Goetz Neumann wrote,

>  John Burke wrote:
> > Here is the setting. MPE/iX 6.5 PP3+. I'm trying to install
> the latest
> > TurboIMAGE patch, TIXMCX3B, by staging it.
>
> I assume that was the only patch that you installed, and that
> your system was not booted from any other staging area, but
> from a (committed) base ?

Correct.

>
> > Of course I downloaded the latest version of Patch/iX (you
> have to now
>
> Would you let us know which version you got ?
>

Patch/iX Version B.01.09 downloaded from ITRC 3/12.

> >  ERROR  - from hpdir_find_path in stage_get_syslib_ufid
> >           status - subsys: #518  info: # -1
> > File name = /SYS/hpstage/TIXMXC3B/NL.PUB.SYS
> > Cannot find NL in the new staging area.
> > Use ISL> STARTBASE NORECOVERY to recover.
> >
> > SYSTEM ABORT 16 FROM SUBSYSTEM 146
> > SECONDARY STATUS:  INFO = -1, SUBSYS = 518
> > SYSTEM HALT 7, $0010
> >
...

> >
> > Anyone ever see something like this before?
>
> Not exactly like that, no.
> A similar SA16 would occur if the /SYS/hpstage directory was
> relocated off LDEV 1 (during a INSTALL/RELOAD), but that only
> happened back in MPE 5.5 days. Since 6.0 /SYS/hpstage has a
> volume restriction pinning it to LDEV 1 when recovered by a
> RESTORE.
> I know of a defect where Patch/iX missed to build a new NL
> and START on a subsys_only patch, when that patch was installed
> 'cumulative' on top of an existing (and booted from) stage,
> but IIRC this was fixed long ago. AFAIR that only happened
> with patches that had no system library components (e.g.
> Linkedit or FTP).
>
> It is relatively seldom that Patch/iX(+ Stage/iX) is used to install
> just one non_NL patch, so it is possible that this is a defect
> that has not been uncovered so far.
>
> Design-wise I believe it should be possible to stage pure SUBSYS
> patches which do not require a new NL and a new START image: either
> the code should be smart enough to detect a 'non_NL' stage and use
> the original, unmodified START/NL files, or it should be smart enough
> to always copy the NL and make a new START image ( a bit of a waste
> of disk space on the latter approach, IMO).
>
> Assuming that you can still produce such a NL-missing stage after
> re-initializing your staging areas (stageman uninstall + init) then
> please report this to your Response Center, and have your support
> contact pass this on as an investigation about a possible defect
> to the Expert Center for MPE.

Actually, this is exactly the process I followed prior to attempting the
patch (stageman uninstall + init). The problem has been reported and is
3201577016. I will forward your comments.

In this particular case, LDEV1 is the entire MPEXL_SYSTEM_VOLUME_SET so
migration of anything off LDEV1 should not be an issue.

Since this is a test machine, I decided to do the obvious based upon the
error message, copy NL.PUB.SYS into the staging area and retry (the ITRC had
already confirmed that NL.PUB.SYS was not updated by the patch). This seems
to have fixed the problem, though I booted back to BASE after rudimentary
testing.

Thus, my WAG is that there is a defect in Patch/iX when it creates the
staging area (it claimed to be creating a new start PME) or, worse, in the
start-up code that processes staging areas.

>
> HTH,
> Goetz.
>

Thanks for the feedback.

John

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *