🚀 go-pugleaf

This just popped up on the INFO-VAX list, and I felt that since Ned failed to
indicate he was posting this to WP50-L@UBVM, that I'd better forward it on.
Hopefully Barb @WPCORP is listening in, or somebody else in her stead, so that
it can be addressed before VMS v5.5 gets into widespread usage (it just went to
general release this month, I think).  Ned's address, by the way since it's
truncated in the following extract from VAX Notes, is NED@HMCVAX.CLAREMONT.EDU,
where he's the postmaster.  He's also one of the authors of PMDF, in case his
name sounds familiar to some WP50-L readers.

>                   <<< USER:[NOTES$LIBRARY]INFO-VAX.NOTE;1 >>>
>                                 -< info-vax >-
>===============================================================================
   =
>Note 11.169         Info-VAX Postings for 01/12/92 - 01/18/92         169 of 18
   3
>PSULCS::INFO_VAX "Ned Freed, Postmaster <NED@HMCVAX" 58 lines  16-JAN-1992 02:1
   1
>     -< Word Perfect mixed with VMS V5.5 can kill symbionts in other pro >-
>-------------------------------------------------------------------------------
   -
>I have a special PostScript symbiont that I like to use for various things.
>There's nothing particularly special or interesting about it -- it is hacked up
>version of one of the symbionts from the DECUS library, but we use it here and
>like it and so do a few other sites.
>
>We upgraded to VMS V5.5 some time back and the symbiont worked fine with the
>new version of VMS and all its new queue stuff. But when we upgraded to VMS 5.5
>at the Harvey Mudd College we immediately got into trouble on one particular
>system. All the queues on the one system failed with an access violation. But
>the symbiont worked fine everywhere else in the cluster. What the blazes was
>going on?
>
>After a little debugging the problem became fairly obvious. All user-written
>symbionts call the routine SMB$INITIALIZE to initialize their connection with
>the job controller. This routine was stepping on some registers that it should
>have preserved. In particular, R4 was getting trounced.
>
>But when I looked at the code for SMB$INITIALIZE it was clear that R4 was
>specified to be saved by the register save mask, just as it should be!
>
>To make a long story short, further analysis showed that something had
>ingratiated itself into the code path from the symbiont to the symbiont routine
>library. Specifically, it is Word Perfect that does this. It defines a exec
>mode logical name in the group logical name table for [1,*] that translates
>SMBSRVSHR to WPCORP_PTR51_SRV. This image apparently is some kind of switcher
>that redirects calls to the symbiont routines to other things in some cases.
>For the most part it just falls through to the regular symbiont routine
>library, it seems.
>
>However, when a product indulges in this sort of redirection there's a
>potential problem. Your code has to specify a register mask that saves at least
>as many registers as the routine you've replaced (assuming you plan on calling
>this routine eventually). Under VMS V5.4 SMB$INITIALIZE only needed to save R2
>and R3. Under VMS 5.5 the routine is a lot more complex and it now saves R2
>through R8. But the mask in Word Perfect's replacement routine did not know
>this, and as a result the registers were not saved and got corrupted. And
>of course this only happened on our one system running Word Perfect.
>
>The solution for me was fairly obvious -- get rid of this redirection nonsense.
>Once I did this things started to work again. It may have broken Word Perfect
>in the process. Frankly, I didn't give a damn about that, but it should also be
>possible to correct Word Perfect's register save masks with a little judicious
>patching of WPCORP_PTR51_SRV. Saving all possible registers would be an
>excellent idea...
>
>This is a potentially fatal problem for any symbiont on any system running
>Word Perfect and VMS 5.5. Anyone running this combination should be aware of
>this problem and should contact Word Perfect Corp. and ask them to fix this.
>
>I'm not going to verbalize my opinion of the modification technique Word
>Perfect has chosen to use because doing so would involve extensive use of some
>of the breezier copulative verbs.
>
>I'm cross-posting this to info-pmdf and info-multinet since I know that both
>PMDF and Multinet make use of special-purpose symbionts that could be affected
>by this problem.
>
>                                        Ned
>

+---------------------------------------------------------------------------+
| J.Lance Wilkinson ("Lance")            BitNet:    JLW@PSULIAS.BITNET      |
| Systems Design Specialist - Lead       InterNet:  <not yet>               |
| Library Computing Services             AT&T:      (814) 865-1818          |
| E8 Pattee Library                      "I'd rather be dancing..."         |
| Penn State University         A host is a host from coast to coast,       |
| University Park, PA 16802     And no one will talk to a host that's close |
|                               Unless the host that isn't close            |
| <POSTMAST@PSULIAS.BITNET>     Is busy, hung or dead.                      |
+-"Never Underestimate the bandwidth of a station wagon full of mag tapes"--+