From: CRDGW2::CRDGW2::MRGATE::"SMTP::CRVAX.SRI.COM::RELAY-INFO-VAX" 28-FEB-1991 03:18:43.27 To: ARISIA::EVERHART CC: Subj: SMGSHR Patch for ANU-NEWS dropping into newsgroups while scrolling From: RELAY-INFO-VAX@CRVAX.SRI.COM@SMTP@CRDGW2 To: Everhart@Arisia@MRGATE Received: by crdgw1.ge.com (5.57/GE 1.87) id AA21267; Thu, 28 Feb 91 03:10:16 EST Received: From UCBVAX.BERKELEY.EDU by CRVAX.SRI.COM with TCP; Wed, 27 FEB 91 04:47:18 PST Received: by ucbvax.Berkeley.EDU (5.63/1.42) id AA25949; Wed, 27 Feb 91 04:32:55 -0800 Received: from USENET by ucbvax.Berkeley.EDU with netnews for info-vax@kl.sri.com (info-vax@kl.sri.com) (contact usenet@ucbvax.Berkeley.EDU if you have questions) Date: 27 Feb 91 09:15:44 GMT From: infopiz!mark@decwrl.dec.com Organization: INFO COMM - Computer Consulting, Redwood City, Ca Subject: SMGSHR Patch for ANU-NEWS dropping into newsgroups while scrolling Message-Id: <6667.27cb0840@infocomm.com> References: <1991Feb18.013223.789@vixvax.mgi.com>, <1991Feb19.085915.1@sao.aarnet.edu.au>, <1991Feb19.123812.1@rogue.llnl.gov> Sender: info-vax-request@kl.sri.com To: info-vax@kl.sri.com I'm posting this to comp.os.vms and vmsnet.misc also so I review the problem: In article <1991Feb18.013223.789@vixvax.mgi.com>, glass@vixvax.mgi.com writes: In article <1991Feb9.134458.1@ptavv.llnl.gov>, oberman@ptavv.llnl.gov writes: > > > > This is a VMS 5.4 SMG problem. What happens is that after certain keyboard > > commands, the next key that is defined is read as a . This includes cursor > > keys, numeric keypad keys and defined control keys. > > Confirmed - it is a SMG bug. It was a long an frustrating debugging session, > but I finally found the sucker. The bug is in the READ_COMPOSED_LINE > routine. Basically, there is an uninitialized variable that is initially > set to what ever random stuff is on your stack when you call the routine. > The problem is triggered whenever you do a scrolling operation on the > newsgroups directory, and you go to one of the following newsgroups numbers: > 44, 556, 1545, 1553, 1617, or 2096 (those happen to be the numbers for > SS_ABORT, SS_TIMEOUT, SS_CONTROLO, SS_CONTROLY, SS_CONTROLC, and SS_CANCEL). > > The stack offset of the key variable is slightly different between VMS 5.3 > and 5.4, so I can very well believe that the problem is sensitive to the > release of VMS, the version of your C compiler, and whether or not you > enabled optimization when compiling. > > Is there anyone out there who has some good contacts with DEC and would > like to pass this on (I can provide more details)? Believe it or not, > I don't really work on VMS that much, so I've never done any trouble > reports to DEC. In article <1991Feb19.123812.1@rogue.llnl.gov>, oberman@rogue.llnl.gov writes: > In article <1991Feb19.085915.1@sao.aarnet.edu.au>, gih900@sao.aarnet.edu.au writes: >> >> I'M IMPRESSED! >> >> nice work Bill! > > I, too an impressed! This was more than a little tricky. > > At Bill's request I have submited his findings to DEC and hope to have a > response shortly. NOTE: That's a response, not a fix. But I will keep this > newgroup updated on whatever DEC comes up with. > I also am VERY impressed with Bill's efforts! Now that I've been teased with knowledge of why I've been bothered on and off by this problem for a LONG time, I couldn't possibly wait for Digital to get around to fixing and have that fix propagate through the field test and release cycle. I tried to reproduce Bill's debugging efforts since he didn't actually mention the explicit uninitialized variable. I now know why he was so frustrated in his debugging efforts. I had to pull quite a few tricks out of my hat to get a version of news that reproduced the problem (a newly compiled one didn't). After a while, I simply said, heck, why not convince the routine SMG$READ_COMPOSED_LINE to initialize all of it's local variables to zero, instead of tracking down the specific one. The following patch works for the version of SMGSHR.EXE supplied with VMS V5.4-1. $ SET DEFAULT SYS$COMMON:[SYSLIB] $ PATCH SMGSHR.EXE REPLACE/INSTR 244F 'CLRL B^0FC(FP)' EXIT 'MOVC5 #0, (SP), #0, #2F0, (SP)' EXIT UPDATE EXIT $ RUN SYS$SYSTEM:INSTALL SYS$SHARE:SMGSHR/REPLACE $ An UNTESTED patch that does the same basic thing for the SMGSHR for VMS V5.3-1: $ SET DEFAULT SYS$COMMON:[SYSLIB] $ PATCH SMGSHR.EXE INSERT/INSTR 21EE 'MOVAB W^0FD0C(SP),SP' 'MOVC5 #0, (SP), #0, #2F4, (SP)' EXIT UPDATE EXIT $ RUN SYS$SYSTEM:INSTALL SYS$SHARE:SMGSHR/REPLACE $ This patch will remove the problem until DEC gets around to fix things. I'll recreate and post the similar patches to above until DEC gets to it. -- Mark Pizzolato - INFO COMM Computer Consulting, Redwood City, Ca PHONE: (415)369-9366 UUCP: decwrl!infopiz!mark or uunet!lupine!infopiz!mark DOMAIN: mark@infocomm.com