🚀 go-pugleaf

We have been experiencing a maddeningly intermittent problem with two
HP LaserJet IIIs with HP's PostScript cartridge.  In brief, the
printers occasionally hang indefinitely while printing jobs, and have
to be reset or power-cycled before they will start printing again.
The problem does not seem to be bugs in the submitted PostScript
programs, because the exact same files will print perfectly after the
printer is reset.  I am posting in the hope that someone else might
have seen similar problems, and either solved them, or been able to
link them to known bugs in HP's PostScript.


Here are the long, gory details:

We have two LaserJet IIIs with HP's PostScript cartridge, hanging
off the MTI-1600 board of a Sun 3/160.  

Occasionally the printers hang (from 0 to 4 times a week, where each
printer sees 250-450 print jobs per week, totaling 1000-3000 pages).
"PROCESSING DATA" shows up in the message window, and stays there for
tens of minutes.  Because I have been able to print the identical jobs
after resetting the printer, I know that the problem is not just that
the PostScript program is doing a great deal of computation.

Sometimes I have been able to take the printer offline, reset it from
the front panel, and had it work fine afterwards.  More often, when I
try to take the printer offline, the message "OFFLINE PENDING" appears
on the screen, but the printer never goes offline.  In those cases, I
have had to turn the printer off and on to bring it back to the real
world.

Generally speaking, resubmitting the same job after resetting or
power-cycling the printer results in perfect output and no hitches.
The times that I have seen the printer hang again on the same job, it
was on a different page.

There doesn't seem to be anything special about the pages where the
printer hangs (no fancy bitmaps or complicated graphics, for example).
Most of the output that causes the problem has been produced by
dvitps, the DVI-to-PostScript translator that we use, but the printers
have also hung on output from tpscript, a public domain
DITROFF-to-PostScript translator.  Neither of these programs has
produced similar problems on our old Apple LaserWriter I (but I
suppose they could be tickling some subtle incompatibility between
HP's PostScript and Apple's).  

I have some evidence that the difficulty is not a communications
problem between the Sun and the printers:

    First, the spooler logs have frequently shown that as far as the
    host was concerned, the entire job was sent, even though there
    were still several pages to print after the point where the
    printer hung--so it seems that the printer was able to continue
    receiving and buffering characters despite not being able to
    output anything.

    Second, in cases where the host had not sent the entire file, I
    have used Sun's pstat command to look at the STREAMS data
    structures for the printer's serial line and confirmed that the host
    was blocked on that line with a full buffer.

    Third--and this is the weird one--I have enabled debugging output
    in the PostScript that dvitps produces and the resulting messages
    indicate not only that the PostScript program has started work on
    the page after the last page it managed to print, but that
    sometimes that it is _two_ pages ahead of the last page to print.

That last point, combined with the whirring noise that sometimes
accompanies the hanging, makes me wonder if the problem is at least
sometimes associated with the printer's attempt to output a page.
There are, however, no paper jams associated with this problem.

I have modified our spooler to bracket each print job with a
"vmstatus" command, to check for memory leaks.  There don't seem to be
any.  Are there other status commands I could insert to indicate if
some evil condition is gradually building up between jobs?

Now for the treatments I have either tried or considered.  When the
problems first surfaced, shortly after we installed the printers, I
moved the PostScript cartridges from the right to the left slot, and
imagined the problem went away for a while (more likely users just
stopped bothering to tell me about it).  I have considered replacing
dvitps and tpscript with alternative programs (probably Tomas Rokiki's
dvips and GNU's groff), but am reluctant to dragoon our users into
this until I have some clearer indication that it will do some good.
I have also thought of moving the printer's serial lines from the
Sun's MTI boards to the ports on the CPU board, but again am reluctant
to shift equipment around without knowing more about what I am doing.
With the problem so intermittent, even deciding whether it has gone
away is difficult.

So has anybody out there seen similar problems?  Are there known
problems with HP's PostScript?  Has anybody actually managed to read
this far?  Email me, and I will post a summary.

Thanks for your help,
John Jorgensen		jorgnsn@qucis.queensu.ca	(613) 545 6784
Systems Programmer, Dept. of Computing Science, Queen's University