Trailing-Edge
-
PDP-10 Archives
-
BB-D868C-BM
-
4-documentation/tops20.tco
There are 24 other files named tops20.tco in the archive. Click here to see a list.
TOPS20 Change Order Number 4.1880
Written by: MILLER 30-Jan-78 13:12:23
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: APRSRV FORK
Related TCO's:
Related SPR's:
Problem: NEED WAY FOR USERS TO SPECIFY
ARITHMETIC TRAPS AND LUUO DISPATCH BLOCKS
Diagnosis: AS ABOVE
Solution: ADD SWTRP JSYS AND SUPPORTING ROUTINES
TOPS20 Change Order Number 4.1881
Written by: ENGEL 31-Jan-78 11:19:08
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: Vertical Tab (CTRL/K) does not now belong to the
proper wake-up class. It belongs to the punctuation class and should belong
to the TOPS-10 class.
Diagnosis: Change the class designation.
Solution: Make Vertical Tab part of TOPS-10 wake-class (a superset of
punctuation)
TOPS20 Change Order Number 4.1884
Written by: HURLEY 14-Feb-78 14:36:46
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IPCF
Related TCO's:
Related SPR's:
Problem: ADD MUTIL FUNCTIONS TO ALLOW A PROGRAM
TO SPECIFY A PID TO WHICH A MESSAGE IS TO BE SENT IF THIS JOB'S
PIDS GET DELETED.
THIS FUNCTION IS INTENDED FOR USE BY TPS-20 TO DETECT IF ANY OF
ITS AUXILLIARY PROCESSES GET RESTARTED.
THE FORMAT OF THE MESSAGE IS:
WORD 0/ .IPCKP
WORD 1/ PID THAT WAS DELETED
Diagnosis:
Solution:
TOPS20 Change Order Number 4.1885
Written by: PORCHER 17-Feb-78 13:12:04
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK
Related TCO's:
Related SPR's:
Problem: TOPS-10 compatable programs with "CCL start"
addresses cannot be started at the CCL start address if they are execute-only.
Diagnosis: Lack of feature
Solution: Add new functionality to SFRKV monitor call.
If the program being started with SFRKV has a TOPS-10 style entry vector
(JRST in the left half), then the left half of AC2 in the SFRKV call
is the start address offset. The only legal offsets are 0 or 1, and
only legal for entry vector position 0 (the start address).
Thus, for TOPS-10 style vectors, the left half of AC2 will
be added to the contents of .JBSA before using this value to
start the program.
TOPS20 Change Order Number 4.1888
Written by: MURPHY 2-Mar-78 16:39:00
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: STG DSKALC
Related TCO's:
Related SPR's:
Problem:
DIFFICULT TO BRING SYSTEM UP WITH SOME STRUCTURE NOT CALLED PS
AS PUBLIC STRUCTURE, ESPECIALLY IF A STRUCTURE CALLED PS IS ALSO
ON LINE.
Diagnosis: NO WAY TO CHANGE DEFAULT PUBLIC STRUCTURE NAME
Solution: ADD CELL TO HOLD DEFAULT PUBLIC STRUCTURE NAME. IF NON-0
USE IT, OTHERWISE, USE PS;.
TOPS20 Change Order Number 4.1889
Written by: MILLER 10-Mar-78 09:03:29
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK SCHED
Related TCO's:
Related SPR's:
Problem: Need to propogate SJPRI setting
to all forks in the job
Diagnosis: Only forks created subsequent to the SJPRI get
the desired capabilities.
Solution: Interrupt top fork of job and propogate capability
to all forks
TOPS20 Change Order Number 4.1900
Written by: HALL 12-Apr-78 12:03:02
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: ATSSRV FILATS IO STG PROLOG GLOBS
PARAMS
Related TCO's:
Related SPR's:
Problem: THERE IS NO SUPPORT FOR APPLICATIONS
TERMINALS IN TOPS-20, AND TPS-20 NEEDS IT.
Diagnosis: HAVE TO ADD MONITOR SUPPORT OF ATS AND NRM
Solution: ADD CODE AS DESCRIBED IN FUNCTIONAL SPEC (NEW MODULES
ATSSRV AND NRMSRV)
TOPS20 Change Order Number 4.1901
Written by: MURPHY 12-Apr-78 13:14:40
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem:
Need to give system administrators control over degree to which system
favors interactive or computational jobs.
Diagnosis:
Solution:
Set of flags in scheduler which affect certain scheduling decisions.
See memo of 11-April-78, D. Murphy
TOPS20 Change Order Number 4.1904
Written by: MURPHY 18-Apr-78 13:31:01
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM SCHED STG
Related TCO's:
Related SPR's:
Problem:
IMPLEMENT WORKING SET SWAPPING.
Diagnosis:
Solution:
TOPS20 Change Order Number 4.1905
Written by: MURPHY 18-Apr-78 13:54:00
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Related TCO's:
Related SPR's:
Problem:
THIS IS IDENTICAL TO TCO #1891 IN THE RELEASE 3 FILE AND
WAS FIRST INCLUDED IN RELEASE 3A.
Diagnosis:
Solution:
TOPS20 Change Order Number 4.1906
Written by: OSMAN 25-Apr-78 10:44:04
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem:
IF GJ%CFM (REQUIRE CONFIRMATION) IS ON FOR THE .CMFIL CALL TO COMND,
BAD EFFECTS HAPPEN (SUCH AS TYPIST HAVING TO TYPE TWO CARRIAGE RETURNS).
Diagnosis:
GTJFN AND COMND BOTH TRY TO DO CONFIRMATION IF USER PROGRAM TURNS ON
GJ%CFM.
Solution:
USER PROGRAMS SHOULD NOT USE GJ%CFM IN CONJUNCTION WITH COMND. THEY
SHOULD USE .CMCFM INSTEAD.
COMND IS BEING CHANGED TO ADD GJ%CFM AS YET ANOTHER PART OF THE GTJFN
BLOCK THAT COMND IS IN CONTROL OF. THE LIST OF GTJFN BITS GIVEN
UNDER THE .CMFIL DOCUMENTATION SHOULD BE AUGMENTED WITH GJ%CFM AS YET
ANOTHER BIT THAT COMND CONTROLS.
TOPS20 Change Order Number 4.1907
Written by: DBELL 8-May-78 13:37:59
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem:
SIN CAN SOMETIMES RETURN DATA WHEN THERE IS ACTUALLY NO DATA THERE.
Diagnosis:
WHEN SFPTR IS DONE TO SEVERAL PAGES IN A NONEXISTANT SECTION, THE
INDEX BLOCK CURRENTLY MAPPED IS STILL LOOKED AT, AND THUS CAN RETURN
DATA WHERE THERE SHOULD BE NONE.
Solution:
DON'T UPDATE FILLFW IN THE CASE WHERE THE INDEX BLOCK DOESN'T EXIST.
IN DISC, AT NEWWNA+6 IN THE LITERAL, JRST TO NEWWNZ INSTEAD OF NEWNA1.
TOPS20 Change Order Number 4.1911
Written by: MURPHY 18-May-78 13:36:15
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's:
Problem:
INPUT TO JOB BEING RUN OVER PTY APPEARS TO BE LOST
Diagnosis:
PROGRAM SETS NO WAKEUP ON CR; LINE APPEARS HUNGRY AT LEFT MARGIN
BUT BUFFER BECOMES FULL IF FULL LINE OF TEXT IS INPUT.
Solution:
IF PTY, FORCE WAKEUP ON CR REGARDLESS OF WAKEUP MASK.
TOPS20 Change Order Number 4.1912
Written by: PORCHER 19-May-78 15:53:24
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK
Related TCO's:
Related SPR's:
Problem: No way to read an extended PC word and
PC flags of a process. Also no way to tell if a process is execute-only.
Diagnosis: Lack of feature.
Solution: Extend RFSTS monitor call.
See LNGRFS.MEM for details.
TOPS20 Change Order Number 4.1916
Written by: DBELL 26-May-78 19:15:29
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem:
IF OFNS OR PAGES RUNS OUT IN NEWLFS WHEN TRYING TO MAKE A SHORT FILE
INTO A LONG FILE, THE SYSTEM CRASHES WITH KIBADU BUGHLT.
Diagnosis:
AT NEWFLL+2, THERE IS A SUB P,BHC+1 WHICH IS INCORRECT AND SO THE
STACK IS TRASHED.
Solution:
REMOVE THE INSTRUCTION.
TOPS20 Change Order Number 4.1917
Written by: R.ACE 2-Jun-78 14:53:39
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: CTRL/W DOESN'T WORK AROUND CR/LF.
Diagnosis: NEVER CODED TO WORK AROUND END-OF-LINE.
Solution: CHANGE CODE THAT CHECKS FOR <CR-LF> -- TREAT IT AS A
PUNCTUATION CHARACTER
TOPS20 Change Order Number 4.1918
Written by: R.ACE 2-Jun-78 15:14:56
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: TEXTI JSYS RETURNS WRONG ERROR CODE IF
ARGUMENT BLOCK IS TOO SHORT
Diagnosis: TEXTI RETURNS RDTX1 (INVALID STRING POINTER) INSTEAD OF
ARGX17 (INVALID ARGUMENT BLOCK LENGTH)
Solution: CHANGE TEXTI TO RETURN ARGX17 IS ARGUMENT BLOCK TOO SHORT
TOPS20 Change Order Number 4.1921
Written by: MCLEAN 15-Jun-78 15:28:34
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem: GDSTS ONLY HAS 2 BITS FOR MAGTAPE DENSITY
Diagnosis: NO SPACE
Solution: MAKE GDSTS ALWAYS RETURN 0 FOR DENSITY (DEFAULT)
DOCUMENTATION SHOULD NO LONGER DEFINE THESE BITS.
TOPS20 Change Order Number 4.1924
Written by: MILLER 22-Jun-78 09:02:12
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV STG
Related TCO's:
Related SPR's:
Problem: Need way to automatically
enable XON and XOFF for DPY type terminals
Diagnosis: As above
Solution: Add new filed to TTYPE1 table. Bit 0 now indicates
whether XON/XOFF is automatically enabled
TOPS20 Change Order Number 4.1925
Written by: FORTMILLER 22-Jun-78 14:41:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem: MTOPR function .MOINF always returns
a 1 in word .MOITP which is the MTA type code.
Diagnosis: The code in MAGTAP is MOVEI T1,1
Solution: Get the MTA type code from the UDB.
TOPS20 Change Order Number 4.1926
Written by: OSMAN 23-Jun-78 15:42:55
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem:
IF BACKGROUND PROGRAMS RUNNING, "^EQUIT" FOLLOWED BY "START" LEFT
BACKGROUND PROGRAMS FROZEN.
Diagnosis: MONITOR WAS FREEZING FORKS ON ^EQUIT AND ^P.
IT'S NECESSARY ON ^P, SINCE ^P IS FOR EMERGENCY, AND YOU WANT EVERYTHING
TO STOP. HOWEVER, ^EQUIT HAS NO NEED TO FREEZE THE FORKS.
Solution: MOVE THE FFORK SUCH THAT IT ONLY HAPPENS ON ^P AND NOT
ON ^EQUIT
TOPS20 Change Order Number 4.1927
Written by: OSMAN 28-Jun-78 10:54:43
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: STG MEXEC PROLOG
Related TCO's:
Related SPR's:
Problem: BATCH SYSTEM NEEDS TO KNOW IF JOB
IS AT TOPS20 COMMAND LEVEL OR NOT
Diagnosis:
Solution: PUT IN NEW GETJI/SETJB OFFSETS .JIT20/.SJT20 WHICH
RETURN/SUPPLY -1/0 FOR JOB IS AT TOPS20 COMMAND LEVEL / JOB AIN'T.
EXEC WILL SET IT TO -1 WHEN GOING FROM PROGRAM LEVEL TO EXEC COMMAND
LEVEL, AND TO 0 WHEN GOING TO PROGRAM LEVEL.
TOPS20 Change Order Number 4.1928
Written by: KONEN 28-Jun-78 14:01:38
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MSTR
Related TCO's:
Related SPR's:
Problem: SECG37 BUGHLT
Diagnosis: NO ARGUMENT CHECK FOR OBTAINING STATUS OF DISK UNIT IN MSTR JSYS
Solution: CHECK ARGUMENTS
TOPS20 Change Order Number 4.1929
Written by: MILLER 29-Jun-78 12:50:48
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem: Need special high priority queue
and special low priority queue.
Diagnosis: As above. Need way to create "background" jobs
(low priority) and special high-priority ones as well.
Solution: Add two more queues to scheduler.
TOPS20 Change Order Number 4.1930
Written by: MURPHY 30-Jun-78 15:39:15
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED PROLOG
Related TCO's:
Related SPR's:
Problem:
Page faults while NOSKED causing performance degradation.
Diagnosis:
Excessive use of NOSKED.
Solution:
Provide CRSKED (critical section) scheduling control. Allows
process priority while holding critical system resource (e.g.
global lock), but allows other process to be run if page
fault or other block occurs.
TOPS20 Change Order Number 4.1931
Written by: HEMPHILL 7-Jul-78 16:02:30
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's: 11229
Problem:
The monitor cannot get a JFN for DSK*:<directory>file.typ
unless the directory exists on PS:.
Diagnosis:
If directory lookup fails on the first structure (PS:) no
check is made to determine whether DSK*: had device field.
Solution:
Check the structure step flag, and step the device if set.
TOPS20 Change Order Number 4.1932
Written by: HEMPHILL 7-Jul-78 16:14:58
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's: 4.1931
Related SPR's:
Problem:
The monitor cannot get a JFN for a filespec which includes a
logical name definition of DSK*:, unless a directory of the given
name is on PS:. (A directory must exist on PS: even to find files
on other structures.)
Diagnosis:
If the directory lookup fails on the first structure (PS:),
the logical name code fails to check to see if DSK*: was
specified.
Solution:
Add code to check for DSK*:.
TOPS20 Change Order Number 4.1933
Written by: HEMPHILL 7-Jul-78 16:24:12
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: LOOKUP
Related TCO's:
Related SPR's: 11229
Problem:
The monitor reports files obtained with DSK*: as being on the
PS: structure, regardless of the actual structure that they reside
on.
Diagnosis:
The routine DEVSTP in the module LOOKUP steps the structure
number but neglects to store the new structure name in the JFN data
block.
Solution:
Add code to store structure name string.
TOPS20 Change Order Number 4.1934
Written by: MILLER 8-Jul-78 14:21:59
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MONSYM PAGEM JSYSA
Related TCO's:
Related SPR's:
Problem: Can't unmpa a page and have
its changed contents thrown away. This is necessary
for data base applications.
Diagnosis: As above. Unmap always puts data back on disk
Solution: Define new PMAP bit, PM%ABT.
TOPS20 Change Order Number 4.1935
Written by: HEMPHILL 8-Jul-78 19:04:17
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MSTR
Related TCO's:
Related SPR's: 11507
Problem:
.MSRNU function of MSTR JSYS is supposed to return the number of pages
for swapping on the structure in word .MSRSW of the argument block.
Instead, it returns the number of swapping sectors on this unit.
Diagnosis:
The code does one thing, the documentation says another. Since the
.MSCRE function requires swapping pages per structure, it makes
sense to make the trivial code change required rather than change
the documentation.
Solution:
Divide sectors by 4, multiply by number of units in the structure.
TOPS20 Change Order Number 4.1936
Written by: MCLEAN 9-Jul-78 00:26:46
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's:
Problem: PROBLEM ALREADY FIXED
Diagnosis:
Solution:
TOPS20 Change Order Number 4.1937
Written by: MILLER 10-Jul-78 11:14:08
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF PAGEM
Related TCO's:
Related SPR's:
Problem: DBMS needs a way to request
a UFPGS without blocking. This would allow writes to
start and to be overlapped with computing.
Diagnosis: Can be done, but not if the OFN needs updating
also.
Solution: Provide a flags field in UFPGS JSYS to be the left-half
of AC2. Define UF%NOW, no wait. This will cause code not to
block if OFN does not need updating. NOTE: The JSYS will
still block if the OFN needs updating.
TOPS20 Change Order Number 4.1938
Written by: MILLER 10-Jul-78 13:27:16
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem: Need a "class" scheduler
Diagnosis: Certainly
Solution: Add class scheduler to release 4
TOPS20 Change Order Number 4.1940
Written by: MCLEAN 12-Jul-78 03:11:31
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem: NEED TO KNOW MORE ABOUT STATE OF
DEVICES (TAPE IN SPECIFIC)
Diagnosis: NEED TO ADD A NEW MTOPR JSYS
Solution: ADD .MOSTS TO RETURN MTA STATUS
TOPS20 Change Order Number 4.1941
Written by: MCLEAN 12-Jul-78 03:39:31
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's: 4.1991
Related SPR's:
Problem: NEED NEW SMON FUNCTIONS FOR TAPE LABELS
Diagnosis:
Solution: ADD .SFLBT (DEFAULT TAPE LABEL TYPE) AND .SFMTA (ABILITY
TO ASSIGN AND/OR OPENF MTA.
TOPS20 Change Order Number 4.1946
Written by: MCLEAN 14-Jul-78 14:14:42
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: NEED NEW SMON FUNCTION FOR UPDATING OF
LAST ACCESS READ TIME IN DIRECTORY TO PREVENT EXTRANEOUS WRITES OF
DIRECTORY PAGES
Diagnosis:
Solution: ADD .SFRDU (READ DATE UPDATE) FUNCTION TO SMON JSYS
TOPS20 Change Order Number 4.1949
Written by: MCLEAN 21-Jul-78 01:05:56
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's: 4.1946
Related SPR's:
Problem: NEED NEW BIT IN OPENF TO
FORCE AN UPDATE OF THE READ TIME
Diagnosis:
Solution: ADD OF%FDT FORCE TIME UPDATE IN OPENF
TOPS20 Change Order Number 4.1950
Written by: 21-Jul-78 03:35:16
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Related TCO's:
Related SPR's:
Problem: MTAINF CAN
Diagnosis:
Solution:
TOPS20 Change Order Number 4.1953
Written by: PORCHER 25-Jul-78 10:31:05
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem: Some of the I/O monitor calls don't recognize
indexed/indirect ERJMP/ERCALs.
Diagnosis: Routine CHKERT checks for ERJMP/ERCAL on the entire left half
of the instruction.
Solution: Make CHKERT only use the OPCODE and AC fields for comparison.
TOPS20 Change Order Number 4.1956
Written by: PORCHER 27-Jul-78 14:31:36
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem: A RENAME command that is aborted by ^C
hangs.
Diagnosis: The RNAMF monitor call goes OKINT with a JFN locked.
Solution: Remove extraneous OKINT at the end of DSKREN routine.
Note that this causes the RNAMF monitor call to complete NOINT--
this is because RELJFN unlocks the JFN without an OKINT.
TOPS20 Change Order Number 4.1966
Written by: R.ACE 3-Aug-78 10:45:58
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: IF COMND IS PARSING A NODE NAME,
TERMINATING THE NODE NAME WITH AN ALTMODE CAN CAUSE SOME
UNEXPECTED RESULTS.
Diagnosis: COMND DOES NOT DELETE THE ALTMODE AFTER THE NODE NAME
FROM THE USER'S TYPEIN BUFFER.
Solution: CHANGE XCMNOD TO EXIT TO XCOMXR INSTEAD OF XCOMXI
TOPS20 Change Order Number 4.1973
Written by: OSMAN 7-Aug-78 09:31:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: SOMETHIMES, ON SCREENS WHEN YOU TYPE
^R, THE CURSOR RIDES UP THE SCREEN
Diagnosis: IN ^R LOGIC IN TEXTI, CODE WAS SEEING IF LAST CHARACTER
WAS LF WITHOUT FIRST CHECKING IF THERE EVEN IS A LAST CHARACTER.
HENCE CLEVER GARBAGE COULD TRICK IT
Solution: CHECK FOR BEGINNING OF BUFFER BEFORE FETCHING THE CHARACTER
TOPS20 Change Order Number 4.1974
Written by: HEMPHILL 7-Aug-78 12:31:19
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 11448
Problem:
COMND DOES NOT PROPERLY HANDLE ESCAPE AFTER COLON WHEN PARSING
DEVICE NAMES, I.E. IT TREATS THE COLON AS TERMINATING THE FIELD.
Diagnosis:
IT SHOULD READ ONE MORE CHARACTER IF THE USER TYPED A COLON.
IF THE CHARACTER IS NOT AN ESCAPE, COMND SHOULD PUT IT BACK.
Solution:
ADD THE NECESSARY CODE.
TOPS20 Change Order Number 4.1975
Written by: HEMPHILL 7-Aug-78 12:35:33
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's: 4.1974
Related SPR's:
Problem:
COMND DOES NOT PROPERLY HANDLE AN ESCAPE FOLLOWING THE PARSING OF
A TOKEN.
Diagnosis:
IT SHOULD READ ONE MORE CHARACTER AFTER THE TOKEN IS PARSED. IF
THE CHAR IS NOT AN ESCAPE, COMND SHOULD PUT IT BACK.
Solution:
ADD THE NECESSARY CODE.
TOPS20 Change Order Number 4.1983
Written by: HELLIWELL 12-Aug-78 20:05:26
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: CAN'T SAY "BUILD FOO:" OR "SET DIR MUMBLE FOO:"
WHERE FOO: IS A LEGAL LOGICAL NAME.
Diagnosis: CRDIR JSYS WILL NOT ACCEPT STRING WITH NO
EXPLICIT DIRECTORY NAME.
Solution: FIX CRDIR TO DO A DIRST AND PARSE THE RESULTS WHEN NO
DIRECTORY NAME WAS SPECIFIED.
TOPS20 Change Order Number 4.1984
Written by: HEMPHILL 14-Aug-78 12:47:41
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MSTR
Related TCO's:
Related SPR's:
Problem:
SECNX BUGHLT WHEN TRYING TO CREATE OR RUN ON 3 OR MORE PACK STRUCTURES.
Diagnosis:
MODEL B PROCESSOR BUG. LOGICAL UNIT NUMBER,,STRUCTURE NUMBER IS
STORED IN MSTRSS. THIS VALUE IS USED TO INDEX INTO STRUCTURE TABLES.
IF THE LOGICAL UNIT NUMBER IS ZERO OR ONE, EVERYTHING IS COOL.
IF IT'S TWO OR GREATER, THOUGH, IT TRIES TO REFERENCE SECTION 2 OR
ABOVE.
Solution:
CLEAR THE LEFT HALF OF THE AC BEFORE DOING ANY INDEXING
TOPS20 Change Order Number 4.1985
Written by: HALL 15-Aug-78 10:19:27
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FREE
Related TCO's:
Related SPR's:
Problem: WHEN RELFRE DETECTS AN ERROR, IT ISSUES A BUGCHK BUT DOES
NOT PROVIDE ANY INFORMATION ABOUT THE BLOCK BEING RETURNED. IT IS
HARD TO TELL AFTER THE FACT WHAT HAPPENED.
Diagnosis: BUGCHK DOES NOT HAVE OPTIONAL DATA
Solution: ADD OPTIONAL DATA AS FOLLOWS:
BUGCHK RELFRA: B (ADDRESS OF BLOCK BEING RETURNED), C (LOWER LIMIT
FREE SPACE), D (UPPER LIMIT ON FREE SPACE)
THIS BUGCHK OCCURS ONLY WHEN THE BLOCK BEING RETURNED DOES
NOT START WITHIN THE RANGE OF THE FREE SPACE TO WHICH IT IS
BEING RETURNED
BUGCHK RELBAD: CX (PC AT TIME OF JRST TO THE BUGCHK),A,B,C,D (CONTENTS
VARY ACCORDING TO THE PC)
TOPS20 Change Order Number 4.1986
Written by: R.ACE 17-Aug-78 10:30:21
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem: NEW MAGTAPE MTOPR FUNCTION .MOSTA SOMETIMES RETURNS THE
WRONG DEVICE TYPE IN THE .MODVT WORD
Diagnosis: ACCESS TO DEVICE TYPE FIELD IN UDBSTS NOT DONE SYMBOLICALLY
Solution: MAKE ACCESS TO UDBSTS DEVICE TYPE FIELD BE SYMBOLIC
TOPS20 Change Order Number 4.1989
Written by: LCAMPBELL 22-Aug-78 11:26:38
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: STG GLOB MONSYM MEXEC JSYSA IMPDV
IMPANX NETWRK TTNTDV
Related TCO's:
Related SPR's:
Problem: WE CAN'T TALK TO ENOUGH HOSTS ON THE ARPANET.
Diagnosis: NOT ENOUGHT BITS IN HOST FIELD OF LEADER.
Solution: IMPLEMENT EXTENDED LEADERS.
TOPS20 Change Order Number 4.1991
Written by: R.ACE 22-Aug-78 17:29:28
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's: 4.1941
Related SPR's:
Problem: EXTRA SMON PARAMETER WAS DEFINED THAT SHOULDN'T BE IN
THE MONITOR
Diagnosis: .SFLBT FUNCTION FOR SMON/TMON JSYS WAS DEFINED AS
SYSTEM-WIDE DEFAULT TAPE LABEL TYPE, BUT IT HAS BEEN DECIDED TO
IMPLEMENT THIS FEATURE ANOTHER WAY.
Solution: REMOVE THE .SFLBT DEFINITION FROM MONSYM AND REMOVE
THE SUPPORTING CODE FROM JSYSA.
TOPS20 Change Order Number 4.1992
Written by: R.ACE 23-Aug-78 10:34:57
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem: SYSJOB LEAVES DEAD SETSPD FORKS LYING AROUND WHEN IT
RUNS SETSPD AS A RESULT OF MTAJB0 BEING NON-ZERO.
Diagnosis: MTASSP ROUTINE IN MEXEC CALLS RUNDIR, WHICH STARTS
SETSPD AND FORGETS ABOUT IT.
Solution: CHANGE MTASSP TO CALL RUNDII, WHICH STARTS SETSPD AND
WAITS FOR IT TO HALT.
TOPS20 Change Order Number 4.1994
Written by: ENGEL 28-Aug-78 16:04:03
Edit checked: NO Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: FFFFP IS SLOW
Diagnosis: FFFFP USES THE RPACS JSYS TO FIND THE FIRST FREE PAGE. THE RPACS
OVERHEAD ACCOUNTS FOR ITS SLOWNESS.
Solution: HAVE FFFFP LOOK AT THE INDEX BLOCKS DIRECTLY.
TOPS20 Change Order Number 4.1997
Written by: FORTMILLER 30-Aug-78 09:11:15
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PHYX2
Related TCO's:
Related SPR's:
Problem: If a Rewind is issued while the DX20 is powered down US.REW is
set in the UDB making the drive unusable.
Diagnosis: US.REW is being set at Start I/O time instead at interrupt level
Solution: Remove setting of US.REW in DX2SIO and move it to the Attention
Interrupt code.
TOPS20 Change Order Number 4.1998
Written by: FORTMILLER 30-Aug-78 09:34:10
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PHYM2
Related TCO's:
Related SPR's:
Problem: None Observed
Diagnosis: US.REW is being set at Start I/O time instead of
at interrupt level.
Solution: Remove setting of US.REW in TM2SIO and add code to
set it in the Attention Interrupt code.
TOPS20 Change Order Number 4.1999
Written by: FORTMILLER 30-Aug-78 09:45:55
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PHYSIO
Related TCO's:
Related SPR's:
Problem: If US.REW is set in the UDB on an OVERDUE TRANSFER the
drive will be unusable.
Diagnosis: US.REW is not being cleared on an OVERDUE TRANSFER.
Solution: Clear US.REW in routine HNGIRB which gets called after
the OVERDUE TRANSER is reported.
TOPS20 Change Order Number 4.2002
Written by: R.ACE 6-Sep-78 13:36:24
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: EXEC1
Related TCO's:
Related SPR's:
Problem: SPLIT INFINITIVE IN HELP TEXT FOR EXEC "TAKE" COMMAND
Diagnosis:
Solution: CHANGE TEXT FROM "COMMA TO ONLY ENTER..." TO "COMMA TO ENTER..."
TOPS20 Change Order Number 4.2003
Written by: OSMAN 6-Sep-78 15:51:55
Edit checked: YES Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND monsym macsym
Related TCO's:
Related SPR's:
Problem: Exec wants to take commands like "CANCEL BATCH F*".
Also, programs want to use COMND to input fields such as PM%CNT.
Diagnosis: COMND usually breaks on characters such as * and %. That is,
when it sees the * it assumes the last character ended the field.
Solution: Do the following:
1) If new bit CM%BRK is on in the COMND function block, interpret contents
of new word .CMBRK (comes after .CMDEF) of the function block to contain
a pointer to a 4-word break mask for the field. Hence by supplying
CM%BRK and a mask that doesn't contain a bit on for "*", you can cause
"*" to be allowed as part of the field.
2) Supply new macro called FLDBK. which is like FLDDB. but assumes 7 args
instead of 6, with sixth being break mask pointer and saventh being
list. (In FLDDB., sixth is list)
3) All macros are in MACSYM now instead of MONSYM, including FLDDB. Ref-
erence to FLDDB. in manuals will have to be changed.
4) All of COMND's field break masks are now defined in MONSYM. These can
be used to come up with custom masks.
Function Name of mask
-------- ------------
.CMKEY KEYB0., KEYB1., KEYB2., KEYB3.
.CMFLD FLDB0.,...
5) Several useful macros are now in MACSYM for specifying special
break masks. BRMSK. accepts five args: the first four are the
break mask to start with, and the last is the set of characters to
remove from the initial mask to create the custom mask.
As an example of FLDBK. and BRMSK., here's how to specify a COMND
function block for reading a DDT symbol from a COMND keyword table.
This function block specifies that the symbol table name is SYMS and
that the symbols are composed of standard keyword characters plus
the characters % (percent) and . (dot) but that - (hyphen) should
not be allowed as part of the symbol.
FLDBK. .CMKEY,,SYMS,,,<BRMSK. KEYB0.,KEYB1.,KEYB2.,KEYB3.,<%.>,<->>
TOPS20 Change Order Number 4.2004
Written by: OSMAN 6-Sep-78 16:11:20
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: yes
Related TCO's:
Related SPR's:
Problem: last tco has typo
Diagnosis: the last tops-20 change order has a typographical error
Solution: brmsk. takes 6 args not 5
TOPS20 Change Order Number 4.2005
Written by: OSMAN 13-Sep-78 10:01:26
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV COMND
Related TCO's:
Related SPR's:
Problem: Video editors want to know how many times the screen has
scrolled due to user typein. RFPOS isn't sufficient, since monitor
clears the RFPOS data everytime input happens (so that auto-XOFF won't
happen). Another reason RFPOS can't work is that if TEXTI is being used
to do input, a linewrap may occur which causes a scroll, but CTRL/U may
be typed which undoes the linewrap but not the scroll. By the time user
program gets control back from the TEXTI, there's no evidence of the scroll.
Diagnosis: More data needed from TTYSRV
Solution: Add new MTOPRS. The following MTOPRs return +1 always, and
only work for terminals.
Set line counter:
1/ designator
2/ .MOSLC
3/ n
This sets the designated terminal's line counter. The line counter gets
incremented by the monitor everytime a linefeed is output to the terminal.
The monitor only clears this counter when a line becomes active.
Read line counter:
1/ designator
2/ .MORLC
Returns current value of line counter in AC3.
Set line maximum:
1/ designator
2/ .MOSLM
3/ n
This function sets the maximum value of the line counter seen to far.
The monitor compares the line counter with the maximum every time a linefeed
is typed, and copies line counter to maximum if it is larger.
When TEXTI moves the cursor up on screen terminals, it decrements the line
counter.
Read line maximum:
1/ designator
2/ .MORLM
This function returns the current value of the line maximum.
As an example of how to account for scrolling on a screen terminal, you can
set the line counter and maximum to 0. Then do TEXTI and then read the
line counter maximum. If the cursor was at the bottom of the screen when the
TEXTI started, the resultant maximum will show how many scrolls have happened.
TOPS20 Change Order Number 4.2007
Written by: DBELL 14-Sep-78 16:56:42
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem:
If a /INCREMENTAL save is done with DUMPER, and then a file is renamed,
a later /INCREMENTAL save will not save the file again. It should do so,
since the new file name was never saved on any tape.
Diagnosis:
Rename code copies lots of FDB entries as they were in the original
file. In particular, the backup words get copied. Thus DUMPER thinks
that the new name has already been backed up.
Solution:
Change the masks in the RENMSK table for the .FBBK0-.FKBK4 words of
an FDB, so they do not get copied from the old file.
TOPS20 Change Order Number 4.2008
Written by: GILBERT 14-Sep-78 20:54:40
Edit checked: YES Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IO
Related TCO's:
Related SPR's:
Problem:
Multiple ESOUTs cause error messages to be seperated with blank lines.
Diagnosis:
ESOUT always prefixes the message with a CRLF.
Solution:
Don't do so if already at beginning of line.
TOPS20 Change Order Number 4.2013
Written by: R.ACE 19-Sep-78 06:45:49
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK
Related TCO's:
Related SPR's:
Problem: SOFTWARE INTERRUPTS PENDING WHEN EIR JSYS IS EXECUTED
DON'T GET "SEEN" UNTIL QUITE A BIT LATER IN TIME.
Diagnosis: MISSING CHKINT IN EIR JSYS CODE
Solution: ADD CHKINT TO EIR JSYS
TOPS20 Change Order Number 4.2019
Written by: OSMAN 22-Sep-78 12:54:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: If you type ESC at the beginning of a quoted string, COMND
barfs because it's not a quote!
Diagnosis: COMND is stupid
Solution: Use CMRFLD instead of CMCIN to read the first character.
CMRFLD knows to buzz if it sees ESC.
TOPS20 Change Order Number 4.2021
Written by: R.ACE 26-Sep-78 08:17:08
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: WHEN DOING OPENF JSYS, IF BITS OF%RD, OF%WR, OF%EX, AND OF%APP
ARE ALL RESET, OPENF RETURNS +2 (INDICATING THAT IT SUCCEEDED IN OPENING
THE FILE) WITH THE FILE NOT OPEN.
Diagnosis: OPENF JSYS HANDLER WAS CODED TO DO THIS!
Solution: CHANGE OPENF JSYS HANDLER TO RETURN OPNX13 (INVALID ACCESS
REQUESTED) IF ALL OF THE BITS NAMED ABOVE ARE RESET.
TOPS20 Change Order Number 4.2022
Written by: R.ACE 26-Sep-78 08:40:04
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: RELD JSYS PERMITS RELEASING OF A DEVICE WHEN THE USER HAS A
JFN OPEN TO THAT DEVICE. IT WAS CONCLUDED IN THE MONITOR MEETING ON
SEPT. 25, 1978 THAT AN ERROR SHOULD BE RETURNED WHEN RELD IS ISSUED
TO A DEVICE THAT THE USER HAS OPEN.
Diagnosis: RELD JSYS WAS CODED TO OPERATE IN THIS FASHION.
Solution: CHANGE RELD TO RETURN ERROR DEVX6 (JOB HAS OPEN JFN ON DEVICE)
IF RELD IS ISSUED TO A DEVICE WITH DV%OPN BIT SET IN DEVCHR TABLE.
TOPS20 Change Order Number 4.2023
Written by: OSMAN 26-Sep-78 13:29:07
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: EXEC
Related TCO's:
Related SPR's:
Problem: WE'RE NOT GOING TO KEEP DOUBLE LINE LOGIN
Diagnosis: TOO BIG A CHANGE
Solution: GO BACK TO SINGLE LINE
TOPS20 Change Order Number 4.2025
Written by: MCLEAN 26-Sep-78 18:01:22
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: .ASND RETURNS AN ERROR CODE OF 0 FOR TTY'S
Diagnosis: BAD RETERR MACRO
Solution: ADD ,
TOPS20 Change Order Number 4.2026
Written by: DBELL 28-Sep-78 15:01:59
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem:
Indirect files lose if they are full of nulls.
Diagnosis:
Code does not eat them.
Solution:
Do so. At CMIND1+3, eat ignore nulls.
TOPS20 Change Order Number 4.2027
Written by: DBELL 1-Oct-78 21:57:08
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TIMER SCHED STG MEXEC
Related TCO's:
Related SPR's:
Problem:
People want to know what time a job started, in an unprivileged way.
Diagnosis:
No way.
Solution:
Add a new argument to GETJI, which is the time at which this job
was initialized. Notice this has nothing to do with being logged in
or not.
TOPS20 Change Order Number 4.2028
Written by: R.ACE 2-Oct-78 10:02:16
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: OCCASIONALLY, THE SYSTEM DEVICE ALLOCATOR IS NOT INFORMED
WHEN ALLOC'ED DEVICES ARE RETURNED VIA RELD OR LOGOUT.
Diagnosis: SPURIOUS INSTRUCTION IN RELDEV ROUTINE IN JSYSA AT RELDD0+11:
JUMPE Q1,RELDD1
Solution: REMOVE INSTRUCTION
TOPS20 Change Order Number 4.2032
Written by: R.ACE 4-Oct-78 15:49:36
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DATIME
Related TCO's:
Related SPR's:
Problem: WHEN USING IDCNV JSYS WITH IC%JUD (JULIAN DATE) SET, THE
FOLLOWING INVALID ARGUMENTS ARE ACCEPTED BY IDCNV IN THE RIGHT
HALF OF AC2, WITHOUT AN ERROR RETURN:
1. ZERO
2. A VALUE OF 366 FOR A NON-LEAP YEAR
FURTHER, IF A VALUE GREATER THAT 366 IS GIVEN, THE ERROR CODE THAT
IS RETURNED (DATEX2 - MONTH IS NOT LESS THAN 12) DOES NOT APPLY
BECAUSE THE MONTH WASN'T SPECIFIED.
Diagnosis: CODING DEFICIENCY
Solution: RETURN ERROR DATEX7 (JULIAN DAY OUT OF RANGE) IF IC%JUD
IS SET IN AC4 AND THE RIGHT HALF OF AC2 CONTAINS ZERO OR A VALUE
GREATER THAN THE MAXIMUM FOR THE SPECIFIED YEAR.
TOPS20 Change Order Number 4.2033
Written by: KIRSCHEN 6-Oct-78 13:58:10
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DTESRV KDPSRV
Related TCO's:
Related SPR's:
Problem:
There is no CPU-independent way to obtain the protocol
status of communications lines. (The current .BTSTS
function of the BOOT JSYS does not suffice because it
refers to DTE20's).
Diagnosis:
A new BOOT JSYS function is needed to provide the required
information.
Solution:
Add BOOT JSYS function .BTLST, Return Line Status.
See Release 4 Network Control Facilities Spec for
more info.
TOPS20 Change Order Number 4.2034
Written by: MILLER 6-Oct-78 14:04:05
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's:
Problem: Job on the CTY is detached on power fail recovery.
This causes the operator jobs to be detached and logged-out
Diagnosis: As above
Solution: Don't detach CTY job on power-fail recovery
TOPS20 Change Order Number 4.2036
Written by: OSMAN 6-Oct-78 18:41:44
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TIMER
Related TCO's:
Related SPR's:
Problem: The TIMER JSYS elapsed-time feature doesn't give interrupts at the
right time.
Diagnosis: code at TIMSC3 was using T3 instead of T2. Also, at TIMSC3+5,
result was being stored in TMMLTM instead of correct TMDTTM.
Solution: Make the code at TIMSC3 use T2 instead of T3. Also, at TIMSC3+5,
store the result in TMDTTM instead of TMMLTM.
TOPS20 Change Order Number 4.2038
Written by: R.ACE 9-Oct-78 10:54:34
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem: FILE RENAME JSYS (.RNAMF) DOES NOT ALWAYS UPDATE
REMAINING-PAGES COUNT FOR SOURCE AND/OR DESTINATION DIRECTORIES.
Diagnosis: MISSING CODE TO UPDATE PGLFT ENTRY OF SOURCE DIRECTORY,
AND INCORRECT CODE TO UPDATE PGLFT ENTRY OF DESTINATION DIRECTORY.
Solution: ADD AND REPLACE CODE TO CORRECT THE ABOVE PROBLEM.
TOPS20 Change Order Number 4.2041
Written by: ZIMA 11-Oct-78 22:27:27
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DSKALC
Related TCO's:
Related SPR's: 12227
Problem:
Last entry in the BAT block is not used. It is left zero and so
CHECKD screams about address 0 marked in BAT blocks.
Diagnosis:
Code in ADDPR routine increments count of monitor added entries
and then calls BATSPC to find out if any space left. BATSPC says
no for last entry position, not knowing of bumped count, so ADDPR
does not write into entry.
Solution:
Test for room in BAT block prior to incrementing count, adjust
address calculation based on room free.
TOPS20 Change Order Number 4.2044
Written by: ZIMA 14-Oct-78 20:15:43
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DSKALC
Related TCO's:
Related SPR's: 12020
Problem:
The HOMHOM word of the homeblocks is being written differently
by different people. Documentation calls for <this block#>,,<other#>
and the installation floppies do this. CHECKD ends up with the
form <other block#>,,<this block#>.
Diagnosis:
Routines CPYH1, CPYH2 in DSKALC don't keep track properly.
Solution:
fix CPYH1/CPYH2 to write the correct HOMHOM word.
TOPS20 Change Order Number 4.2045
Written by: DBELL 16-Oct-78 17:34:12
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's:
Problem:
If confirmation is desired on a GTJFN, comma works as a confirmation
error if recognition is used, but if recognition is used the comma
causes an "invalid confirmation character" error.
Diagnosis:
GTJFN only checks for the "TERMS" class of characters at ENDL33 for
valid confirmation chars. But comma is in the different class "COMMA"
and so loses.
Solution:
Check for the COMMA class also.
TOPS20 Change Order Number 4.2046
Written by: OSMAN 16-Oct-78 17:46:40
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND tv monsym
Related TCO's:
Related SPR's:
Problem: If user types a non-space as first command character to TV, then
rubs it out and types a space, TV fails to print more.
Diagnosis: TV does BIN to read the first command character, then TEXTI
to read the rest if the first one wasn't special. However, TEXTI
doesn't return if that first one is deleted.
Solution: Implement RD%BEG in TEXTI (bit 9) which causes TEXTI to return
when .RDBKL pointer is reached, and TEXTI is about to wait for
more input.
Without RD%BEG set, TV waits for a break character to be typed,
regardless of the existence of .RDBKL.
TOPS20 Change Order Number 4.2050
Written by: KIRSCHEN 18-Oct-78 14:27:47
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DTESRV kdpsrv
Related TCO's:
Related SPR's:
Problem:
New BOOT JSYS function .BTLST is not really needed.
Diagnosis:
Incorrect analysis.
Solution:
Remove this new function.
TOPS20 Change Order Number 4.2060
Written by: OSMAN 23-Oct-78 16:17:31
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Related TCO's:
Related SPR's:
Problem: See PS:[DOC-SPECS]LOCATE.MEM
Diagnosis:
Solution:
TOPS20 Change Order Number 4.2063
Written by: MCLEAN 23-Oct-78 20:51:46
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: NEED BATCH STREAM NUMBER IN JOB DATA AREA
Diagnosis: ADD WORD SETTABLE BY BATCON
Solution: ADD NEW GETJI SETJB FUNCTION FOR THIS FEATURE
SEE GAL GROUP FOR DETAILS OF HOW BITS ARE IMPLEMENTED.
TOPS20 Change Order Number 4.2064
Written by: MURPHY 24-Oct-78 16:06:26
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DIRECT MONSYM
Related TCO's:
Related SPR's:
Problem: Putting pack with large directories on system
without them produces non-informative bugchks.
Diagnosis: Insufficient differentiation of failed checks
in DR0CHK.
Solution: Regular error return and specific error code for
directory too large.
TOPS20 Change Order Number 4.2065
Written by: DBELL 25-Oct-78 16:17:40
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM
Related TCO's:
Related SPR's:
Problem:
If a PMAP of another process fails because the share count in the
fork's SPT entry for his page table is full, the error returned is
PMAPX6, which is quota exceeded. This is incorrect, since he was
not writing to a file.
Diagnosis:
Wrong code.
Solution:
Use the correct error code, one of the "insufficient resources" codes.
TOPS20 Change Order Number 4.2066
Written by: DBELL 26-Oct-78 09:50:41
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's: 11952
Problem:
If a BUILD command is given which specifies a particular directory
number, and that directory does not exist, the command fails with the
"No such device" error message, if no device was specified.
Diagnosis:
The user's string is examined to parse the device name. But a null
name is illegal.
Solution:
Default the device name to DSK: if it was not specified.
TOPS20 Change Order Number 4.2068
Written by: OSMAN 26-Oct-78 14:08:15
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: exec
Related TCO's:
Related SPR's:
Problem: If you type CTRL/U out of LOGIN command during password field,
echoing remains off for next command (which is usually another LOGIN command)
Diagnosis: Exec was neglecting to turn echoing on before command
Solution: turn on at appropriate times (but be careful not to turn it on
more often as necessary for efficiency reasons)
TOPS20 Change Order Number 4.2071
Written by: R.ACE 27-Oct-78 14:04:19
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IPCF
Related TCO's:
Related SPR's:
Problem: IF THE TOP FORK OF A JOB DOES AN MRECV WITH -1 IN THE .IPCFR
WORD OF ITS PDB (I.E. RECEIVE MESSAGE FROM ANY PID OWNED BY THE FORK
EXECUTING THE MRECV), IT CAN RECEIVE MESSAGES FROM JOB-WIDE PIDS THAT
IT DIDN'T CREATE AND DOESN'T KNOW OR CARE ABOUT. THIS LOSES BECAUSE
THE TOP FORK (USUALLY THE EXEC) GETS STUFF IT DOESN'T WANT, AND THE
PROGRAM THAT IS USING THE JOB-WIDE PID DOESN'T GET ITS MESSAGES.
Diagnosis: IPCF MAKES JOB-WIDE PIDS APPEAR AS IF THE TOP FORK OF THE
JOB OWNS THEM (SO WHEN THE JOB LOGS OUT, THE TOP FORK IS KILLED AND
THE JOB-WIDE PIDS GO WITH IT).
Solution: MAKE THE FNDNMF ROUTINE IN IPCF BYPASS JOB-WIDE PIDS DURING
ITS SEARCH FOR PIDS FOR THE CALLING FORK.
TOPS20 Change Order Number 4.2072
Written by: R.ACE 28-Oct-78 08:41:09
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IPCF
Related TCO's:
Related SPR's:
Problem: WHEN DOING MSEND WITH IP%CPD AND IP%JWP SET TO CREATE A
JOB-WIDE PID, THE PID IS FLAGGED AS BEING JOB-WIDE; HOWEVER, WHEN
THE CREATING FORK IS KILLED, THE PID DISAPPEARS TOO.
Diagnosis: WHEN MSEND CREATES A PID, IT ALWAYS SETS THE PID'S OWNING
FORK TO BE THE FORK THAT IS DOING THE MSEND.
Solution: CHANGE MSEND TO SET THE OWNING FORK TO BE THE TOP FORK OF THE
JOB IF THE IP%JWP FLAG IS SET IN THE PDB.
TOPS20 Change Order Number 4.2074
Written by: OSMAN 30-Oct-78 16:57:09
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MONSYM COMND
Related TCO's:
Related SPR's:
Problem: $ AND _ ARE ALLOWED IN FILESPECS BUT NOT IN DEVICE OR LOGICAL
NAMES
Diagnosis: COMND WAS USING SAME ROUTINE FOR READING DEVICES AS ARBITRARY
FIELDS.
Solution: MAKE COMND USE NEW DEVBRK MASK.
TOPS20 Change Order Number 4.2075
Written by: KIRSCHEN 31-Oct-78 14:47:05
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: monsym
Related TCO's:
Related SPR's:
Problem:
Incorrect solution implemented for missing facility.
Diagnosis:
Fuzzy thinking.
Solution:
Implement function provided by .BTNOD argument of .BTSTS
in NODE JSYS as per spec "RELEASE 4 NETWORK CONTROL FACILITIES",
# DBK-78-003-01, and remove .BTNOD.
TOPS20 Change Order Number 4.2076
Written by: DBELL 31-Oct-78 16:28:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
If a program does a BKJFN while it is detached, system will crash
with a SECNX BUGHLT.
Diagnosis:
BKJFN calls LCKTTY with a line number of -1, since the job is
detached. This causes the error when that value is indexed by.
Solution:
Have BKJFN check for a detached terminal, and fail in that case.
There may be similar errors in other JSYSes, but I can't find them
in a reasonable search.
TOPS20 Change Order Number 4.2078
Written by: OSMAN 4-Nov-78 02:07:45
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: If you try to TBADD an entry which has a flag word (in front of
the string), you lose.
Diagnosis: TBADD was always assuming the string started immediately.
Solution: Use CHKTBS in XTADD instead of merely conjuring up a byte pointer.
TOPS20 Change Order Number 4.2081
Written by: ZIMA 6-Nov-78 11:23:48
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MFLOUT
Related TCO's:
Related SPR's:
Problem:
Calls to the routines in MR (EDFAD.,EDFMP.,EDFDV.) destroy the
P5 register FLOUT thinks is preserved and uses to hold errors.
Results in garbage error numbers returned on column overflow,etc.
Diagnosis:
As above.
Solution:
For the time being, preserve the error register across these calls.
TOPS20 Change Order Number 4.2083
Written by: OSMAN 7-Nov-78 16:13:31
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem: Consider the program that wants to run an exec and make the user
think it's a top-level exec. The program interrupts on inferior process
termination, and types "?No higher command level" when the exec halts (due to
POP command), and restarts the exec.
The problem is, if user types LOGOUT to the exec, he sees the beginning of the
"?no higher..." message and then gets logged out.
Diagnosis: The monitor implements the LGOUT JSYS by interrupting the top-level
fork in the job, and halting the other fork that did the LGOUT JSYS. This
gives an interrupt to a superior trapping termination.
Solution: Change LGOUT to do a WAIT instead of HALTF.
TOPS20 Change Order Number 4.2088
Written by: R.ACE 15-Nov-78 11:33:55
Edit checked: YES Document: YES
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's: 4.2098
Related SPR's:
Problem: NEED SYSTEM-WIDE DEFAULTS FOR MAGTAPE PROCESSING
Diagnosis:
Solution: ADD NEW FUNCTION TO SMON AND TMON JSYS'S:
.SFTDF==:60 SET/READ TAPE MOUNT DEFAULT PARAMETERS
ARGUMENT WORD IS AS FOLLOWS:
SYMBOL MASK CONTENTS
------ ---- --------
SF%TDN 777B8 DEFAULT DENSITY CODE (SAME VALUES AS MTOPR .MOSDN)
SF%TDR 777B17 DEFAULT DRIVE TYPE
.TMDR9==:1 9-TRACK
.TMDR7==:2 7-TRACK
SF%TLT 777B26 DEFAULT LABEL TYPE
.LTUNL==:1 UNLABELED
.LTANS==:2 ANSI
.LTEBC==:3 EBCDIC
.LTT20==:4 TOPS-20
TOPS20 Change Order Number 4.2089
Written by: DBELL 16-Nov-78 10:42:03
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 12309
Problem:
Typing a few control-g's, followed by a control-u causes the cursor
to move upwards a line when it shouldn't do so.
Diagnosis:
FNDLIN routine in COMND is wrong. It doesn't count ^G's as a non-
printing character, and it doesn't check the last character in the
string anyway. Thus its count returned is wrong, and this makes
its callers do wrong things.
Solution:
1. When checking the COC bits for a control character, count chars
echoed literally as non-printing since the cursor is not moved then.
2. Change the loop around so that the last n characters on the line
are really looked at, instead of the n preceeding the last one.
TOPS20 Change Order Number 4.2097
Written by: ZIMA 23-Nov-78 21:51:32
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's: 12105
Problem:
GNJFN tries to unlock the file and go OKINT even for non-disk
JFNs, causing BUGCHKs for overly decremented INTDF.
Diagnosis:
No check is made for disk file JFN to decide whether to OKINT.
Solution:
Make the check.
TOPS20 Change Order Number 4.2100
Written by: ZIMA 29-Nov-78 19:37:57
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12131
Problem:
CRJOB JSYS with CJ%NAM zero doesn't work. It still wants
a password.
Diagnosis:
CRJOB sets up name but forgets to set up for skip password
check at CRJOB2.
Solution:
If CJ%NAM is zero, set CJ%NPW at CRJOB2.
TOPS20 Change Order Number 4.2101
Written by: ZIMA 29-Nov-78 19:42:47
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12078
Problem:
The REFUSE SYSTEM MESSAGES flag is unchanged across logout/login
which causes the new user to miss system messages.
Diagnosis:
As above.
Solution:
Set terminal to ACCEPT SYSTEM MESSAGES on login just before BP$020.
TOPS20 Change Order Number 4.2102
Written by: R.ACE 1-Dec-78 11:16:27
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SERCOD
Related TCO's:
Related SPR's:
Problem: MTCON NEEDS SYMBOLIC DEFINITION OF SYSERR CODE FOR
CONFIGURATION STATUS CHANGE ENTRY
Diagnosis:
Solution: ADD SYMBOLIC DEFINITIONS TO SERCOD.
DOCUMENTATION NOTE:
THE DEFINITION OF THE CONFIGURATION STATUS CHANGE ENTRY SHOULD BE ADDED
TO APPENDIX B OF THE TOPS-20 ERROR DETECTION, RECOVERY AND REPORTING
REFERENCE MANUAL. THE WRITER IS REFERRED TO THE TOPS-20 LABELED TAPE
PROCESSING SPEC (PS:<DOC-SPECS>LTPROC.MEM) AND THE MONITOR MODULE
SERCOD FOR FURTHER DETAIL. IF ALL ELSE FAILS, COME AND TALK TO ME.
TOPS20 Change Order Number 4.2104
Written by: R.ACE 4-Dec-78 09:29:00
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SYSERR
Related TCO's:
Related SPR's:
Problem: SYSERR ENTRIES MADE VIA THE SYERR JSYS HAVE 4 EXTRA WORDS OF
ZEROS AT THE END. THIS DOES NOT AFFECT THE SYSERR DISPLAY PROGRAM,
BUT IT ISN'T VERY ELEGANT EITHER.
Diagnosis: APPARENT CODING OVERSIGHT IN SYERR JSYS
Solution: CHANGE SYERR JSYS SO IT ALLOCATES THE CORRECT NUMBER OF WORDS
TOPS20 Change Order Number 4.2105
Written by: BELANGER 4-Dec-78 10:00:37
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Related TCO's:
Related SPR's:
Problem:
ON SYSTEMS WITH LARGE NUMBERS OF AUTO-BAUD LINES, A BOOT LOAD CAN
FAIL BECAUSE OF PORT CONTENTION WHEN THE AUTO-BAUD TASK (SETSPD)
IS REQUESTED.
Diagnosis:
THE AUTO-BAUD PROCESS, AS WELL AS THE MODEM ANSWERING PROCESS PAY NO
ATTENTION TO PROTOCOL LEVEL.
Solution:
TOPS20 Change Order Number 4.2106
Written by: BOSACK 4-Dec-78 19:37:22
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PHYP4
Related TCO's:
Related SPR's:
Problem:
SOME RP04 DRIVES MAY NOT YIELD RELIABLE ECC CORRECTION ON FIRST RETRY.
Diagnosis:
TRUE.
Solution:
DELAY APPLICATION OF ECC UNTIL 3 RETRIES HAVE BEEN DONE.
TOPS20 Change Order Number 4.2112
Written by: OSMAN 7-Dec-78 16:17:53
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: In exec, if you try to use RUBOUT or CTRL/W to edit a SAVE command
to change the page bounds, you get "?Invalid guide words"
Diagnosis: COMND was falling into the guideword function without checking
for comment. So, causing a reparse on
@SAVE (ON FILE) FOO.EXE !New file! (PAGES FROM)...
caused guideword function to be upset by "!New file!".
Solution: Instead of setting CMNOD (no default possible) in the COMND
function table, explicitly check for .CMNOI, and if escape didn't
end previous field, do regular comment checking before calling the function.
TOPS20 Change Order Number 4.2113
Written by: DBELL 7-Dec-78 16:50:02
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM
Related TCO's:
Related SPR's:
Problem:
Assigning long file OFNS succeeds when it shouldn't, when the OFN
table begins to get full.
Diagnosis:
Wrong ac used at ASOF4+1 in literal. By chance the ac which is used
almost contains the right value, so the bug is not as bad as it could be.
Solution:
Fix the ac reference.
TOPS20 Change Order Number 4.2122
Written by: OSMAN 15-Dec-78 15:59:39
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: Sometimes CTRL/H typed after an erroneuos command fails to
cause a redo.
Diagnosis: COMND does BIN to read the first character on the line
in order to see if it's CTRL/H. However, COMND fails to set the wakeup
bits. Hence the BIN may not wake up when the first character is typed.
Solution: Do SFMOD before the BIN to guarantee that the BIN will wake up.
TOPS20 Change Order Number 4.2123
Written by: ZIMA 16-Dec-78 21:20:13
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's: 12597
Problem:
TERMINAL RAISE does wrong things.
Diagnosis:
Routine TTRAIS makes incorrect comparisons.
Solution:
Do the correct comparisons to convert lower-case alphabetic
to upper-case and convert the old ASCII altmodes to ESCAPE.
TOPS20 Change Order Number 4.2124
Written by: MILLER 18-Dec-78 17:27:10
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: APRSRV
Related TCO's:
Related SPR's:
Problem: COBOL needs access to accounting meters
Diagnosis: As above
Solution: Add METER JSYS
TOPS20 Change Order Number 4.2126
Written by: OSMAN 20-Dec-78 16:02:52
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem:
1) In exec, "SYS @FOO" was failing because COMND tacks the contents
of FOO onto the SYS without leaving the space in! Hence exec sees
SYSxxx instead of SYS xxx, and types "?Unrecognized command".
2) If you type something like "COPY @FOO" to exec, and there's no
FOO file, you see an error message printed twice on your terminal.
3) If you type "PRINT /COPES:3 FOO.BAR", you get "?no such file",
even though the file exists. (You should get "no such switch")
Diagnosis: 1) COMND calls CMRSET when it sees "@". CMRSET resets pointer to
beginning of field, and hence loses the space after "SYS".
2) The exec is calling COMND twice, once to read a real filespec,
and then if that fails, it calls it again to try to parse the
input as a parse-only filespec. COMND prints the error message
for the failing indirect file once for each call.
3) Exec passes COMND a chain of functons. COMND merely returns
the last error encountered, which depends on the exec's ordering
of the links!
Solution:
1) Call CMFSET instead of CMRSET, which leaves the spaces in (by
scanning over them).
2) Change COMND to not print error messages itself upon encountering
indirect file problems. Instead, return whatever error code
happened (usually GTJFN/OPENF error). This change means that
COMND will no longer print "?INDIRECT FILE NOT CONFIRMED". Instead,
that case will return NPXNC just like .CMCFM would.
3) Fix COMND to return "best" error instead of last in chain. "Best"
is one associated with most characters having been read into the
atom buffer before a break character is encountered. This tends
to report a semantic error over a syntactical one. For instance,
/FOO will qualify more as a switch than a filespec since for a filespec
"/" is a break character and hence atom buffer will have nothing
in it, but as a switch, atom buffer will have all four characters in
it.
Items 2) and 3) require documentation.
TOPS20 Change Order Number 4.2128
Written by: OSMAN 20-Dec-78 16:55:41
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: If you type an indirect file, like "SYS @FOO." and
use recognition to finish the filespec, two wrong things occur:
1) No space is echoed after the end of the filespec.
2) If you type rubout, nothing appears to happen because the escape
is still in the buffer!
Diagnosis: COMND checks to see if it's reading an indirect filespec, and
if so, it neglects to execute the code that gets rid of the escape and types
the space.
Solution: Make COMND more consious
TOPS20 Change Order Number 4.2129
Written by: ENGEL 21-Dec-78 10:56:58
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem: LONG FILE BUG - INVALID ACCESS
Diagnosis: MONITOR IS DELETING STATUS BITS ON THE FILE THAT IS STILL
"SHORT". WHEN TWO PROCESSES HAVE THE FILE OPEN AND ONE OF THEM MAKES
THE FILE LONG, THE MONITOR DELETES THE ACCESS BITS FROM THE OFN
OF THE PROCESS THAT THINKS OF THE FILE AS SHORT. WHEN THE "SHORT"
USER TRIES TO ACCESS THE LONG PORTION THE BITS DON'T MATCH.
Solution: LEAVE THE STATUS BITS IN THE OFN.
TOPS20 Change Order Number 4.2131
Written by: DBELL 21-Dec-78 21:15:56
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem:
A JFNS of the spec DSK*: types out as PS:. That is, wild devices
do not output as wild devices.
Diagnosis:
No code to do so.
Solution:
Add it. In the .JFNS routine, check for STRSF and type DSK* if set.
TOPS20 Change Order Number 4.2134
Written by: BERKOWITZ 27-Dec-78 15:04:24
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: COMND TYPES OUT THE TEXT "ONE OF THE FOLLOWING" WHEN THERE IS ONLY
ONE KEYWORD OR SWITCH IN THE TABLE.
Diagnosis: THE TABLE SIZE IS NOT CHECKED ON SWITCH AND KEYWORD DISPLAYS
Solution: BEFORE THE DEFAULT HELP TEXT IS TO BE TYPED OUT CHECK THE SIZE
OF THE KEYWORD OR SWITCH TABLE. IF THE SIZE OF THE TABLE IS 1 THEN
DO NOT TYPE THE MESSAGE "ONE OF THE FOLLOWING", BUT TYPE A " " TO SEPARATE
THE KEYWORD OR SWITCH FROM THE "OR".
TOPS20 Change Order Number 4.2138
Written by: ZIMA 28-Dec-78 20:27:18
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's: 12670
Problem:
GDSTS documents AC3 to have [hardware byte count,,0] on
successful return, but right half of AC3 remains unchanged.
Diagnosis:
HRLM B,3 in literal at MTGTS+5 should be a HRLZM instead.
Solution:
As above.
TOPS20 Change Order Number 4.2139
Written by: DBELL 1-Jan-79 15:18:56
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: CDKLDV
Related TCO's:
Related SPR's:
Problem:
Card reader service reads ahead in cases when it shouldn't, such as
when a read error occurs. Thus the operator doesn't know whether or
not to replace one or two cards.
Diagnosis:
Interrupt code only checked for MO%HEM (hopper empty) to stop. Other
reasons for stopping where not checked.
Solution:
Check them. Stop reading ahead on any type of error condition.
TOPS20 Change Order Number 4.2142
Written by: OSMAN 2-Jan-79 15:07:35
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: lpksdv
Related TCO's:
Related SPR's:
Problem: ks lineprinter crashes system
Diagnosis: no check for page-counter-exceeded condition
Solution: make CLRERR clear error if page-counter-exceeded
TOPS20 Change Order Number 4.2146
Written by: DBELL 3-Jan-79 00:25:45
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's: 12326
Problem:
A RCUSR JSYS WHICH CONTAINS A LONG WILDCARD NAME STRING CAN TAKE MINUTES
TO FINISH, AND WILL BE NOINT THE WHOLE TIME.
Diagnosis:
A STRING IS COPIED FROM THE USER SPACE INTO JSB STORAGE, MEANING THE
PROCESS IS NOINT THE WHOLE TIME.
Solution:
PUT THE STRING ON THE JSB STACK, SO THE JSYS CAN RUN OKINT. CHANGES
ARE MADE TO ROUTINES RCUS0 AND RCUCLN.
TOPS20 Change Order Number 4.2147
Written by: DBELL 3-Jan-79 17:53:12
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 12312
Problem:
NOISE WORDS DO NOT GET TYPED OUT WHEN A QUOTED STRING IS FOLLOWED
BY AN ESCAPE. ALSO, IF ESCAPES OR ^F IS USED INSIDE THE QUOTED
STRING, QUESTION MARKS FOLLOWING THE FIELD DO NOT WORK.
Diagnosis:
NO CHECKS ARE MADE FOR A TERMINATING ESCAPE. ALSO, FLAGS ARE NOT
CLEARED FOR CHARACTERS INSIDE THE QUOTED STRING.
Solution:
FIX CODE A LITTLE. DIDDLE THE XCMQST CODE TO CHECK PROPERLY FOR
ALTMODES OR CONTROL-F'S.
TOPS20 Change Order Number 4.2148
Written by: DBELL 3-Jan-79 19:45:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FREE
Related TCO's:
Related SPR's:
Problem:
JSB STACK IS NOT PROCESSED PROPERLY, SO THAT ENTRIES ARE LEFT ON
THE STACK.
