../../compile/KERNEL_NAME_HERE
|
|
make depend
|
|
make
|
|
make install
|
|
|
|
If this fails, go to the "To build a kernel" section.
|
|
|
|
To rebuild everything and install it on the current system.
|
|
-----------------------------------------------------------
|
|
# Note: sometimes if you are running current you gotta do more than
|
|
# is listed here if you are upgrading from a really old current.
|
|
|
|
<make sure you have good level 0 dumps>
|
|
<maybe fix /etc/fstab> [7]
|
|
make buildworld
|
|
make buildkernel KERNCONF=YOUR_KERNEL_HERE
|
|
make installkernel KERNCONF=YOUR_KERNEL_HERE
|
|
[1]
|
|
<reboot in single user> [3]
|
|
mergemaster -p [5]
|
|
make installworld
|
|
mergemaster [4]
|
|
<reboot>
|
|
|
|
|
|
To upgrade from 4.x-stable to current
|
|
-------------------------------------
|
|
# 5.x uses more space than 4.x. Also, the location of kernel
|
|
# modules has changed. If you are installing 5.0 onto a 4.x
|
|
# system, you'll need about 30MB of free disk space on your /
|
|
# partition. If you have less than this, you may encounter difficult
|
|
# to back out of problems with this proceedure. If /tmp is on
|
|
# the / partition, you may want to completely remove all its content
|
|
# before upgrading, as this can be a common source of shortage of
|
|
# space on /.
|
|
|
|
<make sure you have good level 0 dumps>
|
|
<maybe fix /etc/fstab> [7]
|
|
make buildworld
|
|
make buildkernel KERNCONF=YOUR_KERNEL_HERE
|
|
cp src/sys/${MACHINE_ARCH}/conf/GENERIC.hints /boot/device.hints [2]
|
|
make installkernel KERNCONF=YOUR_KERNEL_HERE
|
|
cd src/sys/boot ; make install [6]
|
|
[1]
|
|
<reboot in single user> [3]
|
|
mergemaster -p [5]
|
|
rm -rf /usr/include/g++
|
|
make installworld
|
|
mergemaster [4]
|
|
<reboot>
|
|
|
|
Make sure that you've read the UPDATING file to understand the
|
|
tweaks to various things you need. At this point in the life
|
|
cycle of current, things change often and you are on your own
|
|
to cope. The defaults can also change, so please read ALL of
|
|
the UPDATING entries.
|
|
|
|
Also, if you are tracking -current, you must be subscribed to
|
|
freebsd-current@freebsd.org. Make sure that before you update
|
|
your sources that you have read and understood all the recent
|
|
messages there. If in doubt, please track -stable which has
|
|
much fewer pitfalls.
|
|
|
|
[1] If you have third party modules, such as vmware, you
|
|
should disable them at this point so they don't crash your
|
|
system on reboot.
|
|
|
|
[2] If you have legacy ISA devices, you may need to create
|
|
your own device.hints to reflect your unique hardware
|
|
configuration.
|
|
|
|
[3] From the bootblocks, boot -s, and then do
|
|
fsck -p
|
|
mount -u /
|
|
mount -a
|
|
cd /usr/src
|
|
adjkerntz -i # if CMOS is wall time
|
|
Also, when doing a major release upgrade, it is required that
|
|
you boot into single user mode to do the installworld.
|
|
|
|
[4] Note: This step is non-optional. Failure to do this step
|
|
can result in a significant reduction in the functionality of the
|
|
system. Attempting to do it by hand is not recommended and those
|
|
that pursue this avenue should read this file carefully, as well
|
|
as the archives of freebsd-current and freebsd-hackers mailing lists
|
|
for potential gotchas.
|
|
|
|
[5] Usually this step is a noop. However, from time to time
|
|
you may need to do this if you get unknown user in the following
|
|
step. It never hurts to do it all the time. You may need to
|
|
install a new mergemaster (cd src/usr.sbin/mergemaster && make
|
|
install) after the buildworld before this step if you last updated
|
|
from current before 20020224 or from -stable before 20020408.
|
|
|
|
[6] 4.x boot blocks can be used to boot a 5.x system, however
|
|
it is difficult to do that at best. If you are concerned
|
|
about being able to boot 4.x if the reboot to single user
|
|
fails, then you should do something like
|
|
mv /boot /boot-4x
|
|
mkdir /boot
|
|
before this step. You may be able to then boot the 4.x boot
|
|
blocks if you catch boot2, and enter /boot-4x/loader rather than
|
|
/boot/loader. However, this step hasn't been torture tested.
|
|
|
|
[7] Before you upgrade, please make sure that you are not using
|
|
compatibility slices. These are device names of the form, on i386
|
|
and other architectures that use MBR slicing, /dev/ad0a without the
|
|
actual slice name. Chances are excellent that these will break.
|
|
You generally must update these entries to use the post FreeBSD
|
|
2.x form of /dev/ad0s1a.
|
|
|
|
FORMAT:
|
|
|
|
This file contains a list, in reverse chronological order, of major
|
|
breakages in tracking -current. Not all things will be listed here,
|
|
and it only starts on March 15, 2000. Updating files can found in
|
|
previous releases if your system is older than this.
|
|
|
|
Copyright information:
|
|
|
|
Copyright 1998, 2002 M. Warner Losh. All Rights Reserved.
|
|
|
|
Redistribution, publication, translation and use, with or without
|
|
modification, in full or in part, in any form or format of this
|
|
document are permitted without further permission from the author.
|
|
|
|
THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
|
|
IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
|
|
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
|
|
DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
|
|
INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
|
|
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
|
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
|
|
STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
|
|
IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
POSSIBILITY OF SUCH DAMAGE.
|
|
|
|
If you find this document useful, and you want to, you may buy the
|
|
author a beer.
|
|
|
|
Contact Warner Losh if you have any questions about your use of
|
|
this document.
|
|
|
|
$FreeBSD$
|