Well... One of my 3350s just became wild mad. Hopefully it was one I was
trying to copy a volume INTO, not FROM. I eventually decided to run DSF on it
and the poor thing choked lamentably... It assigned a dozen skip displacements,
about 5 alternate tracks, and then came up with a subtle philosophical state-
ment that would deserve to be kept in the AI book of records:
ICKxxxxx TOO MANY DATA CHECKS ON VOLUME
Thanks pal. That powerful conclusion is worth waiting 30 minutes. But... no,
DSF didn't abandon the task. It just restarted (without even telling me). Ano-
ther 15 minutes, and it had come to the same conclusion after reassigning the
skip displacements on the same tracks. Keep on the good job guy, try again...
But then it ran into a worse problem. The alternate tracks themselves were
defective... :-) I was wondering whether it would assign alternates to the
alternates, but it did not. Instead, it decided to switch from the end of the
volume to cylinder 0, all of a sudden. More good news:
ICKxxxxx CYL-0 TRK-0 DEFECTIVE, PACK UNUSABLE FOR IPL
I'll be running the microcode diags on the poor thing tomorrow. Meanwhile I
did not have time to write a decent error recovery routine for the DISTRIBUTE
function, so what it does when there is some inconsistency in the tables is it
just distributes to all the remaining recipients directly. That SURE can not
generate any loop :-)
As I said before I sent 1.5f out only because of the fixes. There is no doc
on the new features. Alan, the Reply-to= Both is implemented. Jose Maria, I
have not had time for the filemode in filelists. Udo, your STORE problems
should disappear with this version. Richard, the one-line "clipped" message
problem is fixed. Nick, you shouldn't have any more mail forwarded from the
UCLAMVS mailer, not anything sent to it. Harold, DELETE (GLOBAL is now implemen
ted. Dick, setting your release level to 1.05f will automatically disable the
DISTRIBUTE command and have an error message sent to the invoker, but you can
also, alternatively, force INFORM=MAIL on all files distributed by your server
by defining a new variable in LSV$PROF: DISTINFORM = 'MAIL' (defaults to
'USER') -- don't forget to put in the GLOBALV SET too :-) There is a new hard-
coded GET = NAD code, and a new GIVE command (GIVE userid@node fileid) for
file owners (it bypasses the givee's GET code checking, ie I can GIVE a file
to someone who is not allowed to GET it). I *** TEMPORARILY *** added a
"if link weight = 9 then link weight = 100" clause in LSVBITFD, but I'll do it
more neatly later and will also allow for 0-weight links (eg local CTCA link)
as someone asked me. Kent, Dave's problems with the GET command and the "#"
userid (or Dick's "RSCS" userid) from MVS should be solved too. Harold, the
Notify keyword is now used for the "cc: owner" portion of info mailfiles on
MOVE, DEL, ADD, SET, etc. Kent, Scott, when 1.5f has been installed by every-
body you'll be able to try this huge DISTRIBUTE job again and it should not
crash on GLOBALV's 255 chars limitation any more.
Sorry for the untidy format of the explanations but I'm in a real hurry.
Eric
|