Diagnosis:
ARGUMENT TO JSFRMV IS A HALFWORD VALUE ONLY, BUT CALLERS SOMETIMES
HAVE JUNK IN THE LEFT HALF OF T1. THUS NO MATCH IS MADE FOR THE
DATA ON THE STACK.
Solution:
CLEAR THE LEFT HALF OF THE DATA, SO COMPARES WILL WORK PROPERLY.
THIS IS AT JSFRMV+5.
TOPS20 Change Order Number 4.2155
Written by: GILBERT 8-Jan-79 06:38:30
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: APRSRV BOOT DSKALC FILATS FREE globs
JSYSA JSYSF LDINIT MEXEC mstr PAGEM
params monsym POSTLD PROLOG SCHED STG
EDDT MDDT
Related TCO's:
Related SPR's:
Problem:
There is no room to add any new code to the monitor, especially
the 2020 monitor.
Diagnosis:
The monitor's address space is completely full.
Solution:
Remove the symbols from the monitor's address space. See
HIDSYM.MEM for more information.
TOPS20 Change Order Number 4.2157
Written by: ZIMA 8-Jan-79 17:15:54
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's: 12516
Problem:
RCDIR or RCUSR may incorrectly return any of the calculated
bits RC%NOM, RC%AMB, RC%NMD, RC%WLD.
Diagnosis:
Bits may have been set in directory mode word, which is
simply ORed with calculated bits.
Solution:
Don't let the IORs near RCD555: and RCU190: set the bits
improperly.
TOPS20 Change Order Number 4.2158
Written by: DBELL 9-Jan-79 17:18:22
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 12311
Problem:
COMND function .CMUQS for unquoted strings does not work properly
when the break mask includes the special characters "?", altmode,
and control-F.
Diagnosis:
No checks are made to handle this case.
Solution:
Make the checks. This is in the routine XCMUQS.
TOPS20 Change Order Number 4.2160
Written by: DBELL 12-Jan-79 20:49:22
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DIRECT
Related TCO's:
Related SPR's: 12277
Problem:
If a file is in the process of being created in a directory, and that
file is alphabetically the first one, then any new GTJFN of the files
in that directory with the file spec *.*.* will fail if the GJ%OLD bit
was not specified. (Thus most commands like DIRECT didn't see this
bug).
Diagnosis:
In the process of setting up the first file found for the *.*.* file
spec, you should check for the nonexistance of the files you are
examining. At EXTSCN, this check is bypassed if OLDNF was not set,
since it always thinks you are then creating a new file.
Solution:
Check for stepping, and bypass the check for OLDNF if so.
TOPS20 Change Order Number 4.2161
Written by: ZIMA 13-Jan-79 17:56:24
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 11634 12587 12635 12503 12310
Problem:
IN GENERAL, CONTINUATION LINES DON'T WORK.
Diagnosis:
EXCEPT FOR COMMENTS, AND ONE OTHER CASE, THERE IS NO CODE
TO EVEN CHECK FOR THEM.
Solution:
DISABLE THE TWO CURRENT CHECKS AND PUT IN CODE AT CMCIN TO
DO CONTINUATION LINES. THIS WILL HANDLE ABOUT ALL OF THE CASES
COMND IS THEORETICALLY CAPABLE OF DOING.
TOPS20 Change Order Number 4.2162
Written by: GILBERT 15-Jan-79 08:23:33
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM
Related TCO's:
Related SPR's:
Problem:
The RPACS% JSYS can return PA%CPY in the composite half of the
result even though there is a write-locked indirect pointer
in the chain of indirect pointers being described before the
final copy-on-write page. This means that a write reference to
the page will get an Illegal memory write even though the RPACS%
result had PA%CPY set.
Diagnosis:
RPACS% just ORs together all the copy-on-write bits encountered,
as it is documented to do.
Solution:
Change the code and the documentation to ignore copy-on-write page
pointers after a write-locked pointer has been encountered while
chasing the indirect page pointer chain.
TOPS20 Change Order Number 4.2164
Written by: ZIMA 16-Jan-79 15:23:35
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12717
Problem:
User defined USAGE entries don't work, not even the example
given in ACTSYM.MAC.
Diagnosis:
Code in UFNFNA routine incorrectly reads argument list and
fails to find first USUAR. entry properly. Complementary bug
in USAGE. macro definition in ACTSYM builds incorrect number of
words for space fill entry in argument list, so USAGE JSYS call
for example given in ACTSYM succeeds, but incorrect entry written
to accounting file.
Solution:
Fix code in UFNFNA to know about real form of argument list,
and fix macro definition in ACTSYM to generate proper number of
words in all cases.
TOPS20 Change Order Number 4.2165
Written by: OSMAN 16-Jan-79 17:06:46
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: THE .CMACT function of COMND neglects to check the length of
the typed account string. Hence programs that use loop to copy the
account from the atom buffer to another buffer, looking for null, will
clobber whatever's after that other buffer in the case where the user types
a real long string, unless the program has a special check to monitor the
account string length.
Diagnosis: COMND doesn't check the length of the typed account string
Solution: Make COMND refuse the account string if it is longer than MAXLC,
currently defined as 39. The error given is VACCX1 (accunt string too long)
TOPS20 Change Order Number 4.2166
Written by: HALL 17-Jan-79 11:40:36
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PROLOG FREE
Related TCO's:
Related SPR's:
Problem: IT'S HARD TO LOOK AT RESIDENT FREE SPACE AND TELL WHICH
BLOCKS ARE SUPPOSED TO BE ASSIGNED
Diagnosis: THE BIT TABLE IS NICE AND FAST BUT IT'S HARD TO LOOK AT.
THE FREE BLOCK HEADERS LOOK THE SAME REGARDLESS OF WHETHER A BLOCK
IS ASSIGNED OR RELEASED.
Solution: SET A BIT (RS%REL) IN THE HEADER WORD OF A BLOCK WHEN IT
IS RELEASED
TOPS20 Change Order Number 4.2170
Written by: ZIMA 24-Jan-79 12:50:16
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's: 12580 12583
Problem:
If someone deletes CHECKPOINT.BIN.1 and it doesn't get
expunged, the system will not restart because it gets
UXXCKP BUGHLTs.
Diagnosis:
Code written that way. If file doesn't exist, tries to
create CHECKPOINT.BIN.1, but fails because of deleted version.
Solution:
Add code to handle this special case. If GTJFN to create
new file fails, try to get deleted copy and CHFDB it to undelete.
TOPS20 Change Order Number 4.2171
Written by: HEMPHILL 24-Jan-79 15:04:46
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's: 12634
Problem:
NOPGT0 BUGHLTs when trying to open a file for write.
Diagnosis:
If the FDB already exists, and the FB%NXF and FB%LNG bits are on, and
the file's index block address is zero, the DSKOPN code notes the fact
that the previous file is to be considered invalid. It clears FB%LNG,
but only if the previous file's index block was non-zero, the assumption
being that if there was no previous index block, FB%LNG couldn't be on.
The open for write case finally joins the common disk open code, which
tries to read in the super index block, since FB%LNG is on. However,
this page is all zeros, which causes the BUGHLT.
The assumption that FB%LNG must be off if there is no index block address
must be invalid, since it has happened in the field. Evidently there is
some window in which the system can crash to produce this result in the
FDB. Also, whenever you try to open this file, the crash occurs before
the disk can be updated with an index block address, so the FDB is still
in the same state and will continue to cause crashes in the future.
Solution:
Always clear FB%LNG when re-using an old FDB.
TOPS20 Change Order Number 4.2173
Written by: ZIMA 25-Jan-79 15:43:39
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's: 12065
Problem:
JFNS works differently, at least in the default output modes,
for batch than is does for normal timesharing.
Diagnosis:
In actuality, problem is caused by the ACCES JSYS done by the
batch controller to connect the user to the proper directory.
Since there is no code to allocate free space in another job's JSB,
ACCES just turns off the valid bit for the connected directory
string in the object JSB, so the JSB as a result has no string.
When JFNS gets to the point that it wants to use the string, and it
is not there, it decides to do things like always printing the
connected directory, which is not right.
Solution:
As a crock, near the check at JFNS0 put in code to do an
ACCES to connect the user to his connected directory. That will
correct the missing string, and things should look ok from there
on.
TOPS20 Change Order Number 4.2174
Written by: HALL 26-Jan-79 12:17:01
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PROLOG
Related TCO's:
Related SPR's:
Problem: SYSERR BLOCKS THAT ARE DEFINED FOR BUGCHKS, BUGINFS, AND
BUGHLTS HAVE ROOM FOR 4 AC'S OF OPTIONAL DATA. HOWEVER, NO CODE
PREVENTS THE USE OF MORE THAN 4. THE CODE IN APRSRV THAT WRITES
THE BLOCK WILL SIMPLY OVERWRITE SUBSEQUENT WORDS OF THE BLOCK
IF EXTRA AC'S ARE PROVIDED.
Diagnosis: NO CHECKS ON THE NUMBER OF AC'S ANYWHERE
Solution: MAKE THE BUG MACRO COMPLAIN IF MORE THAN 4 AC'S
ARE PROVIDED (AND IGNORE THE EXTRAS)
TOPS20 Change Order Number 4.2175
Written by: OSMAN 30-Jan-79 16:42:59
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: EXEC MONITOR
Related TCO's:
Related SPR's:
Problem: LA38, LA120 SUPPORT
Diagnosis: NEEDED
Solution: TWO NEW TERMINAL TYPES:
LA120 LA38
----- ----
LENGTH 60 66
------
WIDTH 132 132
-----
LOWERCASE? YES YES
---------- --- ---
PADDING NEEDED? NO NO
--------------- -- --
TABS WORK? YES YES
----------
FORM FEED? YES NO
----------
SYMBOL .TT120(17) .TTL38(18)
------
EXEC COMMAND "TER LA120" "TER LA38"
------------ "TER TYPE LA120" "TER TYPE LA38"
TOPS20 Change Order Number 4.2176
Written by: OSMAN 30-Jan-79 16:50:32
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: MONITOR EXEC
Related TCO's:
Related SPR's:
Problem: SUPPORT LA38, LA120
Diagnosis: O.K.
Solution:
LA120 LA38
----- ----
TOPS20 Change Order Number 4.2177
Written by: OSMAN 30-Jan-79 16:52:42
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Related TCO's:
Related SPR's:
Problem:
Diagnosis:
Solution:
TOPS20 Change Order Number 4.2178
Written by: BOSACK 31-Jan-79 12:06:59
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem: IF NOSKD1 AND OKSKD1 ARE DONE AT PI LEVEL
AND INTERRUPT HAVE BEEN DEFERRED (I.E. INTDFF
HAS BEEN CHANGED TO "XPCW PIPC"), THE MONITOR WILL
CRASH.
Diagnosis: AS ABOVE. PIPC NOT PROTECTED FROM REENTRANT
USES.
Solution: CHANGE ROUTINES NOSK11 AND OKSK11
TO CHECK FOR ANY PI IN PROGRESS. IF SO, DON'T
TRY TO CHANGE NSKED OR INTDF.
TOPS20 Change Order Number 4.2179
Written by: ENGEL 31-Jan-79 15:39:28
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC PAGEM
Related TCO's:
Related SPR's:
Problem: FILE CAN BE DELETED AND EXPUNGED WHIL OPEN
Diagnosis: LONG FILES ARE NOT BEING PROPERLY CHECKED FOR ACCESS BITS IF
THE FILE IS ALSO OPEN AS A SHORT FILE.
Solution: TWO FIXES HAD TO BE MADE.
1) EVERY CALL FOR AN ASSIGNING OF A SUPER INDEX BLOCK HAD TO CHECKED AGAINST
THE SECTION ZERO INDEX BLOCK. THIS REQUITED TWO CHANGES. ASGOFP: ADDED
TO PAGEM AND ASPOFN: ADDED TO DISC.
2) EVERY INDEX BLOCK NOW HAS ACCESS BITS. THIS IS A CHANGE FROM BEFORE WHEN
ONLY THE SUPER INDEX BLOCK HAD THEM.
TOPS20 Change Order Number 4.2185
Written by: DBELL 8-Feb-79 14:53:09
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's: 12667
Problem:
TLINK DOES NOT ACCEPT OBJECT DESIGNATORS WHICH AREN'T A TERMINAL YOU OWN,
EVEN IF YOU ARE A WHEEL.
Diagnosis:
TLINK CALLS CHKTTM, WHICH REJECTS TERMINALS WHICH ARE ASSIGNED TO
ANOTHER JOB.
Solution:
CHECK FOR WHEEL OR OPERATOR, AND IF PRIVILEGED, ALLOW ANY TERMINAL
DESIGNATOR AS THE OBJECT.
TOPS20 Change Order Number 4.2186
Written by: OSMAN 9-Feb-79 09:35:57
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: lpksdv
Related TCO's:
Related SPR's:
Problem: On 2020, system won't come up with lineprinter on-line
Diagnosis: SETSPD sends formfeed to lineprinter during startup if printer
is online. If printer gets a GO error, scheduler test fails to notice it
so it hangs forever.
Solution: In OURTLS, when setting .DVFHE, also set .DVFFE. In addition,
smarten up CLRERR to not do error reset if the problem is that page counter
reached 0.
TOPS20 Change Order Number 4.2187
Written by: DBELL 9-Feb-79 16:24:56
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FREE
Related TCO's:
Related SPR's: 12678
Problem:
Consistancy check when releasing free storage is not correct.
Diagnosis:
At RELFR6+1, the last address of a block os computed. But junk can
be in the left half, invalidating the check.
Solution:
Do the add in the other order so it is correct.
TOPS20 Change Order Number 4.2188
Written by: OSMAN 12-Feb-79 14:50:37
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: Device names always require colons, but node names do not
Diagnosis: To be consistent, require two colons after node names
Solution: In .CMNOD function of COMND, require two colons after node
name. If user tries to do recognition, put the colons and a space in
for him
TOPS20 Change Order Number 4.2193
Written by: DBELL 19-Feb-79 13:46:48
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's:
Problem:
Doing recognition of a device name which is the "DSK*:" name doesn't
type that, it types "PS:" instead.
Diagnosis:
Thats what is stored in the device field.
Solution:
At DFDVTY, check for wild device, and if so output DSK* instead.
TOPS20 Change Order Number 4.2202
Written by: OSMAN 6-Mar-79 10:45:16
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem: RANDOM SKED% JSYS MODS
Diagnosis:
Solution:
1) Make .SKRCV return SK%DRG if batch jobs set to run on dregs queue
2) Change .SKRJP to return share/utilization instead of u/s.
This makes it compatible with .SKRCS.
TOPS20 Change Order Number 4.2213
Written by: OSMAN 8-Mar-79 15:39:47
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: monsym COMND
Related TCO's:
Related SPR's:
Problem: When something like "ASSIGN<cr>" typed, the error is confusing
and says "?Invalid device terminator"
Diagnosis: COMND used to have one check for the colon and the nullness,
so it was easy for COMND to skimp on error analysis. Now that COMND
needs two separate checks, it's sack religious to skimp.
Solution: Create DEVX7 which means "?Null device name given"
TOPS20 Change Order Number 4.2215
Written by: DBELL 15-Mar-79 13:15:27
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's:
Problem:
If DSK*: is the default device, the next wildcard field loses any
non wild characters. thus a spec like DSK*:FOO*.MAC becomes instead
DSK*:*.MAC.
Diagnosis:
In the process of defaulting DSK*:, the wrong flag is cleared at STRDEV.
Solution:
Clear the proper flag. Changes at STRDEV and DEFSDV.
TOPS20 Change Order Number 4.2219
Written by: LCAMPBELL 19-Mar-79 10:37:32
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem: Deleting MAIL.TXT accidentally is unrecoverable.
Diagnosis: Permanent files get expunged when deleted.
Solution: Don't.
TOPS20 Change Order Number 4.2221
Written by: OSMAN 21-Mar-79 11:04:53
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: The new exec can't do DIR MT0: correctly all the time.
Diagnosis: The exec depends on the filespec that the user typed (MT0:
in this case) being in the COMND atom buffer after a .CMFIL function.
Solution: hold it, not yet...
DIAGNOSIS (cont.):
The atom buffer seems to have a carriage return after the filespec.
The problem is that COMND always assumes that the filespec terminator is
one character long, when in fact it may be two characters if it's crlf.
It used to be only one character in the old days when COMND always told
TEXTI to convert crlf to lf. When COMND was changed to be able to handle
crlf OR lf, .CMFIL was never updated to be prepared to handle crlf.
The solution is to change code in CMFIL0 to measure length of atom buffer
after filespec is read. That way, when GTJFN appends terminator onto
atom buffer, CMFIL0 can look at new length, subtract, and know how many
characters long the terminator is. I like this solution, since no special
crlf check is needed.
TOPS20 Change Order Number 4.2224
Written by: ZIMA 2-Apr-79 21:54:46
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12433
Problem:
If a user does an OPENF for write access, a page table for
section zero of the file is implicitly created. If the user then
tries to write the file via PMAP without specifying PM%WR, the
writes will succeed to pages 0-777, but no others since the lack
of the PM%WR will prevent page table creates. This is especially
bothersome if the user is writing the file sequentially, since his
program bombs out just as things are getting interesting and hard
to recover.
Diagnosis:
As above, the only check is for page table creates, but the
OPENF implicitly creates one. Therefore, another check is needed.
Solution:
Add a check to PMAP to prevent file writes if PM%WR is not
specified. Bomb out with error PMAPX1. Implement just before
PMAP10: where it has been determined that the source is the process
and the destination a file.
TOPS20 Change Order Number 4.2225
Written by: ZIMA 2-Apr-79 22:29:20
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12496 12522
Problem:
Various JSYSes (but for the purpose of this TCO, NOUT, FLIN, DFIN)
will not recognize ERJMP/ERCAL following them on an error return.
Diagnosis:
Code doesn't bother, error returns JRST to MRETN.
Solution:
Use EMRETNs on error returns.
TOPS20 Change Order Number 4.2226
Written by: HALL 9-Apr-79 12:55:20
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: STG
Related TCO's:
Related SPR's:
Problem: CURRENTLY THERE ARE TWO MAXIMUM VALUES FOR A JOB'S FORKS:
1. THE NUMBER OF FORKS IN THE JOB
2. THE NUMBER OF RELATIVE FORK HANDLES THAT ANY FORK CAN HAVE
THE RESULT IS THAT A JOB MIGHT HAVE MORE FORKS THAN ANY ONE
FORK CAN HAVE ACCESS TO VIA A FORK HANDLE.
Diagnosis: THE LIMIT ON RELATIVE FORK HANDLES IS LESS THAN THE
MAXIMUM FORKS IN A JOB. THIS WAS NECESSARY TO CONSERVE PROCESS
SPACE BEFORE THE RELEASE 4 DDT PROJECT PROVIDED EXTRA ADDRESS
SPACE.
Solution: MAKE THESE TWO LIMITS THE SAME. THIS WILL ALLOW THE
TOP FORK IN THE JOB TO GET HANDLES ON ALL OTHER FORKS. IT WILL
ALSO ALLOW US TO RUN MORE FORKS UNDER SYSJOB.
TOPS20 Change Order Number 4.2227
Written by: DBELL 9-Apr-79 14:15:32
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: POSTLD
Related TCO's:
Related SPR's:
Problem:
Pages can be missing from the monitor when it is saved.
Diagnosis:
POSTLD incorrectly computes the number of pages in a contiguous
section of the monitor, so the SSAVE JSYS misses a page.
Solution:
Do the computation correctly. Instead of subtracting the addresses
of the end and beginning of a region, subtract the page numbers of the
end and beginning.
TOPS20 Change Order Number 4.2228
Written by: MILLER 9-Apr-79 15:57:01
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TAPE
Related TCO's:
Related SPR's:
Problem: Need a way for MOUNTR to clear VV for an MT
Diagnosis: As above
Solution: Add new MTU function, .MTCVV
TOPS20 Change Order Number 4.2229
Written by: OSMAN 10-Apr-79 13:59:46
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: LINEPR
Related TCO's:
Related SPR's:
Problem: On KS, SETSPD sometimes dies, if a printer error occurs.
Diagnosis: LPTCHK calls PSIRQ with 77, since LPIRP is on.
LPIRP can be on, since LPTOPN inits LPPSI to INIPSI, but doesn't clear LPIRP.
Solution: Don't call PSIRQ if LPPSI is INIPSI (77)
TOPS20 Change Order Number 4.2230
Written by: HALL 12-Apr-79 14:42:02
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC STG
Related TCO's:
Related SPR's:
Problem: NOACB BUGHLT'S DURING SYSTEM STARTUP
Diagnosis: A MULTITUDE OF PROBLEMS
1. USAGE FILE HAD BEEN CLOBBERED, AND USAGE CODE WAS NOT
DEFENSIVE ENOUGH
2. CODE IN RUNDD TO SKIP TO THE NEXT LOCATION ON AN ERROR
DID NOT CLEAR INTERRUPTS, DID NOT RESET ACBAS, DID NOT
RESET THE STACK.
3. CODE IN RUNDD DID NOT WARN YOU THAT CODE HAD BEEN SKIPPED
Solution:
1. FIX USAGE CODE TO TEST FOR NON-EXISTENT PAGE AND ERJMP
2. ADD ROUTINE TO MEXEC TO RESET THE WORLD WHEN AN INTERRUPT OCCURS
3. MAKE MONBK POINT TO THE NEW ROUTINE; MAKE CODE IN RUNDD SET
UP A NEW LOCATION THAT POINTS TO THE NEXT TRANSFER POINT
4. PRINT AN ERROR WHENEVER THE SYSTEM JUMPS TO THE NEXT TRANSFER
POINT
TOPS20 Change Order Number 4.2231
Written by: DBELL 17-Apr-79 15:02:25
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Related TCO's:
Related SPR's:
Problem:
Pages of a disk which were entered into the BAT blocks by diagnostics
such as DDRPI can be used by the monitor anyway.
Diagnosis:
DDRPI uses sector addresses, and checks each sector individually. When
a bad sector is found, a BAT block pair is created, with a count of 1
sector, and that address. However, the monitor uses pages for almost
all of its IO, and checks the sector address of a page against the entries
in the BAT block. Unfortunately, unless the bad sector was the same as
the start of the page, the address will not match. Thus 3/4 of all the
bad sectors are not recognized by the monitor.
Solution:
When checking against the BAT block entries, clear the lower order two
bits of each entry so that the page addresses will match. This is in
the routine SEEADR in DSKALC.
TOPS20 Change Order Number 4.2233
Written by: ZIMA 17-Apr-79 17:21:14
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's: 12707
Problem:
If the BAT blocks for PS: become inconsistent, whenever the
system is booted the operator is notified, but the BAT blocks are
not corrected, and the BAT blocks are not scanned for pages in the
swapping area to be marked as used.
Diagnosis:
The call to CHKBAT in MEXEC at EXEC0+15 does not set up any
flags in T3 for the call. The artifacts left there have both
MI%MSG and MS%FNB set. Only MI%MSG should be.
Solution:
Load up T3 with the right stuff before the call to CHKBAT.
TOPS20 Change Order Number 4.2234
Written by: LCAMPBELL 18-Apr-79 14:11:04
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: IMPDV STG
Related TCO's:
Related SPR's:
Problem: 2020s can't talk to the ARPANET.
Diagnosis: Add a device driver and yet more flavors of monitors.
Solution: IMAN22.MAC is the new driver. The following modules
needed minor changes: IMPDV.MAC, IMPANX.MAC, STG.MAC.
TOPS20 Change Order Number 4.2235
Written by: OSMAN 19-Apr-79 10:13:59
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: lpksdv
Related TCO's:
Related SPR's:
Problem: lineprinter on 2020 sometimes types garbage
Diagnosis: If a "GO ERR" occurs, the code restarts the printer. If
this error happens to be when loading the vfu (as it was!), the restart
prints the vfu data as characters!!
Solution: Don't restart the printer on GO ERR. Handle it like any other error.
(If any future change involves trying to restart the printer on a GO ERR,
then the code must be careful to set the appropriate bits when restarting,
so the printer knows whether a vfu-load or real data is being restarted.)
TOPS20 Change Order Number 4.2240
Written by: OSMAN 23-Apr-79 15:39:42
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: lpksdv
Related TCO's:
Related SPR's:
Problem: Random GO ERR on lineprinter on 2020
Diagnosis: Need to do INIT when done loading ram.
Solution: Implement LP%LOR bit meaning "loading ram". Set it when loading
the ram. Test it and do INIT if it's set when done received. Clear it
when done received too.
TOPS20 Change Order Number 4.2241
Written by: R.ACE 24-Apr-79 06:10:09
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: CHFDB WON'T LET WHEEL OR OPERATOR CHANGE BITS 1-17 IN .FBBK0
Diagnosis: NOT CODED TO DO SO
Solution: CHANGE CHFDB TO LET WHOPER CHANGE ALL BITS OF .FBBK0
TOPS20 Change Order Number 4.2243
Written by: HALL 10-May-79 15:20:13
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK
Related TCO's:
Related SPR's:
Problem: BUGHLT FOR 'ILLEGAL INDIRECT' -- LDB T1,T1 WITH
GARBAGE IN T1
Diagnosis: IN .FFORK, CODE CALLS FFORK1 EXPECTING T1 TO
BE PRESERVED. IT ISN'T.
ALSO, SAME ROUTINE IS MISSING THE FOLLOWING INSTRUCTION:
ADD T2,FKCTYP(T3)
6 LINES BELOW THE CALL TO FFORK1.
Solution: PRESERVE T1 AND ADD THE INSTRUCTION
TOPS20 Change Order Number 4.2248
Written by: DBELL 17-May-79 11:38:34
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
GTFDB will return words off the end of the FDB if the FDB is shorter
than the maximum size. This was introduced when the FDB size was
increased for archiving. Also, CHFDB will hang your job if a word off
of the end of a short FDB is changed.
Diagnosis:
GTFDB has no code to handle varying length FDBs. CHFDB goes to the
wrong location if a word off the end is changed, and so leaves the
directory locked.
Solution:
Rewrite GTFDB to consider varying lengths of FDBs, and go to the right
place in CHFDB.
TOPS20 Change Order Number 4.2252
Written by: DBELL 18-May-79 11:18:09
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's: 4.2525
Related SPR's:
Problem:
If a file spec of the form *.*.-2 is given (to mean the oldest version
of all files) and the oldest version of a file is deleted, then GNJFN
will skip the file type completely. It should return the oldest undeleted
version of a file.
Diagnosis:
GNJFN sets IGDLF always which allows deleted files, and then ignores
deleted files which are returned to it. This loses for -2, since
the low level routines will then not return the undeleted oldest version.
Solution:
Only set IGDLF if we want to see deleted files. This is at GNJFN1+5.
TOPS20 Change Order Number 4.2253
Written by: DBELL 18-May-79 16:26:35
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12171
Problem:
Account validation is incorrect, and doesn't validate directory groups
properly.
Diagnosis:
The user group is looked at when the directory group should be at SCNDN1.
Solution:
Fix the instruction at SCNDN1.
TOPS20 Change Order Number 4.2254
Written by: R.ACE 20-May-79 17:37:37
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: BUGS IN .ARSST FUNCTION OF ARCF JSYS:
1) FILE CONTENTS DO NOT GET DELETED
2) JFN LOCK DOESN'T GET CLEARED
Diagnosis: ARCF WAS CALLING FSHFIL WITH GARBAGE IN T4
Solution: ADD MISSING INSTRUCTION BEFORE CALL TO FHSFIL.
INSTRUCTION IS... MOVE T4,P3
TOPS20 Change Order Number 4.2256
Written by: OSMAN 22-May-79 13:22:58
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: IO
Related TCO's:
Related SPR's:
Problem: We can't read tapes from the rest of the world
Diagnosis: The rest of the world has text lines as individual records with
no crlf's on the tape. When we try to copy that tape to disk, we lose the
interrecord gaps.
Solution: Have the monitor detect that such a tape is being read, and if
so, give crlf's to the user after each record.
TOPS20 Change Order Number 4.2257
Written by: SCHMITT 23-May-79 10:15:32
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: THE NIN JSYS IGNORES CARRIAGE RETURNS IN THE INPUT
STRING.
Diagnosis: SAME AS THE PROBLEM
Solution: MODIFY THE BYTE INPUT ROUTINE FOR THE JSYS SO AS TO CHECK FOR
CARRIAGE RETURNS WHEN IT HAS BEEN CALLED FROM THE NIN JSYS AND MARK THAT
A CR HAS BEEN SEEN. WHEN A CR HAS BEEN ENCOUNTERED AND THE NEXT BYTE IS
A LF THEN RETURN THE LF TO THE NIN ROUTINE. IF THE NEXT BYTE IS NOT A LF
THEN DO A BKJFN, MODIFY THE USER'S BYTE POINTER TO POINT TO THE PREVIOUS
BYTE AND RETURN A CR TO THE NIN ROUTINE.
TOPS20 Change Order Number 4.2262
Written by: DBELL 30-May-79 11:14:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's: 11858
Problem:
Doing a GNJFN of the file spec DSK*:<FOO>FOO.FOO.1 where the only
wildcard is the structure, will loop forever returning the same file
over and over.
Diagnosis:
Structures are only stepped when the rest of the file spec stepping
fails. But in this case, no failures occur and nothing is stepped.
Solution:
Have GNJFN set a flag saying it is being done, and in DIRECT stop
returning the same file if the flag was set. This is not the best
possible fix, but that would take rewriting lots of the stepping
code!
TOPS20 Change Order Number 4.2266
Written by: DBELL 1-Jun-79 13:35:10
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM
Related TCO's:
Related SPR's:
Problem:
Release 4 monitor will not work properly on a model A processor, and
crashes with many different BUGHLTs in a few minutes.
Diagnosis:
The new routine in PAGEM called XGCCHK fakes a page fault when a process
has accumulated enough runtime. In faking the page fault, it builds an
extended style flag and PC always. The page fault code then trashes the
PC when it converts the "old style" PC into the new style PC. Thus
on returning from the "page fault", it returns to a random exec mode
address.
Solution:
In XGCCHK, build an old style PC word if running on a model A cpu.
TOPS20 Change Order Number 4.2280
Written by: DBELL 8-Jun-79 13:47:22
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK
Related TCO's:
Related SPR's:
Problem:
If a program is written which creates lots of forks, and each fork
runs that program, the system can die with an ILWRT2 BUGHLT, and maybe
other things.
Diagnosis:
There is a race in CFORK where a check is made to see if any job fork
slots are available, and then later on one is assigned. But the assignment
is not made in NOSKED code, and so they can run out.
Solution:
Move the call to ASSJFK around BP$019 up into the NOSKED code which
checks for enough resources.
TOPS20 Change Order Number 4.2281
Written by: MILLER 9-Jun-79 16:28:09
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem: If a program is doing non-blocking dump mode input
(DUMPI) and a short or long record is encountered
at the EOF, the GDTST subsequently executed returns
with MT%IRL and MT%EOF set, but with a record length
of zero. Because of this, it is not possible to correctly
recover from such an error.
Diagnosis: MAGTAP's GDSTS code always waits for all outstanding
I/O to complete and returns the composite status of
all unreported I/O.
Solution: Fix MAGTAP interrupt routine to abort
outstanding transfers if short record is encountered.
TOPS20 Change Order Number 4.2283
Written by: DBELL 11-Jun-79 19:27:26
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK
Related TCO's:
Related SPR's:
Problem:
You can GET a program, and then CONTINUE it without it ever having
been started. This usually causes the program to bomb out at PC 0.
Diagnosis:
EXEC no longer checks now that the SFORK JSYS has the continue bit.
It expects the monitor to catch the error.
Solution:
Make the monitor catch the error by remembering if the fork has ever
been started. Then make a continue fail unless it has been started.
TOPS20 Change Order Number 4.2288
Written by: OSMAN 14-Jun-79 15:37:38
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: If you type "LOGIN feuhefueh<ESC><ESC><ESC>...", each <ESC>
was causing another "(PASSWORD)" guideword to be printed.
Diagnosis: COMND neglected to strip off the ESC when doing a parse-only
user name parse. Hence reparse saw the ESC and reguided.
Solution: Just before and after CMDIR8, go to CMDIR9 instead of CMDIR2 so
that ESC will get stripped.
TOPS20 Change Order Number 4.2291
Written by: DBELL 18-Jun-79 12:36:49
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem:
When shutting down the system, the message "shutdown complete"
never gets typed.
Diagnosis:
A TTMSG is done to type the string, but it doesn't use a valid
terminal designator. It tries using just the terminal number.
Release 4 requires the designator form.
Solution:
Set .TTDES at HSYS91+7 or so.
TOPS20 Change Order Number 4.2292
Written by: DBELL 18-Jun-79 13:11:04
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
Doing a DELNF using a JFN which is parse-only will crash the system
with a ASTJFN BUGHLT.
Diagnosis:
DELNF does not weed out parse-only JFNs, and so it calls GETFDB which
loses.
Solution:
Check for a parse-only JFN at .DELNF+9 or so.
TOPS20 Change Order Number 4.2298
Written by: LCAMPBELL 19-Jun-79 16:35:33
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem:
GLREM2 BUGHLTs, potentially bits randomly sprayed over the monitor.
Diagnosis:
Race condition in ELOGO (in MEXEC) can cause job to go away while
the LGOUT is being processed. When this happens, ELOGOO picks up
a garbage fork number from JOBPT, and it goes downhill from there.
Solution:
Cause most of the code in ELOGO to run NOSKED. Also cause ELOGO to
check the fork number it gets for legality and just return if invalid
(because UNICK5 in PHYSIO also calls ELOGO and might have the same
race).
TOPS20 Change Order Number 4.2299
Written by: OSMAN 20-Jun-79 14:37:46
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: TV can get in an infinite loop when you type CTRL/V in a command
string.
Diagnosis: TV does TEXTI to input characters, and then does three things
depending on whether TEXTI returned due to exhausted count, break
character, or neither. If it's "neither", TV merely redoes the
TEXTI. The problem is, if TEXTI saw CTRL/V and count is 1, TEXTI
stuffs the CTRL/V back in the input buffer and tries next time.
Hence TV loops with TEXTI continually rejecting the CTRL/V.
Solution: Fix TEXTI to allow CTRL/V even when count is 1. In order to
determine that the first character of the next character in this case
should be quoted, TEXTI scans backwards when it starts, and decides according
to the parity of the consecutive CTRL/V count whether to quote the first
character.
TOPS20 Change Order Number 4.2305
Written by: OSMAN 21-Jun-79 14:41:42
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: monsym
Related TCO's:
Related SPR's:
Problem: Programs can say
Not a quoted string - does not begin with double quote (error code NPXNQS)
Diagnosis: Should say
Not a quoted string - quote missing at beginning or end
Solution: Do it.
TOPS20 Change Order Number 4.2311
Written by: DBELL 26-Jun-79 11:14:13
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's:
Related SPR's:
Problem:
GTJFN was changed to never expand logical names for parse-only JFNs.
This will break user programs, and DUMPER.
Diagnosis:
GTJFN was changed to be this way in release 4 so that network file
transfer programs would not be confused.
Solution:
Implement a new GTJFN bit, G1%SLN, to suppress logical name expansion.
Thus old programs will continue to work as they used to, and programs
which need the real device name specified can set the bit.
TOPS20 Change Order Number 4.2318
Written by: DBELL 30-Jun-79 19:59:41
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IO
Related TCO's:
Related SPR's:
Problem:
The "controlling terminal" designator of 777777 always refers to the
job's terminal, and not the fork's. This is wrong if SCTTY had been
done for the fork.
Diagnosis:
No code to grab the right terminal.
Solution:
Add the code in CHKJFN.
TOPS20 Change Order Number 4.2321
Written by: ZIMA 10-Jul-79 11:59:14
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: ENQ
Related TCO's:
Related SPR's: 13301 13302
Problem: ENQ may append a request to the wrong lock block.
Diagnosis:
If an OFN and user code (5B2+number) hashes to the same value as the
same OFN and ASCIZ string, a new requestfrom another fork for the same
OFN and user code appends this request to the string lock block.
In the routine STRCMP the case of mismatching types is not checked
thoroughly. For the user code, a zero length byte pointer is extracted,
which makes the comparison succeed incorrectly.
Solution:
In STRCMP, after it is determined that the lock block is a string,
make sure the user-provided argument is not a user code, by looking for
the 5B2, which is an illegal byte pointer. The symmetric case of lock
block code and user provided string already works because of the single
word compare.
TOPS20 Change Order Number 4.2330
Written by: OSMAN 16-Jul-79 09:16:28
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FREE
Related TCO's:
Related SPR's:
Problem: Free space gets slowly used up by the monitor.
Diagnosis: The RELRNG BUGCHK, when it occurs, means some free space probably
just got lost. If this were a BUGHLT instead of a BUGCHK, we would get a
dump and could analyze the problem.
Solution: Change it to a BUGHLT
TOPS20 Change Order Number 4.2337
Written by: ZIMA 23-Jul-79 16:24:16
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's: 12897
Problem: Batch jobs are sometimes logged in the USAGE file as timesharing jobs.
Diagnosis:
The batch bit is kept in the data associated with the PTY controlling
the batch job. If the job should become detached for some reason, such as
a ^ECEASE shutdown logging off BATCON before the controlled jobs, there is
no PTY data for the USAGE JSYS to check, and the jobtype then defaults to
timesharing, disappointing customers who want to do differential charging.
Solution:
In the USAGE JSYS, if the terminal state is detached, make a further
check of the BATSTF word in the JSB to see if OB%BSS is set indicating that
a batch stream has been set for the job, and if so, mark it as a batch job.
TOPS20 Change Order Number 4.2339
Written by: DBELL 24-Jul-79 13:51:24
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PHYX2
Related TCO's:
Related SPR's:
Problem:
ON A DX20, IF TWO DIFFERENT TAPE DRIVES BOTH HAVE ASYCHNONOUS
INTERRUPTS AT THE SAME TIME, THEN A P2RAE2 BUGCHK CAN OCCUR,
AND THE TAPE DRIVES STOP WORKING. THIS IS EASILY CAUSED BY RUNNING
A PROGRAM LIKE MULTIO WHICH WILL RUN SEVERAL DRIVES AT THE SAME TIME
DOING THE SAME THING TO EACH.
Diagnosis:
IN ORDER TO PROCESS ASYCHRONOUS INTERRUPTS, THE STATUS OF THE
DESIRED DRIVE MUST BE READ. THIS INVOLVES WRITING ONE OF THE
REGISTERS TO SELECT THAT DRIVE. BUT IF A PREVIOUS ASYCHRONOUS
EVENT HAD BEEN PROCESSED, THE RH20 CAN BE BUSY STARTING IO FOR
THE PREVIOUS DRIVE, AND SO A REGISTER MODIFICATION ERROR CAN
OCCUR.
Solution:
CHECK TO SEE IF IO IS CURRENTLY IN PROGRESS, AND IF SO DO NOT
START IO ON THIS DRIVE, AND DO NOT CONTINUE LOOKING FOR EVENTS.
THEY WILL BE PROCESSED WHEN THE CURRENT IO IS FINISHED ANYWAY.
THIS CHANGE IS AT ASYNS3-4 IN PHYX2.
TOPS20 Change Order Number 4.2345
Written by: R.ACE 25-Jul-79 15:01:00
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IO
Related TCO's:
Related SPR's:
Problem: SIN (AND POSSIBLY OTHER JSYS'S THAT TAKE BYTE POINTERS)
MAY NOT TRANSFER BYTES PROPERLY IF THE USER SUPPLIES A BYTE
POINTER THAT USES AN INDEX REGISTER.
Diagnosis: CODING ERROR IN BYTBLT ROUTINE
Solution: FIX ERROR
TOPS20 Change Order Number 4.2349
Written by: OSMAN 26-Jul-79 15:32:27
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: ASGINT occurs at UFNSAS (USAGE JSYS)
Diagnosis: ASGSWP is called OKINT
Solution: Move the NOINT from RELSWP to before the ASGSWP
TOPS20 Change Order Number 4.2355
Written by: WEISBACH 27-Jul-79 16:21:50
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's:
Problem: ** KS ONLY **
XON CHARACTERS ARE SENT WITHOUT PREVIOUS XOFF.
Diagnosis: CODED TO DO SO
Solution: CHANGE TO SEND XON CHARACTERS ONLY IF XOFF SENT FIRST
TOPS20 Change Order Number 4.2356
Written by: R.ACE 28-Jul-79 12:58:34
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TAPE
Related TCO's:
Related SPR's:
Problem: VOLUME-PROTECTION FIELD IN VOL2 LABEL IS NOT CHECKED WHEN
OPENING FILES ON TOPS-20 LABELED TAPES
Diagnosis: NOT CODED TO DO SO
Solution: ADD CODE TO CHECK VOLUME-PROTECTION FIELD ON TOPS-20 TAPES
TOPS20 Change Order Number 4.2357
Written by: ZIMA 28-Jul-79 16:23:18
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem:
CRJOB creates not logged in jobs connected to PS:<SYSTEM>. This is
a security bug since any unprivleged user then has owner access to the files
on PS:<SYSTEM>.
Diagnosis:
Job is connected to PS:<SYSTEM> in JBFINI, but code around STCRJB+13
fails to connect it elsewhere since the index AC it is using is trashed,
and the code is wrong anyway.
Solution:
Connect the created job to the creator's login directory. The connected
directory cannot be used because a prior structure mount... may be required.
TOPS20 Change Order Number 4.2364
Written by: OSMAN 1-Aug-79 15:17:23
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: If Alec Carlson types @SYS:QSRMAC.UNV to the exec, he gets
?Command too long for internal buffer
but jfn stays around for the file.
Diagnosis: COMND doesn't close the jfn if CMDIBQ bombs due to space exhaustion
Solution: Make CMDIBQ check for indirect file and close it before bombing
TOPS20 Change Order Number 4.2365
Written by: OSMAN 1-Aug-79 16:06:05
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: COMND sometimes clobbers user's AC4
Diagnosis: IMCALL IDTNC causes it
Solution: save user's AC4 around IMCALL .IDTNC.
TOPS20 Change Order Number 4.2367
Written by: OSMAN 2-Aug-79 08:54:20
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: monsym
Related TCO's:
Related SPR's:
Problem: To the exec, if you type "NUL:_<escape>", the escape following the
underscore causes an unending stream of bells.
Diagnosis: The FILBx. breakset for COMND's filespec functions in MONSYM
were missing underscore.
Solution: Add underscore to the set.
TOPS20 Change Order Number 4.2369
Written by: GRADY 2-Aug-79 10:48:27
Edit checked: YES Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DATIME
Related TCO's:
Related SPR's: 13254
Problem:
The odtim jsys returns an error condition when the od%tmz flag is set and
the time zone is not one of the printable, U.S. or Greenwich, zones. The
user, in this case Mailer, does not usually expect an error when trying
to get the current time of day. In this case, Mailer dies at those sites
in Europe, etc. unless the incorrect time zone is set at start-up.
Diagnosis:
The monitor returns the "unprintable time zone" error code when it does not
happen to have an entry in it's tables for that zone. The only "good" time
zones are in the U.S. and Greenwich, and everybody else loses.
Solution:
Fix the monitor to simply not supply a time zone stamp for those zones for
which it has none. This works out better than printing blanks (which the
idtim jsys would not be very fond of), or "inventing" time zone stamps
for those zones lacking them already.
TOPS20 Change Order Number 4.2371
Written by: GRANT 3-Aug-79 09:38:25
Edit checked: NO Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: LAST SEGMENT OF A DECNET LOGICAL LINK MESSAGE WAS BEING LOST.
Diagnosis: THE MTOPR ROUTINE WHICH DOES THE .MOCLZ FUNCTION DID NOT
PROPERLY CHECK FOR SYNCHRONOUS DISCONNECT.
Solution: FIX THE CODE IN THE NTMTCZ ROUTINE SO THAT IT ONLY LOOKS AT
THE LEFT HALF OF T2 WHEN DECIDING TO DO AN ABORT OR A SYNCHRONOUS
DISCONNECT.
TOPS20 Change Order Number 4.2377
Written by: GRANT 8-Aug-79 09:49:38
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: RLJFN JSYS WITH T1/-1 RELEASES ALL CLOSED JFNS IN THE JOB.
Diagnosis: FIX IT SO T1/-1 MEANS RELEASE ALL CLOSED JFNS IN THIS
PROCESS AND ITS INFERIORS.
Solution: THE RLALL ROUTINE WAS MODIFIED SO THAT T1/-1 CAUSES A
CLZFF WITH CZ%NCL.
TOPS20 Change Order Number 4.2381
Written by: TOMCZAK 10-Aug-79 12:06:35
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FILMSC
Related TCO's:
Related SPR's: 13195
Problem: files don't get completely copied over to an unassigned TTY
Diagnosis: code that closes the TTY doesn't care if anything is still happening
Solution: Install a patch at TTYCLZ that will cause CLOSF to dismiss until
the output buffer of the terminal is empty. If the CZ%ABT bit is set, go
ahead and truncate the file.
TOPS20 Change Order Number 4.2382
Written by: OSMAN 10-Aug-79 13:25:29
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: When you type TERMINAL TYPE VT06, you get ?First non-space is not
a digit
Diagnosis: We'd like to get ?Does not match switch or keyword. COMND
tries to give the best error, which would have happened had not there
been a default saying SYSTEM-DEFAULT which causes COMND into thinking the
number error message is better.
Solution: In CMRFLD in COMND, when using the default in the atom buffer,
stop if an illegal character is seen. Also, use a more exact break mask
for numbers (NUMBRK). Hence the initial "S" of SYSTEM-DEFAULT will be
caught as illegal, leaving 0 character in the atom buffer, which is less
than the 4 characters seen by keyword in VT06, hence the keyword error
will rule, which is the one we want.
TOPS20 Change Order Number 4.2386
Written by: R.ACE 11-Aug-79 19:29:53
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TAPE
Related TCO's:
Related SPR's:
Problem: ERROR CODES BEING RETURNED FOR LABELED TAPE OPERATIONS
ARE MISLEADING:
1) "TAPE HDR1 MISSING" RETURNED WHEN LOGICAL END-OF-TAPE ENCOUNTERED
2) VARIOUS INCORRECT ERROR CODES BEING RETURNED WHEN OPERATION
CANNOT BE PERFORMED BECAUSE TAPE POSITION IS INDETERMINATE
Diagnosis: CODED THAT WAY (I WISH TCO WOULDN'T ASK FOR THIS FIELD)
Solution: CHANGE TEXT FOR ERROR CODE IOX24 FRM "TAPE HDR1 MISSING" TO
"LOGICAL END OF TAPE ENCOUNTERED".
ADD ERROR CODE IOX32 - TAPE POSITION IS INDETERMINATE
TOPS20 Change Order Number 4.2387
Written by: ZIMA 11-Aug-79 22:06:52
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's: 12056
Problem: BKUPDF BUGHLTS
Diagnosis:
DUMPO to magtape of a memory page which has been mapped to a
nonexistent file page dies with an illegal memory read instead of
the correct DUMPO error code, and screws up a lot of stuff, leaving
an inconsistent CST. As soon as the user does a RESET or equivalent,
the system dies.
Solution:
In the device dependent code for tape DUMPOs add a routine to
test access to the pages to be output, and return the proper error
code (DUMPX4) if the access fails. NOTE: lots of code in the module
IO to do this sort of checking has been "historical" for a long time
now, and could be REPEAT 0'ed, but that's another TCO.
TOPS20 Change Order Number 4.2388
Written by: GILBERT 12-Aug-79 17:44:28
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: STG PAGEM
Related TCO's:
Related SPR's:
Problem:
1. The monitor will not come up on a 128K system if there are large
gaps between the psects below the swappable code.
2. The content of SWPCOR is wrong.
Diagnosis:
1. The gaps between the psects are mapped as all-zero swappable monitor
pages. They are eventually swapped out (wasting swapping space) when
GCCOR runs and notices that they are never referenced. On 128K systems,
the monitor never gets up far enough to run GCCOR because there are no
free physical memory pages except those between the psects, so there is
no room to read in the swappable monitor.
2. Its old definition was made obsolete by the re-arrangement of the
psects. It was being used to pass information from STG to PGRINI,
which left it pointing into the middle of the NPVAR psect.
Solution:
1. Don't map the psect gaps at all, and put them one the initial RPLQ
built by PGRINI.
2. Delete SWPCOR and both remaining references to it. Monitor programmers
wishing to test an address or page number against psect boundaries can
compare directly against the symbols that define the limits of the
psects, as follows:
If xxxxx is the name of a psect, then the following symbols exist:
xxxxx The psect origin (defined by LINK)
xxxxxZ The last used word of the psect (in the form nnn777)
xxxxxP The first page of the psect, = xxxxx_-PGSFT
xxxxxL The last used page of the psect, =xxxxxZ_-PGSFT
TOPS20 Change Order Number 4.2392
Written by: DBELL 13-Aug-79 13:26:31
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
RLJFN called with -1 can cause trashing of JFN 1, and free core
problems because of releasing junk.
Diagnosis:
RLJFN AOBJNs over all JFNs to release them. But it skips JFN 0 without
fixing the count in the left half. This makes it try to release a JFN
which is beyond MAXJFN, which causes a junk location between JFN 0 and
JFN 1 to be released.
Solution:
Skip JFN 0 properly, at RLALL+2.
TOPS20 Change Order Number 4.2393
Written by: DBELL 13-Aug-79 15:36:42
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: LOOKUP FILMSC
Related TCO's:
Related SPR's:
Problem:
Doing GNJFN on a file spec of the form NUL:<*>FOO.BAR loses. First,
it will always succeed making your program loop. Second, it could
crash the system with a SECNX BUGHLT.
Diagnosis:
1. No code exists to terminate GNJFN on wildcard directories,
thus it always works.
2. In LOOKUP in the routine NAMSTP, reference is made to DIRORA,
which is valid only for disks.
Solution:
1. Add code in FILMSC in NILSET to make GNJFN return failure when
repeated.
2. Check for disks, and skip references to the directory if it
is not a disk.
TOPS20 Change Order Number 4.2395
Written by: DBELL 14-Aug-79 10:28:48
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem:
While doing a SAVE JSYS, if quota becomes exceeded, then the fork will
hang up and use runtime, and cannot be logged out.
Diagnosis:
The JSYS does SOUTs and BOUTs while NOINT, and do not notice errors.
Solution:
ERJMP any errors, and abort if one occurs.
TOPS20 Change Order Number 4.2398
Written by: R.ACE 15-Aug-79 08:45:10
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Related TCO's:
Related SPR's:
Problem: NEED CHANGES TO PACKAGING:
1) REPLACE STANDARD 2020 MONITOR WITH 202-DECNET MONITOR AND
ELIMINATE THE MONITOR NAMED "2020-DECNET".
2) REMOVE ATS FROM THE MONITOR.
Diagnosis:
Solution: IMPLEMENT CHANGES. FOR FURTHER INFORMATION ON PACKAGING,
REFER TO 2102::PS:<DOC-SPECS>PACK.MEM
TOPS20 Change Order Number 4.2405
Written by: DBELL 16-Aug-79 13:18:36
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED
Related TCO's:
Related SPR's:
Problem:
Can get PSISKD bugchks, and cause a JSYS to run CSKED when it should not.
Diagnosis:
If a JSYS aborts without clearing CRSKED, then the next JSYS will run
that way too, and maybe call PITEST which will cause a BUGCHK.
Solution:
Initialize CRSKED whenever a user JSYS is started, at MENTU+7.
TOPS20 Change Order Number 4.2406
Written by: TOMCZAK 16-Aug-79 13:45:32
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's: 4.2407
Related SPR's: 12899
Problem: THE CTY WILL ALWAYS COME UP WITH "RECEIVE LINKS" SET. AS IT CURRENTLY
STANDS, THERE IS NO AUTOMATIC WAY TO CHANGE THIS WHEN DOING THE ATTACH TO THE
OPERATOR JOB RUNNING PTYCON. MANY CUSTOMERS SEEM TO USE THE CTY FOR CONFIDENTIAL
PURPOSES AND THIS POSES A MINOR SECURITY PROBLEM.
Diagnosis: CTY IS TREATED LIKE ANY OTHER TERMINAL AND DEFAULT IS CURRENTLY
TO RECEIVE LINKS.
Solution: PATCH INSTALLED AT THE TLINK JSYS JUST BEFORE SYSINP TO CHECK
FOR CTY STARTUP. THREE OTHER TLINKS OCCUR IN LOGOUT AND TWO DETACH ROUTINES
WHERE THE SAME PATCH IS INSTALLED. IN TTYSRV AT ABOUT TTSET1+20 CODE WAS ADDED
TO HANDLE THE CASE OF TERMINAL ASSIGNING(TCO#4.2407).
TOPS20 Change Order Number 4.2407
Written by: TOMCZAK 16-Aug-79 13:52:21
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's: 4.2406
Related SPR's: 12899
Problem: SEE TCO# 4.2406 FOR A COMPLETE DESCRIPTION OF THIS PROBLEM RELATED
TO TERMINAL LINKING AND THE CTY.
Diagnosis: RECEIVE LINKS BIT IS BEING SET IN TTY ASSIGN CODE
Solution: PATCH INSTALLED AROUND TTSET1+20 TO LEAVE CTY IN REFUSE LINKS STATE.
MOSTLY INSTALLED FOR CONSISTENCY.
TOPS20 Change Order Number 4.2408
Written by: GILBERT 17-Aug-79 03:27:30
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM STG POSTLD
Related TCO's:
Related SPR's:
Problem:
1. If the monitor hits an EDDT breakpoint while DTE I/O is in
progress, the I/O can get wrong data or write into the symbol
table
2. The typeout from POSTLD is hard to read and interpret.
3. POSTLD does not type out the POSTCD psect, nor does it
type out where the symbols will be moved to at runtime.
4. There are several undocumented restrictions on where
various psects can go that are not checked anywhere,
but which will make a system not come up if they are
accidentally violated.
Diagnosis:
1. DTE I/O is done only into or out of the resident free pool,
which is part of the NPVAR psect. When EDDT gets control
for a breakpoint, it switches the monitor's map. Since DTE
transfers are into exec virtual addresses, the DTE starts
transferring to/from addresses in the symbol address space
that correspond to the NPVAR psect in the monitor's address
space. These addresses contain symbols.
2-4. Coded that way.
Solution:
1. Swap the NPVAR and NRVAR psects, so that the NPVAR psect
will be way up at the end of the address space, out of the
way of the symbols. Since the psect will now be mapped the
same in both address spaces, switching the map will not hurt
any DTE I/O.
2. Reformat the POSTLD typeout to be in columns, and put a
big arrow wherever there is an overflow.
3. Type out more information.
4. Install several new error checks, to warn if the monitor
loaded is too big to run on a 128K system, if the symbols
fit at loadtime but won't at runtime, or if the resident
free pool overlaps BOOT's assigned virtual addresses.
TOPS20 Change Order Number 4.2409
Written by: SCHMITT 17-Aug-79 10:37:34
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DSKALC
Related TCO's:
Related SPR's:
Problem: BUGCHKS ASAASG & ASGBPG
Diagnosis: During bit table initialization, swapping sectors are marked
as used before the BAT blocks are checked. If a bad sector was also in
the swapping area, the INIBTB routine would have attempted to mark it used
twice. Thus causing the ASAASG & ASGBPG bugchks.
Solution: Before attempting to mark BAT blocks as used, check to see if the
sector is part of the swapping space, for if it is there is no need to mark
it again.
TOPS20 Change Order Number 4.2410
Written by: OSMAN 17-Aug-79 13:16:08
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: APRSRV
Related TCO's:
Related SPR's:
Problem: The BUGHLT called APRNX1 prints out ac1, but ac1 isn't useful.
Diagnosis: I'd like to not have APRNX1 referenced twice in APRSRV (once for
KS and once for KL), but the KS version seems not to print out ac1.
Solution: Don't print ac1 anymore. Consolidate the bug into one reference.
TOPS20 Change Order Number 4.2411
Written by: GRANT 17-Aug-79 13:26:31
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPPAR NSPINT NSPSRV
Related TCO's:
Related SPR's:
Problem: FORMAT OF DECNET CI MESSAGE IS INCORRECT.
Diagnosis: THE LINK PRIORITY FIELD OF THE INFO BYTE WAS SET TO 2 AND
THE NSP SPEC SAYS IT MUST BE 1. THIS FIELD IS CURRENTLY (DECNET
PHASE II) IGNORED ON RECEPTION BUT THAT WILL NOT BE THE CASE IN
PHASE III. THE CHANGE IS NECESSARY SO OUR PHASE II MESSAGES CAN BE
CORRECTLY READ BY PHASE III SYSTEMS.
Solution: LNKPRI==1 IN NSPPAR
TOPS20 Change Order Number 4.2412
Written by: OSMAN 21-Aug-79 09:35:20
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: all
Related TCO's:
Related SPR's:
Problem: There's no document describing BUGs, and no way to keep it up-to-date.
Diagnosis: Nobody's ever made the effort.
Solution: Move all the definitions out of modules into a central file called
BUGS.MAC. Hence BUGs such as
BUG(HLT,BLECCH,<Everything's gone SNAFU>)
in source modules get changed to
BUG(BLECCH)
with the rest of the definition appearing in BUGS.MAC.
TOPS20 Change Order Number 4.2414
Written by: R.ACE 22-Aug-79 06:42:47
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem: ILLUUO BUGCHK WHEN LOGGING OUT JOB CREATED BY CRJOB
Diagnosis: CODE EXISTS THAT ENTERS LOGO ROUTINE IN SECTION 0. FURTHER
DOWN THE LINE, MONITOR GETS ILLUUO EXECUTING XBLT AT XBLTE IN SECTION 0.
(REMEMBER, XBLT IS A UUO IF DONE IN SECTION 0.)
Solution: PUT SE1ENT AT LOGO
TOPS20 Change Order Number 4.2416
Written by: R.ACE 23-Aug-79 10:13:27
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: DOING JFNS FOR A NON-DIRECTORY DEVICE WITH A NULL EXTENSION
WILL NOT GIVE THE PUNCTUATION FOR THE TYPE FIELD EVEN IF JS%TYP
CONTAINS A 1. THIS IS A LOSER FOR PARSE-ONLY JFN'S, WHERE IT ISN'T
VALID TO MAKE ANY DECISIONS ABOUT THE CHARACTERISTICS OF THE DEVICE
BASED UPON THE CURRENT SYSTEM DEVICE TABLES.
(ALSO, WITH PARSE-ONLY, DEVICES THAT AREN'T IN THE DEVICE TABLES ARE
CLASSIFIED AS NON-DIRECTORY DEVICES.)
Diagnosis: NOT CODED THAT WAY
Solution: CHANGE JFNS JSYS ALWAYS TO PUNCTUATE TYPE FIELD FOR
PARSE-ONLY JFN'S.
TOPS20 Change Order Number 4.2417
Written by: KONEN 23-Aug-79 10:58:24
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: 'DSK*' does not work in the accounts file for validation
Diagnosis: JSYSA is checking the wrong accumulator
Solution: Use the correct accumulator
TOPS20 Change Order Number 4.2421
Written by: DBELL 24-Aug-79 16:21:04
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: phyx2
Related TCO's:
Related SPR's:
Problem:
TUxx drives do not always tell you when they go offline or come
online, and therefore it is possible that a unit will be marked
offline forever. No amount of fiddling with the drive will restore
its use.
Diagnosis:
PHYX2 is happy to call a drive offline or online whenever that is
noticed.
Solution:
Remove all references in PHYX2 to US.OFS, and don't call PHYONL or
PHYOFL. This way, attempts to use the TUxx drives will always be
made, and the user will find out the drive is offline the hard way.
TOPS20 Change Order Number 4.2423
Written by: ZIMA 27-Aug-79 15:57:44
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MSTR
Related TCO's:
Related SPR's:
Problem: MSTR JSYS always returns MS%LIM as zero.
Diagnosis: Mask of returned bits does not include MS%LIM.
Solution: Include MS%LIM in mask. Document for .MSGSS function in
status word .MSGST
TOPS20 Change Order Number 4.2424
Written by: ZIMA 27-Aug-79 17:12:06
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MFLOUT
Related TCO's:
Related SPR's: 13191
Problem: FLOUT (and DFOUT too) loops forever for some inputs.
Diagnosis: Confusion over abominable floating point numbers (ie zero
mantissa and non-zero exponent - a nonstandard zero).
Solution: Convert to standard form of zero at top of FLOUT. Note: perhaps
future improvements to convert or bomb out on nonstandard floating point
numbers (ie as above or unnormalized) are appropriate instead of taking what
the routines give.
TOPS20 Change Order Number 4.2428
Written by: HEMPHILL 29-Aug-79 11:47:38
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM
Related TCO's:
Related SPR's: 13125
Problem:
Systems with low swapping space may experience any of the following BUGHLTs:
J0NRUN, ILGDA1, ILGDA2 and ILPPT1.
Diagnosis:
The OFN that DDMP is currently working on may be inadvertently released
during a section of code that is OKSKED. The OFN has the SPTLKB bit set,
but it could still be released if the share count goes to zero.
Solution:
Protect the OFN by artificially incrementing the share count during that
section of code.
TOPS20 Change Order Number 4.2429
Written by: R.ACE 29-Aug-79 13:47:54
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: STG
Related TCO's:
Related SPR's:
Problem: IF CDPN=0 AND PCDPN>0 THEN CDPSRV GETS LOADED AND THE SYMBOL
"CDPTIM" IS NOT DEFINED IN STG. CDPSRV REFERENCES CDPTIM, SO CDPTIM
APPEARS AS AN UNDEFINED GLOBAL AT LINK TIME.
Diagnosis: STG DEFINES CDPTIM BASED ON PCDPN AND CDPN. IT SHOULD
DEFINE IT BASED ONLY ON PCDPN.
Solution: CHANGE STG
TOPS20 Change Order Number 4.2431
Written by: SCHMITT 30-Aug-79 11:46:45
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MSTR
Related TCO's:
Related SPR's: 13361
Problem: Ugly code which was rumoured to cause looping
Diagnosis: To use .NULIO as a pointer for a SOUT call is not correct.
Furthermore, there is no reason to call STRST in the MSTDIS routine
when all the code wants to do is get the device string anyway.
Solution: Remove the instruction which loads T3 with .NULIO and replace
the call to STRST with a call to the DEVST JSYS.
TOPS20 Change Order Number 4.2434
Written by: GRANT 31-Aug-79 14:18:25
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: TOPS-20 uses incorrect segment size when attempting to use the
segment size requested by the other end of the DECnet logical link.
Diagnosis: NSP protocol violation.
Solution: Do not subtract length of header from the segment size.
TOPS20 Change Order Number 4.2438
Written by: GRANT 4-Sep-79 16:33:24
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: DECnet-20 node numbers are not being set as defined in
n-CONFIG.CMD.
Diagnosis: The default node number is being set at the wrong time.
Solution: Move the setting of the default node number from NSPINI to
NAMINI so that no one touches the node number after SETSPD runs.
TOPS20 Change Order Number 4.2442
Written by: OSMAN 5-Sep-79 10:43:14
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: monsym
Related TCO's:
Related SPR's:
Problem: There's no symbol denoting what field of AC1 contains the jfn for
GET jsys.
Diagnosis: laziness
Solution: Add GT%JFN==:7777B35 to MONSYM
TOPS20 Change Order Number 4.2452
Written by: HALL 10-Sep-79 16:34:17
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem: DOING A CRJOB AND SPECIFYING AN AC BLOCK CAN PRODUCE
A KIBADU BUGCHK
Diagnosis: BLTMU IS CALLED WITH GARBAGE IN THE LEFT HALF OF T2.
T2 WAS SET UP WITH A BYTE POINTER BEFORE A PREVIOUS GTJFN, AND
ONLY THE RIGHT HALF IS SET UP FOR THE CALL TO BLTMU
Solution: ZERO THE LEFT HALF OF T2 BEFORE CALLING BLTMU
TOPS20 Change Order Number 4.2455
Written by: HELLIWELL 11-Sep-79 16:33:05
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TIMER
Related TCO's:
Related SPR's:
Problem: TIMER JSYS FUNCTION 4 (.TIMDD) DOES NOT REMOVE THE PROPER
MILLISECOND REQUESTS FROM THE QUEUE.
