I should mention first that I'm rather new both to the 'IBM world',
and to ADSM-L; also we're not using ADSM for too long yet -
so, if what I'm writing below is 'all normal', just let me know ...

I've lately learned about APAR IX87066 (dated January '96),
which would be real funny for its reply being all besides the point,
if it only didn't inhibit the fixing of a rather blatant bug in the
UNIX ADSM client code.

Paraphrased error description:
        UNIX `dsmc' crashes when during backup of a filesystem,
        it encounters a directory file there, whose name
        contains '*' or '?' wildcard characters.

Paraphrased problem conclusion:
        You cannot specify a directory whose name contains '*' or '?'
        wildcard characters, with any ADSM command, since wherever you
        pass a wildcard character to `dsmc', it'll be treated as such.

        We'll enhance the documentation to make this even clearer.

        This means you cannot ever expect `dsmc' to successfully backup
        UNIX directories whose names contain '*' or '?'.

(end of the APAR conclusion's far-fetched conclusion)

I can confirm that the problem (crash of `dsmc', _without_ any
error message pointing to the 'culprit', which was a directory file
named '???' in our case) exists as of today, and I can add
that the first occurence here cost me several days to analyze
and then bypass, since the affected file system (whose name
has no wildcard in it :-) happened to be pretty huge ...

I can also report that according to the IBM support person to whom
I described our problem, the APAR's conclusion isn't even correct,
since he got back to me with "I've just tried to reproduce this,
and `dsmc' did _not_ crash". (Hint: It seems like there got to exist
a non-empty directory whose name is 'matched' by the 'culprit' wildcard.
Here, `dsmc' crashed while backing up a file in a 'neighbour' directory
whose name was three letters, thus matched by '???' treated as a wildcard.
I'm wondering what "feature" would demand wildcard-matching code
within `dsmc' to ever be invoked, with the name of something to be backed up
used as a _pattern_). So I was allowed to send in the core dump
(a meager 100-odd MB - guess how long it took `dsmc' to arrive at
the point of crash, with us already resorting to MEMORYEFFICIENTBACKUP YES),
and received the APAR reference in return, plus advice to submit
a "feature request" ...

The chance of the problem re-occuring is distinctly non-zero here,
as we happen to backup client's files, and can't sensibly make
prescriptions as to what directory names _we_ or _IBM_ considered
'proper' for them to use, what `tar' files to never unpack, etc.

Allegedly competing backup products do not come with weird restrictions
like the one that could be inferred from the APAR's conclusion
(I didn't see it documented, though, in spite of the APAR's promise,
made 3 years ago ...
... now if the preceding sentence made no sense to you,
that's exactly how I feel about the APAR's logic :-)

IBM, do we really have to submit a "feature request" in order
to get this bug fixed? What were the chances if we did? What
if we did not? And besides, maybe you do want ADSM clients
to be used only in 'closed' shops?

Wolfgang J. Moeller, Tel. +49 551 2011516 or -510, moeller AT gwdvms.dnet.gwdg 
DOT de
GWDG, D-37077 Goettingen, F.R.Germany   |       Disclaimer: No claim intended!
<moeller AT decus.decus DOT de>  ----- <moeller AT gwdg DOT de>  -----  
<w.moeller AT ieee DOT org>