Google
 

Trailing-Edge - PDP-10 Archives - bb-bt99e-bb - t1s702.d10
There is 1 other file named t1s702.d10 in the archive. Click here to see a list.
                 EDIT DESCRIPTIONS FOR TOPS-10-KS-V702                          
  
  
                             EDIT 11113  FOR 702
  
[SYMPTOM]
  
     Stopcodes DOM, CMU.
  
  
[DIAGNOSIS]
  
     IPCSER occasionally can return the MM resource  when  it  doesn't
own  it.  This can only happen on a page mode send when the page to be
sent has already been paged out of the sender's working set.
  
  
[CURE]
  
     Rearrange the code so that MM is obtained  and  released  at  the
proper places.
********************************************************************************
  
  
                             EDIT 11157  FOR 702
  
[SYMPTOM]
  
     "?Illegal data mode for device ..." error message can contain  no
device name if the device is a TTY.
  
  
[DIAGNOSIS]
  
     Routine CHKMOD calls TTYKLQ before dispatching to ERRCON to  type
the error message.
  
  
[CURE]
  
     Call TTYKLQ after the error message is printed.
  
  
********************************************************************************
  
  
                             EDIT 11378  FOR 702
  
[SYMPTOM]
  
     Customer QUEUE. UUO functions are rejected by IPCSER.
  
  
[DIAGNOSIS]
  
     Incorrect range checking.
  
  
[CURE]
  
     CAIL T1,.GTQFT##-.GTQFC## -> CAML T1,[.GTQFT##-.GTQFC##]
  
  
********************************************************************************
  
  
                             EDIT 11390  FOR 702
  
[SYMPTOM]
  
     1.  Cannot clear the aborted status of a lock by using  the  ENQ.
         modify function.
  
     2.  Cannot change the status of a lock from shared  to  exclusive
         by  using the ENQ. modify function, even when the user is the
         only sharer of a lock.
  
     3.  [1,2]  jobs  requesting  shared  global  locks  instead   get
         exclusive global locks.
  
  
[DIAGNOSIS]
  
     1.  The documentation clearly states that the aborted status of a
         lock  can  be cleared by issuing an ENQ. modify function with
         EQ.FAB clear, but the code  also  clearly  never  implemented
         this case.
  
     2.  For  the  second  case,  the  documentation  states  this  is
         possible,  and  there  is  much code to support it.  However,
         this code does not work.
  
     3.  The third case is reasonably obscure.  It requires  having  a
         [1,2]  job  that  doesn't  have  the  JP.ENQ  privilege.  The
         subroutine that causes the problem is  documented  as  saving
         all the ACs, but it calls PRVBIT which normally will preserve
         all ACs, except in this very case.
  
  
[CURE]
  
     1.  Add code to handle clearing the abort  status  with  an  ENQ.
         modify function.
  
     2.  Add missing instruction to skip over  the  lock  block  in  a
         queue block chain.
  
     3.  Call SAVT to  preserve  all  the  temp  ACs,  before  calling
         PRVBIT.
  
********************************************************************************
  
  
                             EDIT 11515  FOR 702
  
[SYMPTOM]
  
     Stopcode KAF.
  
  
[DIAGNOSIS]
  
     Code in VALSEG has a T1 instead  of  a  T2  in  chasing  fragment
pointers.
  
  
[CURE]
  
     T2.
  
  
********************************************************************************
  
  
                             EDIT 11589  FOR 702
  
[SYMPTOM]
  
     KS Networks Don't.
  
     In Particular:
  
     1.  Mongen defined line uses get ignored;  people run crocks like
         ANF and DECnet.EXE.
  
     2.  DDCMP doesn't meet  corporate  spec.   Confuses  others,  and
         ourself.
  
     3.  %Unexpected restart on  ...  messages  unexpectedly  restart,
         producing a black, illegible mess.
  
     4.  Adjacent node congestion.
  
     5.  Adjacent node remote reply timeouts.
  
     6.  Sundries.
  
  
  
[DIAGNOSIS]
  
     We're VERY confused.
  
     1.  BLTing zero over line user in KDL
  
     2.  Was coded to match ANF, almost.  DECnet expects otherwise.
  
     3.  When we get a START on a running line, we don't  halt  DDCMP,
         we just try and restart it.
  
     4.  Data messages aren't ACK'd until the other end sends a  REP5.
         We idle REPs on DECnet lines.
  
     5.  Of course.
  
  
  
[CURE]
  
     Of course.
  