Diagnosis: WHEN THE TIME OF THE REQUEST FOR MILLISECOND QUEUES IS
CALCULATED, IT IS MILLISECONDS SINCE NOW. THE TIME IN THE QUEUE IS
MILLISECONDS SINCE MONITOR STARTUP (TODCLK).
Solution: ADD TODCLK TO P2 PRIOR TO CALL TO TDBEFO.
TOPS20 Change Order Number 4.2456
Written by: HELLIWELL 11-Sep-79 16:36:33
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TIMER
Related TCO's:
Related SPR's:
Problem: TIMER JSYS GIVES ERROR RETURN IF CHANNEL NUMBER IS
NOT LEGAL FOR .TIMDD, .TIMBF, .TIMAL FUNCTIONS.
Diagnosis: THERE IS ALWAYS A TEST FOR LEGAL PSI CHANNEL EVEN
WHEN IT IS NOT NEEDED.
Solution: BYPASS TEST FOR LEGAL CHANNEL NUMBER FOR DELETE FUNCTIONS.
TOPS20 Change Order Number 4.2457
Written by: DBELL 11-Sep-79 17:40:51
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: phyx2
Related TCO's:
Related SPR's:
Problem:
Multiple TXO2's off of one DX20 do not work at all. Therefore the
TXO3 option will not work either.
Diagnosis:
PHYX2 has no idea that more than one TXO2 can be connected to a DX20.
The unit number of a drive controlled by a DX20 is an 8-bit quantity,
where the high order 4 bits are the TXO2 number. Thus if more than
one TXO2 exists in a configuration, it will have to be unit 1 or 2 or etc,
thus making drive numbers in the 20's, 30's, etc. But PHYX2 uses the
drive number as an index into K.DUDB table in several places, and this
table is only 16 words long.
Solution:
Do not index into K.DUDB anymore. Instead add the routine DRVSRC to
search all UDBs looking for the correct one. This is only needed when
an asynchronous event occurs. The K.DUDB table is still only 16 words
long, but now it can contain any 16 drives off of a DX20, no matter
which TXO2/TXO3/TXO5 arrangement is used.
TOPS20 Change Order Number 4.2461
Written by: GRANT 12-Sep-79 15:36:03
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: Incorrect information given in NSPRTH BUGCHK.
Diagnosis: Wrong accumulator loaded with port number.
Solution: Get port number into T2 rather than T1.
TOPS20 Change Order Number 4.2462
Written by: DBELL 12-Sep-79 16:39:40
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem:
GTDIR fails for subdirectories even if you have ACCESSED or CONNECTED
to it's superior. This makes the BUILD command fail.
Diagnosis:
In .GTDIR, a call to CHKMNO trashes T1 just becore a call to SUPCHK.
Thus SUPCHK always gives the error return unless you are enabled.
Solution:
Restore the directory number before the SUPCHK call.
TOPS20 Change Order Number 4.2463
Written by: OSMAN 13-Sep-79 09:49:26
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: The command
@DISMOUNT CURDS:
says "Does not match switch or keyword" but should say "no such device".
Diagnosis: COMND attempts to figure out which error is better by measuring
how many characters are read. Supposedly, the colon should be read for device
and not for keyword, and hence CURDS: should look like a device. However,
COMND doesn't put the terminating colon in the atom buffer, so device doesn't
win out over keyword as we'd like.
Solution: When processing suffixes, like the colon in this case, have COMND
give credit to its length. Hence "CURDS:" is longer than "CURDS" so the
correct error results!
TOPS20 Change Order Number 4.2465
Written by: DBELL 13-Sep-79 11:10:18
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PHYM2
Related TCO's:
Related SPR's:
Problem:
The monitor does not know that ANSI ASCII works only for the TMO2
tape controller.
Diagnosis:
Code didn't check to see if the controller was a TMO2, which can
do ANSI ASCII, or a TMO3, which can't.
Solution:
In TM2CON, disallow ANSI ASCII for a TMO3.
TOPS20 Change Order Number 4.2467
Written by: MILLER 14-Sep-79 12:35:25
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DTESRV
Related TCO's:
Related SPR's:
Problem: Occasionally TOPS20 fails to reload the front-end.
Diagnosis: The most severe problem occurs when an open
line, or a line not honoring X-OFF causes B02 crashes in
RSX20F. While all this is happening, TOPS20 exhausts its
DTE queue scheduling pool trying to send X-OFF characters
to the misbehaving line. The result is that the
reload of RSX20F is hung when the start-up message is
sent (this occurs becasue RSX20F immediately crashes again
and never accepts any messages).
Solution: Flush the DTE queue for the master -11 whenever
it is reloaded. While this will not stop RSX20F from
crashing, it will ensure that TOPS20 keeps trying
to fix the problem.
TOPS20 Change Order Number 4.2468
Written by: GRANT 14-Sep-79 14:54:35
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: In order to allow DECnet to be decoupled from the monitor after
Release 4, all NSP messages must have route headers.
Diagnosis: Add route headers to data, link service, and interrupt messages.
Solution: Insert calls to the routine which builds route headers and adjust
the size of free space blocks in which the messages are built.
TOPS20 Change Order Number 4.2469
Written by: MILLER 14-Sep-79 15:30:49
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA STG MONSYM
Related TCO's:
Related SPR's:
Problem: Need an ACJ function for ACCESS and CONNECT.
Diagnosis: AS above
Solution: Add one
TOPS20 Change Order Number 4.2476
Written by: DBELL 19-Sep-79 15:01:51
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: DIAG PHYPAR PHYX2
Related TCO's:
Related SPR's:
Problem:
The SET ONLINE command for newly existant tape drives will not work
with DX20's which have TXO2's with the TMO3 option, or with multiple
TXO2's.
Diagnosis:
The DIAG code to implement setting on-line "knows" the maximum number
of units on a device, and will disallow more. The limit of 20 is not
correct anymore, since a DX20 can have units up to 377. Furthermore,
the code to check the existance of a unit will no longer work since
you cannot index into the KDBUDB tables anymore to pick up the UDB
that is wanted.
Solution:
Implement a new dispatch address for CDBs and KDBs which will check for
the existance of a unit, and range check the unit number at the same
time. Use this dispatch for the DIAG function. This requires chances
in many modules, to handle the new dispatch entry.
TOPS20 Change Order Number 4.2480
Written by: R.ACE 21-Sep-79 07:39:18
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TAPE
Related TCO's:
Related SPR's:
Problem: SECG37 BUGHLT FROM MTU% JSYS
Diagnosis: CLRLBL ROUTINE DOES SETZM 0(T1) WITH TRASH IN LH(T1)
Solution: FIX CODING ERROR
TOPS20 Change Order Number 4.2482
Written by: R.ACE 24-Sep-79 08:34:28
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TAPE
Related TCO's:
Related SPR's:
Problem: MTU% JSYS CRASHES MONITOR WITH LUUMN0 BUGHLT
Diagnosis: MTGET ROUTINE CALLS GETLBU WITH INDEXED BYTE POINTER IN
T1, GETLBU MODIFIES THE INDEX REGISTER AND THEN USES THE BYTE POINTER.
Solution: FIX MTGET NOT TO USE INDEXED BYTE POINTER.
TOPS20 Change Order Number 4.2483
Written by: DBELL 24-Sep-79 14:02:53
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PROLOG
Related TCO's:
Related SPR's:
Problem:
The standard supportable configuration for release 4 says that a
DX20 can use 16 tape drives if the TXO5 or TXO3 options are used.
But the monitor only has table space for 8 drives.
Diagnosis:
MTAN in PROLOG was 8.
Solution:
Increase MTAN to 16 for most monitors, but leave it 8 for 2020's,
since they cannot run with DX20's.
TOPS20 Change Order Number 4.2485
Written by: GRANT 25-Sep-79 10:25:35
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: RELRNG BUGHLT
Diagnosis: Caused by NSPSRV giving a bad address to FREE when releasing
a message at ADIN10.
Solution: Get the correct address of the message block into T1 before
calling RELMES.
TOPS20 Change Order Number 4.2486
Written by: HELLIWELL 25-Sep-79 11:42:48
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's:
Problem: MTOPR FUNCTION .MOTPS DOES NOT WORK (SET INTERRUPTS FOR NON-CONTROLLING TERMINAL)
Diagnosis: CODE IS FULL OF BUGS
Solution: REWRITE CODE AND FIX A FEW OTHER STRUCTURE DEFINITIONS
TOPS20 Change Order Number 4.2487
Written by: MILLER 25-Sep-79 14:23:32
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF JSYSA
Related TCO's:
Related SPR's:
Problem: It is possible to open an unassigned but assignable
device without invoking and ACJ action. This is a violation
of security.
Diagnosis: Missing code.
Solution: Add a new ACJ function, .GOOAD.
This GETOK request will be made whenever an assignable
device not yet assigned to the requesting job is
being opened. It has one user argument, the device designator
of the device being opened.
TOPS20 Change Order Number 4.2490
Written by: DBELL 26-Sep-79 17:40:13
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PHYSIO PHYM2 PHYX2 MONSYM
Related TCO's:
Related SPR's:
Problem:
When an illegal mode or density is used for a tape drive, the
error returned is simply IOX5, which implies that some IO error
ocurred. This does give the user a good handle on the problem.
Diagnosis:
The device drivers for magtape mark all errors as simply device
errors, and don't pass upwards the fact that the hardware cannot
perform the desired action.
Solution:
Define a new bit for GDSTS to return, which says that an illegal
mode or density, etc, was specified, and return that if appropriate.
Also define a new error code to be returned from an IO operation
which says what is wrong.
TOPS20 Change Order Number 4.2491
Written by: WEISBACH 27-Sep-79 12:08:55
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: KDPSRV
Related TCO's:
Related SPR's:
Problem: References to DUP #2 on a single DUP 2020 system can crash TOPS20
Diagnosis: KDPSRV does incorrect range checking on DUP number. It uses KDPN,
the number of DUPs for which this monitor was assembled, instead of KDPNLN,
the actual number DUPs on the system.
Solution: Fix the range check routines to use the correct parameter.
TOPS20 Change Order Number 4.2493
Written by: ZIMA 27-Sep-79 14:41:50
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 12716
Problem:
If the .CMCFM function is not the first one on the list, COMND never
uses it on the case of a simple carriage return following the prompt.
Diagnosis:
Code not written.
Solution:
Chain through the function blocks in the check around NLINE, and if
a .CMCFM block is found, use it. Otherwise, perform as we do now and
simply reprompt.
TOPS20 Change Order Number 4.2494
Written by: ZIMA 27-Sep-79 14:49:16
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's: 13517
Problem:
If one specifies the CM%SDH bit in the first block of a list, and
does not specify a help text, you get an isolated "or" as the first
of the "?" possibilities.
Diagnosis:
Setting the CMQUE2 flag arbitrarily in CMRTYP forces this to happen
when we get to DOHLP.
Solution:
Don't set CMQUE2, designating that we have typed something and need to
do the "or" until we indeed have typed something as indicated by CM%SDH
off or CM%HPP on.
TOPS20 Change Order Number 4.2495
Written by: ZIMA 27-Sep-79 15:25:48
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem:
Wrong error results if NOUT in COMND fails due to other screwy
circumstances.
Diagnosis:
NOUT is followed by ITERR() but error code is not in T1, JFN is. NOUT
returns errors in T3.
Solution:
Change the ITERR to be ITERR (,<MOVE T1,LSTERR>) so the correct error
code gets passed back.
TOPS20 Change Order Number 4.2503
Written by: GRANT 1-Oct-79 10:58:11
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: Monitor trying to release a block of free stroage that was already
released.
Diagnosis: When attempting to remove a node from the DECnet data base REMNDX
and ADIN10 both released the message block.
Solution: Remove the CALL RELMES in REMNDX. This fixes the problem caused
at ADIN10. The failure return to CALL ADDNOD in INSNOD must now
JRST INSERR to release the block since REMNDX no longer releases
the block.
TOPS20 Change Order Number 4.2504
Written by: OSMAN 1-Oct-79 17:07:13
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: LPKSDV
Related TCO's:
Related SPR's:
Problem: Printer on system 4097 stopped printing around Sep 28, 1979
Diagnosis: Printer interrupts are happening while the printer is still printing.
The KS lineprinter interrupt routine assumes that if an interrupt occurs when
no error bits are set, and done is not set, that page counter reached 0, so
it restarts the printer.
However, if the printer is still printing, restarting it produces demand-timeout
and go errors.
Solution: Check for printer still printing when interrupt occurs. If so,
dismiss interrupt without trying to restart the printer.
TOPS20 Change Order Number 4.2507
Written by: DBELL 2-Oct-79 16:38:01
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PHYSIO
Related TCO's:
Related SPR's:
Problem:
The RSTSEK routine in PHYSIO is wrong. Callers of it (only caller now
is DIAG JSYS) can crash the system.
Diagnosis:
RSTSEK doesn't know about KDB's properly, thus P2 and P3 are set up
incorrectly and point to the wrong stuff.
Solution:
Fix RSTSEK to call DGUMAP which knows how to loop over all the units
on a channel properly.
TOPS20 Change Order Number 4.2511
Written by: MILLER 5-Oct-79 16:01:33
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IO
Related TCO's:
Related SPR's:
Problem: Over quota error during BOUT causes
duplication of character in file.
Diagnosis: Code in IO that backs up the user PC is
incorrect. Backing up the PC causes the BOUT
to be executed twice. Since the monitor recovers
from over quota internally, this changing of the
user's PC seems unnecessary.
Solution: Remove code in IO that backs up user's
PC.
TOPS20 Change Order Number 4.2513
Written by: OSMAN 8-Oct-79 09:10:06
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: monsym
Related TCO's:
Related SPR's:
Problem: COMND doesn't allow dollarsigns and underscore in user names
Diagnosis: Monitor does, so COMND should too.
Solution: Fix MONSYM break masks (xxxBn. symbols) to allow those characters in
user names. Note that they already were allowed in filespecs.
TOPS20 Change Order Number 4.2514
Written by: OSMAN 8-Oct-79 09:37:18
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: COMND
Related TCO's:
Related SPR's:
Problem: If you ^O during help output, you don't get reprompted for next input.
Diagnosis: COMND doesn't have special code to cause that to happen
Solution: COMND already has special code to cause the initial prompt for
a command to be seen even if ^O. Make this code a subroutine and call it at
CMRTYP!
TOPS20 Change Order Number 4.2515
Written by: GRANT 8-Oct-79 10:52:25
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: kdpsrv
Related TCO's:
Related SPR's:
Problem: DECnet loopback test didn't work.
Diagnosis: KMC was never receiving the looped start message.
Solution: The DUP maintenance bits for cable and controller loopback
were reversed, fix them.
TOPS20 Change Order Number 4.2523
Written by: TOMCZAK 11-Oct-79 13:30:49
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's: 13588
Problem: Any user can increase their working and permanent storage by creating
a new subdirectory with negative working/permanent quotas.
Diagnosis: The only check currently made on a new subdirectory is only to make
sure that the subdirectory does not exceed it's superior directory's quotas. No
check is made for negative quantities.
Solution: Add a SKIPE instruction around CRDI0P-2 and CRDI0L-2 in the CRDIR
JSYS.
TOPS20 Change Order Number 4.2525
Written by: SCHMITT 15-Oct-79 15:08:47
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: GTJFN
Related TCO's: 4.2252
Related SPR's:
Problem: The patch for SPR 20-11916 broke GNJFN on a group of
files which had a deleted file in the middle of the group.
Diagnosis: The original patch would clear a flag to the version lookup
routine which would signify that the lookup should not notice
the existence of deleted files. If it found one, it would
give an error return telling the GNJFN JSYS that no file with the
specified generation number was found. The GNJFN JSYS interprets this
as meaning that there was no more files which matched the
specifications in the group.
Solution: It is not possible to make GNJFN smarter and interpret
this error return correctly because by this time the internal
organization of the directory may have changed which would cause
great confusion later on and probably BUGHLT.
This means that when stepping version numbers in the lookup
one should tell the lower level lookup routine to return all
deleted files which match the specification and let the GNJFN
routine weed out the deleted ones.
But, the only effective way of performing a GNJFN for groups
of files with specifications of the lowest non-deleted
version (ie. Foo.*.-2) is to tell the device dependent
lookup routine not to consider deleted files. The cure,
therefore, is to clear the IGDLF flag, which tells the device
dependent routine to not consider deleted files in its lookup, only when
a GNJFN on a file spec of the flavor Foo.*.-2 is performed and at no
other time.
TOPS20 Change Order Number 4.2527
Written by: DBELL 16-Oct-79 11:29:55
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PHYSIO
Related TCO's:
Related SPR's:
Problem:
Lots of BUGINFs, BUGCHKs, and BUGHLTs would like to type the channel,
controller, and unit that the bug is concerned with, but there is no
easy way to get this information. Furthermore, those BUGs which try
to do this by hand do it wrong (like OVRDTA).
Diagnosis:
Too much code needed for every bugchk to do it by hand.
Solution:
Add a routine to PHYSIO which will return these quantities given the
CDB, KDB, and UDB pointers, so BUGCHKs can set up the ac's easily.
TOPS20 Change Order Number 4.2528
Written by: GRANT 17-Oct-79 09:00:07
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: Unexpected NSP segment sizes observed during DECnet transfers.
Diagnosis: TOPS-20 not correctly adjusting segment sizes when the other
system requested a larger size.
Solution: Fix CIPSRV to compare with SEGSIZ and not NSPMAX(X).
TOPS20 Change Order Number 4.2529
Written by: OSMAN 17-Oct-79 09:21:26
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: lpksdv
Related TCO's:
Related SPR's:
Problem: On KS, if VFU entry is before RAM entry in 4-CONFIG.CMD, and
the system is brought up with the lineprinter on-line, the system hangs and
never comes up completely.
Diagnosis: It is hanging in the LPTDIS scheduler test, which is waiting for
a lineprinter interrupt to happen, which would either cause an error or say
the printer is done. Unfortunately, the printer is idle, so no interrupt is
forthcoming.
An observation is that the SNDEOF routine seems to be called from LINEPR in
various places in order to wait for printer activity to cease. The KS doesn't
have such a clever SNDEOF routine. Hence when SETSPD sends its formfeed to
the printer followed by vfu data, the vfu data gets started to the printer
while the printer is still actively printing the formfeed.
Solution: The following "solution" seems to allow the system to come up,
but isn't a complete solution.
Instead of making SNDEOF wait for activity to stop, which would probably involve
setting up a LPTDIS scheduler test, put in checks in FESNDS and FESDTA which
cause the requested action (print some characters or load ram or vfu) to be
refused if the printer is not idle.
This seems to allow the system to come up. The real solution, however, would
be to fix SNDEOF.
TOPS20 Change Order Number 4.2540
Written by: MAGRATH 21-Oct-79 18:04:23
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: ENQ
Related TCO's:
Related SPR's: 13481
Problem:
Using the ENQ/DEQ facility to enqueue multiple lock
requests on the same resource using different JFNs can produce
the following two problems:
1. Unexpected PSI interrupts can occur (usually on channel 0
since no channel was specified). These interrupts usually
occur sometime after the locks have been dequeued.
2. A CLOSF to a file for which all locks have been properly
dequeued may fail with an error indicating the JFN is still
locked.
Diagnosis:
These problems occur only when lock requests are nested
for a particular resource. When the second request for an existing
lock is processed:
1. The pointer chain used to string all requests for a single
ENQ call together is broken. This chain is used by the lock
scheduling routine, and since some lock requests for the
current call exist outside the chain, they will not be locked,
and by default will appear to be waiting to generate a PSI
interrupt.
2. The JFN of the original lcok request will remain in the lock
request block, and the JFN of the current request will be lost.
This happens because nesting of the same lock request merely
increments a count, which in turn must be decremented to zero
before the lock is removed. Therefore, CLOSF will either not
find a JFN which has a lock request outstanding, or it will
find a JFN which does not actually have a lock request in the
queue.
Solution:
Add JFN as a criteria for locating lock requests in the
queue. This will cause requests for the same lock on different
JFNs to create seperate lock request blocks, properly maintaining
the 'by request' pointer chain, and allowing CLOSF to properly
detect locked JFNs.
TOPS20 Change Order Number 4.2543
Written by: GRANT 23-Oct-79 15:19:34
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: SCHED JSYSA STG BUGS
Related TCO's:
Related SPR's:
Problem: The monitor needs to insure that the ACJ is processing GETOK
requests in a timely fashion.
Diagnosis: The ACJ should be forced to do RCVOKs within a certain time
period if a GETOK is available.
Solution: Add a RCVOK timer. The monitor will no longer recognize the
presence of an ACJ if the RCVOK timer goes off.
TOPS20 Change Order Number 4.2547
Written by: DBELL 24-Oct-79 16:18:53
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: fulili
Related TCO's:
Related SPR's:
Problem:
If the routines to copy strings from user core to monitor core are
handed a byte pointer with 1B12 set, the monitor loses.
Diagnosis:
Such a byte pointer indicates a double word byte pointer. Thus using
this pointer will trash the next AC unexpectedly, causing all sorts
of problems later.
Solution:
Save the next AC when the byte pointer is used.
TOPS20 Change Order Number 4.2548
Written by: ZIMA 25-Oct-79 10:50:20
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTNTDV
Related TCO's:
Related SPR's:
Problem: NSKDIS BUGHLTs on APRAnet systems.
Diagnosis:
Calling sequence to TTCHI was changed, but changer did not find all
of the callers and fix them to handle the RET/RETSKP returns.
Solution:
Insert a no-op after the call to TTCHI at about NVTUPB+5 so that the
OKSDK1 will not be skipped and cause NSKDIS when IMPBP2 tries to MDISMS.
