From: Andreas F. Borchert on
On 2006-12-22, Thommy M. Malmstr�m <thommy.m.malmstrom(a)gmail.com> wrote:
> Here's something that is totally screwed up. A patch should only
> fix/patch something that is already installed (*). A patch must never
> install new packages/programs/functionality.
>
> I hope this is a mistake...

That was likewise done in the past. Patch 112438-01 for Solaris 8,
for example, introduced /dev/random.

And this was a good thing as this eased the installation of OpenSSH
which requires /dev/random or some equivalent.

Andreas.
From: Thommy M. Malmström on
Andreas F. Borchert wrote:
> On 2006-12-22, Thommy M. Malmstr�m <thommy.m.malmstrom(a)gmail.com> wrote:
>> Here's something that is totally screwed up. A patch should only
>> fix/patch something that is already installed (*). A patch must never
>> install new packages/programs/functionality.
>>
>> I hope this is a mistake...
>
> That was likewise done in the past. Patch 112438-01 for Solaris 8,
> for example, introduced /dev/random.
>
> And this was a good thing as this eased the installation of OpenSSH
> which requires /dev/random or some equivalent.

Yeah, I know. And I wasn't happy back then either. For me it's a totally
bad idea to add functionality with patches, but maybe I'm just to damn
old... ;-)

In my world you end up with an uncontrolled system. IRL there are
actually companies trying to have control over what's installed.
From: Rich Teer on
On Fri, 22 Dec 2006, "Thommy M. Malmstr�m" wrote:

> Yeah, I know. And I wasn't happy back then either. For me it's a totally
> bad idea to add functionality with patches, but maybe I'm just to damn old...
> ;-)

I think it depends on the feature, and the binding that's requested
at the ARC. Some feature request patch binding, others Micro, and
so on.

So I think that although it's true that SOME (or even many) new
features become available in the form of patches, I don't think
ALL new features become available via patches.

Thinking about it a bit more, isn't the rule something like patches
can only change existing packages, not install new ones? For something
like /dev/random, the change was presumably made to SUNWcsr, which was
a preexisting package.

--
Rich Teer, SCNA, SCSA, SCSECA, OpenSolaris CAB member

. * * . * .* .
. * . .*
President, * . . /\ ( . . *
Rite Online Inc. . . / .\ . * .
.*. / * \ . .
. /* o \ .
Voice: +1 (250) 979-1638 * '''||''' .
URL: http://www.rite-group.com/rich ******************
From: Daniel Rock on
Rich Teer <rich.teer(a)rite-group.com> wrote:
> Thinking about it a bit more, isn't the rule something like patches
> can only change existing packages, not install new ones? For something
> like /dev/random, the change was presumably made to SUNWcsr, which was
> a preexisting package.

No,

see patch 122640 (SPARC), 122641 (x86): zfs genesis patch. Introduced the
new packages:
system SUNWzfskr ZFS Kernel (Root)
system SUNWzfsr ZFS (Root)
system SUNWzfsu ZFS (Usr)

(just extract the patch and do a "pkginfo -d payload")


or patch 113000 (only x86): SUNWgrub patch. Introduced the new package
system SUNWgrub GNU GRUB - GNU GRand Unified Bootloader


or the new (annoying) SUNWsmapi library introduced with 118833 (SPARC),
118855 (x86)


But these are all Solaris 10 examples. I don't have a Solaris 8/9 system here
so I cannot check these.


--
Daniel
From: Miroslav Zubcic on
Daniel Rock wrote:

> For the impatient:
>
> The patches are on the 11/06 DVD in the directory
>
> Solaris_10/UpgradePatches
>
> Just unzip all *.jar files in a directory and do a
>
> patchadd -M `pwd` 1*
>
> from that directory.
>
>
> But beware: Use at your own risk. In the past it worked for my for
> "upgrading" my 01/06 to 06/06, but I haven't tried the step to 11/06 yet.

I presume that this works only when upgrading from "vanilla" update
release to another (new) update release.

Yesterday, I have upgraded from 6/06 to 11/06, but first, I had to copy
and edit Solaris_10/Misc/analyze_patches from DVD and run it. It
produces the list of patches that must be removed before upgrade.
When I removed them, I was able to install with patchadd -M `pwd` *
without single error or warning (checked logs later too).

One also have to (re)install release package SUNWsolnm by hand from
Product directory to get accurate /etc/release - it is not in any patch.

From new features in 11/06 I can highlight enhanced showrev command:

# showrev -p
Segmentation Fault (core dumped)

Workaround is to edit /var/sadm/pkg/SUNWcsr/pkginfo and remove cca 160
lines from huge PATCH_INFO_118855-33=Installed header (remove patches
that are obsoleted by previous 118855 patch, I hope this is ok). Then
showrev -p will restore old pre-11/06 behaviour.

.... and of course even harder broken system shutdown procedure with
poweroff command that does not sync/umount filesystems in all cases.

Everything else works ok.


--
Man is something that shall be overcome.
-- Friedrich Nietzsche