fresh5.0.7, can't install MP1
Brian K. White
bare bones.
Just the OS and the intel pro/1000 driver. No patches.
The readme for MP1 states that it supercedes oss 646, 656, and 631, so
as far as I could tell, now, the first thing to install should now be
mp1, so that's what I tried. from:
ftp://ftp.sco.com/pub/openserver5/osr507mp/
However, I can't install it.
I get this error:
(\) The custom+ binary (backend) process has died.
the "Details" button on that errorbox yeilds:
Assertion failed: path != NULL, file "dbaseStanza.C", line 484
after OK'ing the error, custom exits immediately (the kernel is not
rebuilt which the readme indicates is supposed to happen during this
update)
I un-installed the updates, but the system remains somewhat broken
since I didn't check the check-box to keep the replaced items on disk,
I tried to re-install but manually selecting only part of the update,
I managed to (apparently) sucessfully install just the rs507 portion.
that install went without error and it did do a kernel link along the
way.
after rebooting,
docview's httpd fails to start
and sshd fails to start due to missing /usr/lib/libz.so.1
I get the same error when trying to install any of the other portions
of MP1 individualy.
I also get the same error when trying to install gwxlibs seperately
from:
ftp://ftp.sco.com/pub/openserver5/opensrc/gwxlibs-1.3.2Ag/
I don't suppose anyone else has run into this, or else I'd have seen
posts about it by now... I was hoping to avoid re-installing the OS
myself. This would make 2nd or 3rd 5.0.7 box in the last year that
I'll have had to re-install for different reasons. Not a good track
record so far. I can't imagine it's the vendor botching the installs
since I've had occasion to talk a little with a couple different
techs, and they seem pretty together to me, but these were only very
brief phone conversations and emails.
Mike Brown
I have seen the error on the gwxlibs, not sure what caused it. I checked
the dbaseStanza.c file at the point the install flaged the failure
message, and it definitely was corrupted at the line indicated.
I played a bit more with the VOLS, and best that I can tell the
routine that uncompresses the bits and parts was generating the corrupt
file. If I copied the compressed parts over to a 5.0.6a machine they
would uncompress fine, but not on the 5.0.7mp1 machine.
I copied an entire directory of compress programs over to the 5.0.7mp1
machine, ran sum and verified they where all okay, then uncompressed
them all and ran sum again. One program did not uncompress properly.
A reload of the same machine, with numerous relinks and reboots ( basically
after every patch ) loaded everything perfectly and had no uncompress
problems.
> I don't suppose anyone else has run into this, or else I'd have seen
> posts about it by now... I was hoping to avoid re-installing the OS
> myself. This would make 2nd or 3rd 5.0.7 box in the last year that
> I'll have had to re-install for different reasons. Not a good track
> record so far. I can't imagine it's the vendor botching the installs
> since I've had occasion to talk a little with a couple different
> techs, and they seem pretty together to me, but these were only very
> brief phone conversations and emails.
Funny, I checked for posts and came to the same decision.
WAG, but maybe a loaded program/lib and a memory resident program/lib
are getting out of sync.
After getting the box loaded correctly, I ran an ftp/sum/compress/uncompress
script for a couple of days and no errors to report.
Is your box still corrupted?
Mike
--
Michael Brown
The Kingsway Group
Brian K. White
Yes, worse than that,
I just finished a fresh, scratch install, with a different divvy
layout so I know the fs's were new, and it failed at the exact same
spot in the same way.
I installed the base os, (which required a aacraid btld)
upon completion oss656B was already installed
(It wasn't when shipped from the vendor.)
so, according to mp1's clear instructions I uninstalled 656b
prior to installing mp1
(possible glitch: I didn't reboot after the kernel link)
then installed mp1, (selecting the entire package)
The install failed with the same error and did not do a kernel link
afterwards, restarting custom shows mp1 as installed, but really it's
not fully installed.
I just spoke to a tech from the vendor and he confirmed they've been
using an old copy of a 507cd to do installs so that explains why 656b
was not installed from them yet my cd installed it by itself. Said
tech said he has only installed mp1 once and it went right through
fine, but they have recently updated their procedure such that from
now on they are going to be installing mp1 on all new 507 boxes by
default. So he doesn't have any ideas yet but will "ask around" in
their resources.