TOPS20 Change Order Number 4.2550
Written by: ZIMA 25-Oct-79 15:11:32
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: TTDZDV
Related TCO's:
Related SPR's:
Problem:
There is no terminal speed zero for DZ11 lines. The code just
ignores the condition and doesn't touch the speed at all.
Diagnosis:
The DZ11 hardware has no real provisions for setting the speed
to zero. All speed select bit patterns are used (17 is 19.2 kbaud)
so the code did not even bother, and took its default action on bad
speed change requests of doing nothing and keeping quiet. That will
result in people who try to set speed zero on troublesome lines to
be very disappointed.
Solution:
If speed zero is selected, turn off the receiver clock enable
for that line. Preserve the output characteristics of the line,
since there is no way to turn off the output, and it is probably
better to see some output like system messages visibly, than to
change the baud rate and see wrong speed garbage come out.
TOPS20 Change Order Number 4.2551
Written by: DBELL 25-Oct-79 18:50:10
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: phyx2
Related TCO's:
Related SPR's:
Problem:
If the microcode of the DX20 halts, the poor users get stuck waiting
for their IORBs to time out and cause OVRDTA BUGCHKs.
Diagnosis:
The PHYX2 code blindly asks for IO to be done, without caring whether
or not the microcode is still running.
Solution:
In the DX2CON routine, check to see if the microcode is still running,
and if not, do a BUGINF and error return.
TOPS20 Change Order Number 4.2552
Written by: DBELL 25-Oct-79 21:25:04
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem:
SKIP FILE commands can loop and never come back, and in some cases
hang the job so ^C will not work. This happens if some I/O errors
occur during the spacing operation.
Diagnosis:
The routines MTAFFL and MTABFL loop until their desired action is
completed, checking only for tape marks or BOT. Errors are not
checked, and so they never give up.
Solution:
Add calls to CHKERO in both routines to abort on any errors.
TOPS20 Change Order Number 4.2553
Written by: MILLER 26-Oct-79 10:39:41
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PROLOG MEXEC
Related TCO's:
Related SPR's:
Problem: Packaging monitors is made difficult as a result of
the value SYSTAD. Since file checksums are unreliable,
it is not possible to automate a verification procedure.
Diagnosis: As above\
Solution: Remove SYSTAD from POSTLD and MEXEC.
TOPS20 Change Order Number 4.2555
Written by: DBELL 26-Oct-79 14:11:27
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MSTR
Related TCO's:
Related SPR's:
Problem:
TCO 4.2431 broke dismounting of structures.
Diagnosis:
DEVST in the MSTDIS routine is being given a structure number, instead
of a device designator.
Solution:
Replace the DEVST with the original CALL STRST. Then fix the original
bug by making STRST not append a string if the pointer is 0.
TOPS20 Change Order Number 4.2556
Written by: TOMCZAK 26-Oct-79 14:24:10
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem: DELNF NOW HAS BITS IN LH AC1 CAUSING CHKJFN CALL TO FAIL
(WHEN MAYBE IT SHOULDN'T)
Diagnosis: DOING A UMOVE ON AC1 WHEN ALL THAT IS WANTED IS THE RH (CONTAINING
JFN)
Solution: CHANGE .DELNF+1 FROM UMOVE JFN,T1 TO XCTU [HRRZ JFN,T1]
TOPS20 Change Order Number 4.2557
Written by: ZIMA 26-Oct-79 14:45:58
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV FILMSC
Related TCO's:
Related SPR's:
Problem:
Nobody finds out if an ACJ program rejects a baud rate change.
Diagnosis:
Routine TTSSPD has only one return, which is a success return.
Solution:
Change TTSSPD to have a nonskip error return and a skip return for
success. Change its only caller in FILMSC to accomodate this. NOTE that
this is a calling sequence change, and the only use of the error return is
for ACJ rejection.
TOPS20 Change Order Number 4.2559
Written by: R.ACE 30-Oct-79 08:26:48
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TAPE
Related TCO's:
Related SPR's:
Problem: DEFERRED VOLUME-SWITCH MTOPR FUNCTION (.MOSDS) IS REJECTED
FOR ANSI LABELED TAPES.
Diagnosis: U-FORMAT FOR ANSI LABELED TAPES WAS ADDED AS AN AFTERTHOUGHT.
THE MTSDVS ROUTINE WASN'T UPDATED TO PERMIT ANSI LABELED TAPES; IT
PERMITS ONLY TOPS-20 TAPES.
Solution: FIX MTSDVS ROUTINE TO PERMIT THE .MOSDS FUNCTION FOR BOTH
ANSI AND TOPS-20 LABEL TYPES.
TOPS20 Change Order Number 4.2561
Written by: DBELL 1-Nov-79 16:29:22
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
UFPGS supposedly takes a count of pages to update in AC2. But
the actual JSYS code only looks at the right half of the ac.
And to make sure a file is totally updated, you want to put <1,,0>
in AC2, which will lose!! This affected PA1050 in release 3A.
Diagnosis:
Code does HRRZ when it should use the whole AC.
Solution:
Use the whole AC (but reduce range to a reasonable value if necessary).
TOPS20 Change Order Number 4.2562
Written by: MILLER 5-Nov-79 12:44:39
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PAGEM
Related TCO's:
Related SPR's:
Problem: Monitor crashes with PTNIC1 BUGHLTs
Diagnosis: As pointed out by Dave King of CMU, there is a
problem in MVPT with the assumption about which page tables
are in memory. His analysis clearly indicated a case
where the destination page table may not be in memory even
though the code at MVLK assume it is. For those
of you interested, the description is in QAR # 20-00813.
Solution: AS Dave suggested, take some explicit action
to insure the destination page table is in memory.
The fix is to try and fault it into memory at the same
time the source PT is faulted in, but it is also
faulted in at the beginning of MVLK.
TOPS20 Change Order Number 4.2563
Written by: R.ACE 5-Nov-79 14:49:58
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: PARAMS
Related TCO's:
Related SPR's:
Problem: CFORK CALLS ACJ IF A JOB IS TRYING TO CREATE MORE FORKS THAN
A SPECIFIED NUMBER. THIS NUMBER IS THE VALUE OF THE SYMBOL DGOFKN.
THE PROBLEM: THE CUSTOMER DOESN'T KNOW IT HAS CONTROL OVER THIS VALUE.
Diagnosis: NO ONE PUT THE SYMBOL IN PARAMS.MAC.
Solution: ADD DGOFKN TO PARAMS.MAC AND BUILD.MEM. ALSO, CHANGE THE
MONITOR CALLS REFERENCE MANUAL TO DOCUMENT THAT THIS IS AN INSTALLATION-
SETTABLE PARAMETER (DOCUMENTED UNDER GETOK% JSYS, .GOCFK FUNCTION).
TOPS20 Change Order Number 4.2564
Written by: DBELL 6-Nov-79 10:01:53
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: IO
Related TCO's:
Related SPR's:
Problem:
CHKJFN returns the wrong error code in some cases, confusing its
callers.
Diagnosis:
The CHKJFS routine checks the JFN against MJFN, instead of RJFN.
MJFN is the number of JFNs possible, but the code wants RJFN,
which is the number of JFNs multiplied by the length of each JFN.
So the comparison made is wrong.
Solution:
Change the MJFN to RJFN, and fix the CAILE to be a CAIL.
TOPS20 Change Order Number 4.2565
Written by: OSMAN 7-Nov-79 09:27:12
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: LPKSDV
Related TCO's:
Related SPR's:
Problem: When KS comes up, garbage gets printed in lineprinter.
Diagnosis: Printer characters and RAM/VFU data both get put in the same
buffer (as determined by GETLPA). So on startup, while printer is printing
formfeed, buffer gets clobbered with ram/vfu data, which printer tries to
display as printing characters!
This would not be a problem if the SNDEOF routine properly waited for the
printer to stop, via some sort of scheduler test, such as LPTDIS. Perhaps
someone will spiff up SNDEOF in the future.
Solution: For now, move the LPWGO check in FESDTA to a slightly earlier
point, such that ram/vfu loading just aborts if the lineprinter is still
printing when the ram/vfu data is about to be put in the buffer.
TOPS20 Change Order Number 4.2571
Written by: DBELL 14-Nov-79 16:25:37
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
When retrieving files which had been archived, they come back invisible.
Diagnosis:
The ARCF function to restore a file doesn't bother to clear the
invisibility flag.
Solution:
Do it. Around ARCFXX-10, clear FB%INV.
TOPS20 Change Order Number 4.2572
Written by: DBELL 15-Nov-79 15:05:01
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: PHYP4
Related TCO's:
Related SPR's:
Problem:
Port contention between the -20 and the front end works badly. When
^\ is typed to invoke the front end parser, the -20 locks up the disk
for a second at a time, and it takes a very long time for the -11 to
succeed in reading in the parser. (The front end also was crashing
lots in the process).
Diagnosis:
There is some code in PHYP4 to supposedly handle port contention by
setting the drive offline(!!), and then when an interrupt comes back
clearing it again. But in the meantime, the monitor has set operator
intervention requested for the device, and so no I/O gets started
until the once-a-second check comes around. So the drive stays locked
up by the -20, which requested it and then ignored it!
Solution:
The REAL fix is to define a new UDB bit to say the drive is in use by
another port, and this will go into 4.1. But for release 4, just
add US.OIR and US.OMS to the flags diddled at RP4CN2 and RP4ATU.
TOPS20 Change Order Number 4.2574
Written by: DBELL 20-Nov-79 11:12:45
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: APRSRV
Related TCO's:
Related SPR's:
Problem:
Due to possible hardware features, it is possible to cause a PI5ERR
BUGCHK which means nothing.
Diagnosis:
If some device raises attention, and then clears it again quickly before
the CPU can respond to it, then an unvectored interrupt occurs instead
of the normal vectored interrupt. This causes the BUGCHK since the system
does not expect such an interrupt. The device which seems to cause this
problem mostly is the DX20.
Solution:
Put in a counter of times that the interrupt occurs, and only give
the bugchk if this counter reaches a threshold. So if any "real"
hardware problems are causing interrupts, they will still be noticed.
TOPS20 Change Order Number 4.2575
Written by: GRANT 20-Nov-79 11:35:32
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: 2020 loopback test did not work. The messages were not going
out over the line.
Diagnosis: The message sending parts of NSP were not paying attention
to the loopback word.
Solution: All local messages are now sent out over the looped line.
TOPS20 Change Order Number 4.2576
Written by: SANICHARA 21-Nov-79 10:16:25
Edit checked: NO Document: NO
TCO Tested: YES Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: DISC
Related TCO's:
Related SPR's:
Problem: Incorrect error message after call to CHKOFN
Diagnosis: If the error return is taken after the call to CHKOFN, the code
in DISC goes to a common location and checks for either of 2 error codes.
If neither is found, an error number with no corresponding error string
is returned resulting in an incorrect/meaningless error message.
Solution: Insert the correct error code after the call to CHKOFN
TOPS20 Change Order Number 4.2579
Written by: R.ACE 26-Nov-79 21:51:31
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: BUGS
Related TCO's:
Related SPR's:
Problem: PROBLEMS WITH ILLUUO BUGCHK:
1) THE MONITOR DOESN'T RECORD SUFFICIENT INFORMATION TO ENABLE YOU
TO DEBUG THE PROBLEM BY LOOKING AT THE SYSERR OUTPUT.
2) THE ACTION THAT IS TAKEN BY THE MONITOR IN THE EVENT OF AN ILLUUO
CAN CONCEIVABLY CAUSE LOSSAGE (E.G. BUGHLT) FURTHER DOWN THE LINE,
AND THE PERSON DEBUGGING THE SECOND FAILURE MIGHT NOT EVER KNOW
ABOUT THE ILLUUO THAT CAUSED ALL OF THE HEADACHES.
Diagnosis: ILLUUO IS A BUGCHK. BOTH OF THE ABOVE PROBLEMS CAN BE
SOLVED BY MAKING ILLUUO A BUGHLT.
Solution: DON'T BE A TOASTER, MAKE ILLUUO A BUGHLT.
TOPS20 Change Order Number 4.2582
Written by: MILLER 28-Nov-79 10:45:28
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: FORK STG
Related TCO's:
Related SPR's:
Problem: Fork lock timeout is not correct. In particular,
a job with many forks will timeout the fork lock much
faster than a job with a small number of forks.
Diagnosis: The timeout mecahnism is based on the number of times
some process has failed to acquire the lock. The more forks
that compete, the faster the timeout will occur
Solution: Make the timeout based on time.
TOPS20 Change Order Number 4.2583
Written by: GRANT 29-Nov-79 08:50:17
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: A GETOK% that fails does not return an updated error status block.
Diagnosis: The building of the GETOK data strucutre is not done correctly.
Therefore, the GETOK return does not find the user's error block
address.
Solution: Change GOKHDL from ^D16 to ^D15 and make ALOCRS put the correct
value in GOKUSA.
TOPS20 Change Order Number 4.2584
Written by: GRANT 29-Nov-79 08:56:00
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA
Related TCO's:
Related SPR's:
Problem: A user-defined function code to GETOK does not get passed to
RCVOK and, therefore, not returned in the GIVOK.
Diagnosis: GETOK2 does not retrieve the user's arguments correctly.
Solution: Change the DMOVE at GETOK2. T1 no longer contains the user's
argument.
TOPS20 Change Order Number 4.2585
Written by: ZIMA 30-Nov-79 10:20:10
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: bugs NETWRK
Related TCO's:
Related SPR's:
Problem: ^ESET ARPANET OFF causes NETRBG BUGHLTs.
Diagnosis: None yet. Have not received dumps to examine. Does not occur on
the 2020T. However, this particular error should probably have been a BUGCHK
instead of a BUGHLT, ala RELRNG.
Solution: None yet. However, turn the BUGHLT into a BUGCHK by modifying the
code and return to prevent the crashes, and solve the problem of turning the
ARPANET off causing the BUGCHKs when more information is available.
TOPS20 Change Order Number 4.2586
Written by: MILLER 30-Nov-79 10:31:45
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MAGTAP
Related TCO's:
Related SPR's:
Problem: Can't read tapes with only EOF1 labels in
trailer label set.
Diagnosis: TAPE and MAGTAP believe different things about the
state of the IORBs in certain instances. In particular,
TAPE flushes all IORBs before label processing, but
MAGTAP believes one is still active.
Solution: Clear all buffer controls when IORBs are flushed
by TAPE.
TOPS20 Change Order Number 4.2588
Written by: GRANT 3-Dec-79 14:57:15
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: Can't establish a DECnet logical link with a Phase III RSX-11M
system.
Diagnosis: TOPS-20 rejects the CC that is sent by the Phase III system.
Solution: Fix CCMSG: and DCMSG: to call LLLKUP instead of LLLKUH. When
a Phase III node is connected to a Phase II node, no route headersders
are put on the messages going to the Phase II node. TOPS-20 was
assuming that CC's and DC's always had route headers.
TOPS20 Change Order Number 4.2591
Written by: R.ACE 7-Dec-79 08:59:34
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: MEXEC
Related TCO's:
Related SPR's:
Problem: JOB CAN'T LOG OUT AFTER DOING SSAVE AND GETTING QUOTA EXCEEDED
Diagnosis: SAVER FORK (I.E., THE ONE DOING THE SSAVE JSYS) MAPS A
PAGE OF THE SAVEE'S FORK INTO JSB FREE SPACE USING SETMPG. WHEN THE
SOUT THAT WRITES THE .EXEC FILE FAILS, THE MAPPING REMAINS IN EFFECT,
AND THE SHARE COUNT OF THE SAVEE'S SECTION-0 PAGE TABLE REMAINS 1
GREATER THAN WHAT IT SHOULD BE. WHEN THE SAVEE IS KILLED, IT HANGS
FOREVER IN KSELF, WAITING FOR THE SHARE COUNT TO GO TO 1. WITH THE
FORK HUNG, THE JOB CAN'T LOG OUT.
Solution: PUT ERJMP AFTER SOUT IN LITERAL AFTER SSAVB4 IN MEXEC.
THE ERJMP HANDLER WILL UNMAP THE WINDOW PAGE.
****** NOTE: THIS IS NOT A COMPLETE SOLUTION ******
THERE IS NO DEFENSE AGAINST A ^C DURING THE SOUT (OR ANY OTHER TIME
THAT THE MAPPING IS IN EFFECT). MAYBE WE HAVE TO ADD SOME NOINT'S.
TOPS20 Change Order Number 4.2592
Written by: KONEN 7-Dec-79 17:00:14
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: TTYSRV
Related TCO's:
Related SPR's:
Problem: ILLUUO
Diagnosis: Stack is one location off for RET.
Solution: Change TTCON1 to call routine to create job, instead of JRST'ing
to it, since it now RETSKP's.
TOPS20 Change Order Number 4.2596
Written by: DBELL 12-Jan-80 15:53:28
Edit checked: NO Document: YES
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: DSKALC
Related TCO's:
Related SPR's:
Problem:
A program written for release 4.1 which uses the DSKOP JSYS might trash
disks if it is run under release 4.0.
Diagnosis:
The DSKOP JSYS had to be extended in release 4.1 to allow the specification
of a controller number for the .DOPPU argument format. This was done by
defining a new bit which if set, indicates that the channel and unit numbers
are in AC4 along with the controller number, instead of being in AC1. But
the release 4.0 (and previous monitors) do not care if that bit is set.
Thus if a release 4.1 program attempts to use the new format for the DSKOP
JSYS, and that program is ever run under release 4.0, the monitor will use
the old format and therefore write to the wrong disk unit.
Solution:
Have release 4.0 return the error DECRSV (DEC reserved bits not zero) if
any undefined flags are used in AC2. Thus a release 4.1 program using
the new format DSKOP will get an error under the old release.
TOPS20 Change Order Number 4.2597
Written by: OSMAN 18-Jan-80 09:36:22
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: YES
Program: MONITOR
Routines Affected: CDRSRV cdksdv
Related TCO's:
Related SPR's:
Problem: On a 2020 monitor without a cardreader, but with CDRN set to non-0,
the system crashes within a few seconds of typing "System in operation"
with a CDILVT bughlt.
Diagnosis: Field CDRTYP has a 0 for KL and a 1 for KS. If CDRN is non-0,
the monitor allows CDRIVE to open the cardreader and do an MTOPR that reads
the status. This MTOPR goes to CDRSTA which dispatches through one of two
vectors, CDVTFE or CDVTKS depending on whether CDRTYP is 0 or 1. If CDRTYP
is wrong, the wrong vector gets dispatched through. The code arranges for
all dispatch addresses in the wrong vector to transfer to CDBHLT which
immediately bughlts with CDILVT.
So how did CDRTYP become wrong?
I thought you'd ask that! Well, KSINI is called to initialize the KS
cardreader database when the system first starts. KSINI calls UBGOOD to
decide if the physical cardreader is really there, and immediately returns
if UBGOOD says it's not. Unfortunately, this immediate return doesn't let
the subsequent KSINI code put a 1 in CDRTYP. So CDRTYP is left wrong!
Solution: Move the setting of CDRTYP to the beginning of KSINI before UBGOOD
is called. Also, don't check PROLOG in CDRINI, since that could also circumvent
the CDRTYP initialization.
TOPS20 Change Order Number 4.2598
Written by: GRANT 24-Jan-80 09:20:57
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSA FORK
Related TCO's:
Related SPR's:
Problem: It was occassionally not possible, even after a reset, to restart
NETCON because the topology server link could not be opened.
Diagnosis: NETCON's JSB was full. The JSB space used by the PRARG JSYS was
not being released upon fork termination.
Solution: Add the code to KSELF which will release JSB free space used by
PRARG.
NOTE: This may not totally cure the observed problem. PRARG was
certainly a contributor to the free space problem but it may not
be the only one.
TOPS20 Change Order Number 4.2599
Written by: GRANT 24-Jan-80 15:51:20
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: NSPSRV
Related TCO's:
Related SPR's:
Problem: DECnet loopback test on 2020 continues to run without error
even when the connector is removed during the test.
Diagnosis: When the driver informs NSP that the line has died, NSP clears
the loopback-running flag. Local messages are then sent
internally rather than to the driver and no errors are seen.
Solution: The DECMCB routine in NSPSRV should not clear the loopback-
running flag. If a line goes down messages will still be
sent to line and the test will hang - an acceptable condition
given that there are no timeout meachanisms.
TOPS20 Change Order Number 4.2603
Written by: MILLER 19-Feb-80 10:30:11
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: LOOKUP
Related TCO's:
Related SPR's:
Problem: BUGHLT ILSTP3 occurs when it shouldn't.
Diagnosis: This BUGHLT is from the TENEX era and really
shouldn't be in TOPS-20. The condition that was at one time
impossible on TENEX is now possible under TOPS-20.
Solution: Remove the BUGHLT.
TOPS20 Change Order Number 4.2604
Written by: DBELL 19-Feb-80 13:41:20
Edit checked: NO Document: NO
TCO Tested: NO Maintenance Release: NO
Hardware-related: NO
Program: MONITOR
Routines Affected: JSYSF
Related TCO's:
Related SPR's:
Problem:
Hung jobs which cannot be logged out when running several programs
simultaneously in the job. (This occurs mostly at sites which run
a multi-forking EXEC).
Diagnosis:
The monitor has races in PMAP code which leave the JFN share count in
the FILLFW word incorrect. When trying to kill a job, a fork will
not kill itself because it is waiting for the share count to go to 1.
The races occur when several programs each are running at once in the
job.
Solution:
For release 4.1, fix PMAP so that the race doesn't exist. This is a
huge change. So for release 4, change the code near CLZFM1-8 in JSYSF
to let the job log out anyway in the above situation. The consequences
of this should not be too serious.