     1.  Don't
  
     2.  Change to conform to spec
  
     3.  Halt DDCMP before restarting.  Eg, send a STRT, not a STACK.
  
     4.  Always ACK data messages immediately.  Note that this may not
         be a seperate message;  it can be piggybacked if the phase of
         the moon is OK.
  
     5.  Don't.  ANF will continue to idle REPs, as  this  is  how  it
         determines  line  down.  This change causes the line timer to
         have a "stopped" state, per the spec.
  
     6.  yes.
  
  
  
********************************************************************************
  
  
                             EDIT 11594  FOR 702
  
[SYMPTOM]
  
     If DECNET.EXE or other user of KDP.  fcn 104 creates a new DECnet
line,  ROUTER's circuit database is trashed, and NML can't talk to the
line.
  
  
[DIAGNOSIS]
  
     MOVE should be MOVEI, off by 1, reversed test, code too big,  and
SUB should be SUBI.
  
  
[CURE]
  
     I hope so.
  
  
********************************************************************************
  
  
                             EDIT 11600  FOR 702
  
[SYMPTOM]
  
     TMR and other STOPCDs when OPR defines a large disk and  believes
the MAX value for   SATs/unit.  Observed on RP07.
  
  
[DIAGNOSIS]
  
     ASKSPU tries to figure out how many  retrieval  pointers  (=  SAT
clusters)  will  fit in the RIB it builds.  It cleverly allows for all
the change of unit pointers, and the prime RIB.  It fails to allow for
the  redundant RIB and the possible EOF pointer.  The TMR happens when
we try to add the spare RIB pointer.  Because TMR is  a  HALT  STOPCD,
20F tries to help things go further downhill.
  
  
[CURE]
  
     Account for worst case when computing what to let the OPR say.
  
     ASKSPU+8.  SUBI    T2,RIBLEN  -1 -> RIBLEN  -3
  
  
********************************************************************************
  
  
                             EDIT 11622  FOR 702
  
[SYMPTOM]
  
     If the new FILOP.  function RENAME with allocation given in words
can  truncate  a  file  incorrectly.   If the word count is on a block
boundary, the file become one block too small.
  
  
[DIAGNOSIS]
  
     If the file is being truncated to  a  block  boundary,  the  last
block in the file is full (128 words) not empty (as currently coded).
  
  
[CURE]
  
     In FILUUO at RENA35-4 change:
  
     TRNE       P4,BLKSIZ-1
  
     TRZ        P4,BLKSIZ TO
  
     TRNN       P4,BLKSIZ-1     ;BLOCK BOUNDARY?
  
     TROA       P4,BLKSIZ       ;YES -- LAST BLOCK IS FULL
  
     TRZ        P4,BLKSIZ       ;NO -- COMPUTE REMAINDER
  
  
********************************************************************************
  
  
                             EDIT 11623  FOR 702
  
[SYMPTOM]
  
     The DIU-10 slave jobs will be created not LOG'd  in.   They  will
then  LOGIN  under the PPN of the requestor.  Since DIU will be a very
large program it would be nice if it could  run  virtual.   Currently,
the  DIU slave job has no VM quota, so it can not run virtual.  If DIU
could set it's MAXIMUM VIRTUAL LIMIT the job could run virtual.
  
  
[DIAGNOSIS]
  
     Allow a non-logged-in job to set  it's  maximum  virtual  limits.
DIU will then set these limits before the slave job is logged in.
  
  
[CURE]
  
     In COMCON at SETTBL + 24 change
  
     XWD        NEDPRV,SETUVL   to
  
     XWD        NEDPRV+NLOGOK,SETUVL
  
  
********************************************************************************
  
  
                             EDIT 11704  FOR 702
  
[SYMPTOM]
  
     The TOPS-10 Software Installation Guide says that  "Erase-to-EOL"
and     "backspace-space-backspace"     characteristics     of     the
customer-defined terminal types can be global  addresses  or  symbols.
There  is  no  easy  way  for  the customer to use this option without
editting the monitor.
  
  
[DIAGNOSIS]
  
     COMDEV does not handle external symbols passed to it through  the
TERMCR macro.
  
  
[CURE]
  
     Check the symbols passed by the TERMCR macro during pass  2.   If
they are NEEDED, declare them to be external.
  
  
********************************************************************************
  
  
  
END OF  TOPS-10-KS-V702