Trailing-Edge
-
PDP-10 Archives
-
bb-bt99e-bb
-
dbmv5a.d05
There are 2 other files named dbmv5a.d05 in the archive. Click here to see a list.
EDIT DESCRIPTIONS FOR DBMS-10-V5A
EDIT 521 FOR DBMS
;SYMPTOM:
;
; WHEN A MEMBER RECORD IN A SET OCCURRENCE HAS BINARY ZEROES IN
; ITS SET POINTERS, FIND NEXT AND FIND PRIOR GET INTO AN INFINITE
; LOOP.
;
;DIAGNOSIS:
;
; FIND RSE 3 IS NOT EXPECTING CORRUPT DBKEYS IN THE SET POINTERS,
; SO WHEN IT SEES ZEROES IN THE PARTICULAR SET POINTERS, IT THINKS
; THAT SOME OTHER CONCURRENT RUN-UNIT IS WORKING WITH THIS RECORD.
; IN ORDER TO GUARD AGAINST THIS POSSIBILITY, IT DEFINES A
; CANONICAL NEXT OF SET, WHICH IS TYPICALLY THE RECORD WHICH IT IS
; CURRENTLY WORKING WITH. THUS, IN THE CORRUPT SITUATION, IT
; RE-FINDS THAT RECORD AND LOOPS INFINITELY. WHEN DBCS HAS FOUND
; A RECORD, IT DOES NOT SPOT-CHECK THE SET POINTERS WHICH IT
; CONTAINS FOR BINARY ZEROES.
;
; HOWEVER, IF CORRUPT SET POINTERS ARE NON-ZERO, THIS FACT WILL
; BE CAUGHT UNDER THE XX22 ERROR-STATUS CODE BECAUSE THIS TYPE
; OF CURRUPT SET POINTER WILL POINT OFF TO AN AREA OR DATA BASE
; PAGE OR RECORD WHICH IS INCONSISTENT WITH THE WITHIN DEFINITION
; FOR THE RECORD TYPE IN THE SCHEMA, OR WILL POINT OFF TO SOME
; DATA BASE PAGE WHICH IS NOT DEFINED FOR AN AREA.
;
; ALSO, IF AN OWNER RECORD CONTAINS BINARY ZEROES IN ITS SET
; POINTERS THIS IS ALSO ALREADY REPORTED UNDER ERROR-STATUS XX22.
;
;CURE:
;
; INSTALL EDIT 521 FOR DBMS-10/20 V5A, WHICH WILL SPOT-CHECK
; SET POINTERS FOR THE PARTICULAR SET IN THE RECORD FOR BINARY
; ZEROES. IF IT FINDS SUCH, IT WILL ISSUE AN ERROR-STATUS CODE
; OF 0356.
;
********************************************************************************
EDIT 524 FOR DBMS
;SYMPTOM
;
; (411)
; USING SCHEMAS CONTAINING DDL USING INTERSPERSED "AREA NAME",
; "RECORD NAME" AND "SET NAME" SYNTAX, DBINFO MAY FAIL WITH
; THE MESSAGE "?INFSAF -- SCHEMA ACCESS FAILURE".
;
; (451)
; WITH EDITS 411 AND 412 INSTALLED, THE DBINFO FUNCTIONS
; /DISPLAY DATA:"SET-NAME" AND /DISPLAY USAGE:"SET-NAME" MAY
; REPORT NO DATA WHEN APPROPRIATE DATA EXISTS ON THE DBMS
; DATA BASE.
;
;
;DIAGNOSIS
;
; (411)
; DBINFO DOES NOT BIND SUCH SCHEMAS PROPERLY BECAUSE THE
; INTERNAL NAME-ID'S FOR THE USER-DEFINED SYMBOLS IN THE
; SCHEMA GET OUT OF SEQUENCE WITH THE TID'S ASSIGNED FOR
; USER'S RECORD TYPES. THIS IS REALLY A SCHEMA PROBLEM, AS
; THE SCHEMA PROCESSOR MAY INADVERTENTLY OVERLAP AN IN-CORE
; WORK AREA FOR PROCESSING RECORD TYPES AND SET TYPES. HOW-
; EVER, COBOL PROGRAMS USING SUCH SCHEMAS BIND AND EXECUTE
; PROPERLY. COBOL DOES THE BIND IN TWO STAGES. FIRST, AT
; COMPILE TIME IT FOLLOWS THE LINKED LIST OF RECORD-TYPES
; IN THE .SCH FILE WHEN IT IS CREATING THE BIND STATEMENTS.
; AS A RESULT, THE NAME-ID'S FOR THE VARIOUS DATA-NAMES ARE
; IN ASCENDING ORDER ALTHOUGH THE BIND STATEMENTS MAY NOT
; BE IN ORDER AS DEFINED IN THE SCHEMA. AT RUN-TIME DBCS
; EXPECTS THE NAME-ID'S TO BE IN ORDER, AND USES THE TID'S
; TO CREATE A TABLE OF REFERENCES TO THE RECORD TYPES IN
; THE IN-CORE DATA BASE. ON THE OTHER HAND, DBINFO TRIES THE
; SAME THING IN ONE PASS, BUT IT IS NOT SMART ENOUGH TO
; SHUFFLE THE RECORD TYPES IN THE SAME WAY. AS A RESULT,
; THE NAME-ID'S GET OUT OF SEQUENCE AS DBINFO ATTEMPTS TO
; BIND THE RECORD TYPES IN SUCCESSION.
;
; IT WAS DECIDED NOT TO FIX THE PROBLEM IN SCHEMA BECAUSE
; OF THE FACT THAT IT CAUSES NO PROBLEMS FOR COBOL (OR FOR
; FORTRAN). AS A RESULT, DBINFO IS BEING MODIFIED IN
; ORDER TO CONTAIN THE PROBLEM AT THE POINT WHERE AN EDIT
; WILL HAVE THE FEWEST POTENTIAL REPERCUSSIONS.
;
; (451)
; AFTER EDIT 411 IS THROUGH SHUFFLING THE RECORD LIST FOR THE
; SUB-SCHEMA BINDING PROCESS IN DBINFO, IT FAILS TO CLEAR THE
; LEFT HALF OF THE WORD FOR THE AFFECTED MEMBERS OF THIS LIST.
; THESE TWO DBINFO FUNCTIONS USE THE ENTRIES OF THIS LIST IN
; ORDER TO SEE IF THE CURRENT DATA BASE RECORD IS AN OWNER OF
; THE SPECIFIED SET TYPE. IF THE LEFT HALF OF THE ENTRY IN
; THE RECORD LIST IS NOT CLEARED, THE COMPARISON WILL FAIL
; WHEN IT SHOULD PASS.
;
;CURE
;
; INSTALL EDIT 524, CAUSING DBINFO TO RETAIN INFORMATION
; IN ITS IN-CORE TABLE OF REFERENCES AS TO WHAT ORDER THE
; RECORD TYPES SHOULD HAVE BEEN IN. THE ORIGINAL ENTRIES
; ARE A SERIES OF HALF-WORD ADDRESSES IN THE RIGHT HALF OF
; EACH WORD, AND THE LEFT HALF OF EACH REMAINS AT 0 FOR
; THE ENTIRE RUN OF DBINFO. IT IS THE LEFT HALF OF THESE
; WORDS WHICH ARE BEING USED FOR THE PURPOSES OF THIS EDIT.
;
; (451)
; ALSO, WHEN THE WORK OF EDIT 411 IS COMPLETE, CLEAR THOSE
; LEFT HALF-WORDS.
;
;
;WARNINGS
;
; WHEN THIS EDIT IS INSTALLED INTO DBINFO, ITS "DISPLAY
; DATA" FUNCTION STILL WILL NOT KNOW WHERE TO LOOK IN THE
; IN-CORE DATA BASE FOR THE INFORMATION ON RECORD TYPES
; WHICH ARE OUT OF ORDER, AND IT WILL PRODUCE THE MESSAGE
; "?INFDAF -- DATA BASE ACCESS FAILURE". IN ORDER TO CLEAR
; THIS PROBLEM, YOU MUST INSTALL EDIT 525.
;
; EDITS 524 AND 525 ARE INTENDED TO FIX A VERY SPECIALIZED
; PROBLEM AS EXPLAINED ABOVE, AND YOU NEED NOT INSTALL THEM
; UNLESS YOU CONSTRUCT YOUR SCHEMA DDL IN THE UNUSUAL FASH-
; ION AS INDICATED UNDER "SYMPTOM" ABOVE AND YOU ARE EXPER-
; IENCING THIS PROBLEM.
;
; ALSO, WHEN THE "DISPLAY MAP" FUNCTION IS USED, DBINFO WILL
; PUT THE LISTINGS FOR THE RECORD TYPES WHICH ARE OUT OF
; ORDER AT THE BOTTOM OF THE "MAP" REPORT. THE CONTENTS OF
; THESE LISTINGS WILL BE ACCURATE. IT WAS DECIDED NOT TO
; FIX THIS PROBLEM AT THIS TIME, AS IT WOULD HAVE REQUIRED
; A MAJOR RE-WRITE OF THE "MAP" FUNCTION IN ORDER TO OVER-
; COME WHAT AMOUNTS TO A MINOR INCONVENIENCE FOR THE USER.
;
; FINALLY, DO NOT INSTALL THIS EDIT UNTIL AFTER YOU HAVE INSTALLED
; EDIT 522 FOR DBMS-20 OR EDIT 523 FOR DBMS-10.