The only funny thing I did during this install was that I loaded smp
and the linker before applying mp1.
I guess I'll try again without doing that, and rebooting after each
kernel link no matter what the directions say.
I probably won't ever need smp or a compiler on this particular box,
but the next box I do I may not be able to leave them out.
John Boland
> I just received a pre-installed 5.0.7 box, but the installation was
> bare bones.
> Just the OS and the intel pro/1000 driver. No patches.
>
> The readme for MP1 states that it supercedes oss 646, 656, and 631, so
> as far as I could tell, now, the first thing to install should now be
> mp1, so that's what I tried. from:
> ftp://ftp.sco.com/pub/openserver5/osr507mp/
Brian,
The MP1 release notes state:
"Before installing the Maintenance Pack, you should remove
OSS646A/OSS646B and OSS656A/OSS656B. It is not necessary to first
remove any of the OSS631 supplements.
When you remove the recommended supplements, you do not need to reboot
the system after the kernel is re-linked. The Maintenance Pack
installation also re-links the kernel -- you can reboot at that point.
"
Let me know if this is the problem.
Thanks,
John
Brian K. White
> Mike Brown <mi...@tkg.ca> wrote in message news:<3FFB06C7...@tkg.ca>...
> > "Brian K. White" wrote:
> > >
> > > I just received a pre-installed 5.0.7 box, but the installation was
> > > bare bones.
> > > Just the OS and the intel pro/1000 driver. No patches.
> > >
> > > The readme for MP1 states that it supercedes oss 646, 656, and 631, so
> > > as far as I could tell, now, the first thing to install should now be
> > > mp1, so that's what I tried. from:
> > > ftp://ftp.sco.com/pub/openserver5/osr507mp/
> > >
> > > However, I can't install it.
> > > I get this error:
> > > (\) The custom+ binary (backend) process has died.
> > >
> > > the "Details" button on that errorbox yeilds:
> > >
> > > Assertion failed: path != NULL, file "dbaseStanza.C", line 484
> > >
> > > after OK'ing the error, custom exits immediately (the kernel is not
> > > rebuilt which the readme indicates is supposed to happen during this
> > > update)
> > >
I think I figured it out...
If anyone else ever has this problem, check your bios and disable the
$#@!% hyper-threading.
What a pain in the ..... this was.
I'm going to have words with my vendor. they are not a small company
and they should know this ever since HT appeared.
FyRE
It's not likely to ever be fixed though since SCO became a legal firm
instead of a software company. As hardware evolves and SCO's already
ancient product line stagnates even further these types of problem are
only likely to increase in frequency. What a shame <chuckle>...
Linux has full HT support, don't you know ;-) Surprising that SCO
doesn't, since, you know, Linux obviously stole it all from Darl and
co ;-)))
--
FyRE < "War: The way Americans learn geography" >
Mike Brown
Hmmm, unsupported HT breaks the uncompress process. Subtle failure.
I think there may be something else wrong, since I have seen this failure
while loading the new GWXLIBS, APACHE, PERL on a 5.0.6a machine that
does not have a HT CPU ( Pentium III 550 I think). Time will tell.
Mike Brown
OSR 5.0.7up1 has support for HT, but not with the CD install kernel.
So, HT needs to be disabled until the kernel is linked with support
for it.
Interestingly, HT may or may not provide any performance improvement.
FyRE
Well luckily this issue was discussed and benchmarked recently.
Apparently HT does make a very BIG difference when running under an OS
that can make use of it...
http://developers.slashdot.org/article.pl?sid=04/01/07/058251&mode=thread&tid=118&tid=137&tid=187
Brian K. White
Nahh, I read those and did it just like it said the 1st through 5th
times.
read my original post.
It's working now and I beleive the culprit was that HyperThreading was
enabled in the motherboard bios. (pre-built, pre-installed box)
Thanks.
Bill Vermillion
Just a comment. I've seen problems on other OSes during installs
with HTT enabled. I've just made it a habit to turn off HTT when
installing or doing upgrades then re-enabling after done.
I will say that I have not run HTT on an SCO system however - but
even in XP land certain things really don't like to have HTT on
when installing.
Bill
--
Bill Vermillion - bv @ wjv . com