;
********************************************************************************
EDIT 525 FOR DBMS
;SYMPTOM
;
; WITH EDIT 524 INSTALLED INTO DBINFO AND USING SCHEMAS
; CONTAINING DDL USING INTERSPERSED "AREA NAME", "RECORD
; NAME" AND "SET NAME" SYNTAX, DBINFO MAY FAIL WITH THE
; MESSAGE "?INFDAF -- DATA BASE ACCESS FAILURE".
;
;DIAGNOSIS
;
; SEE FIRST PARAGRAPH OF THE WARNINGS IN THE WRITE-UP FOR
; EDIT 524.
;
;CURE
;
;
; INSTALL EDIT 525, CAUSING THE DBINFO "MAP" FUNCTION TO
; USE THE ADDRESSING INFORMATION WHICH WAS PUT INTO THE
; IN-CORE RECORD REFERENCE TABLE BY EDIT 524.
;
;WARNING
;
; THIS EDIT MUST NOT BE INSTALLED INTO DBINFO UNLESS YOU
; ARE ALSO INSTALLING EDIT 524.
********************************************************************************
EDIT 527 FOR DBMS
;SYMPTOM
;
; DBINFO DOES NOT RECOGNIZE A DISPLAY REQUEST WHOSE PAGE
; RANGE LIES OUTSIDE THE AREA SPECIFIED. WHEN THIS OCCURS
; DBINFO RETURNS THE MESSAGE "?INFDAF DATA BASE ACCESS
; FAILURE -- IS DBMS PROBLEM IF IT RECURS."
;
;DIAGNOSIS
;
; INAPPROPRIATE DBINFO ERROR MESSAGE. (UNFORTUNATELY NO
; NEW ERROR MESSAGES CAN BE INTERPOLATED INTO THE DBMS V5
; .REL LIBRARIES.)
;
;CURE
;
;
; SELECT AN ERROR MESSAGE WHICH WILL CAUSE THE USER TO
; LOOK IN THE PROPER PLACE FOR REFERENCE MATERIAL, I.E.
; IN HIS SCHEMA FILE. INSTALL EDIT 527 TO GIVE DBINFO'S
; ?INFSAF MESSAGE IN THIS CASE.
********************************************************************************
EDIT 530 FOR DBMS
;SYMPTOM:
;
; DBINFO DOES NOT UNDERSTAND /open all AND /close all, BUT IT WILL
; UNDERSTAND /open ALL AND /close ALL.
;
;DIAGNOSIS:
;
; FOR THE MODIFIERS AFTER THESE VERBS, DBINFO WASN'T CONVERTING
; LOWER CASE TO UPPER CASE PROPERLY CAUSING COMPARISIONS TO FAIL.
;
;CURE:
;
; APPLY EDIT 530 TO DBINFO. EDIT 530 WILL CAUSE LOWER CASE TO BE
; COMPARED PROPERLY TO THE CORRESPONDING UPPER CASE ALPHABETICS
; FOR THESE COMMANDS.
;
********************************************************************************
EDIT 531 FOR DBMS
;SYMPTOM:
;
; DBINFO-10 DOES NOT UNDERSTAND /schema schnam , BUT IT WILL
; UNDERSTAND /schema SCHNAM.
;
;DIAGNOSIS:
;
; FOR THE MODIFIER AFTER THE VERB, DBINFO-10 WASN'T CONVERTING
; LOWER CASE TO UPPER CASE PROPERLY CAUSING COMPARISIONS TO FAIL.
;
;CURE:
;
; APPLY EDIT 531 TO DBINFO-10. EDIT 531 WILL CAUSE LOWER CASE TO
; BE COMPARED PROPERLY TO THE CORRESPONDING UPPER CASE ALPHABETICS
; FOR THIS COMMAND. NOTE, THE SCHEMA NAME WILL BE OUTPUT IN UPPER
; CASE ON YOUR DBINFO LISTINGS.
;
********************************************************************************
EDIT 532 FOR DBMS
;SYMPTOM:
;
; DBMEND APPEARS TO HAVE DIFFICULTY RECOGNIZING THE BEGINNING
; OF A DISK JOURNAL WHEN THE COMMAND /POSITION 1 IS GIVEN.
;
;DIAGNOSIS:
;
; PAGE 1 OF A DISK JOURNAL IS THE JOURNAL LABEL PAGE AND
; ALSO CONTAINS THE TERMINAL FRAGMENT OF THE LAST JOURNAL
; PAGE AND THEREFORE SHOULD NOT BE REFERENCED BY THE /POS
; COMMAND WITH AN ABSOLUTE ARGUMENT.
;
;CURE:
;
; INSTALL EDIT 532 TO GIVE AN ERROR MESSAGE WHEN THE COMMAND
; /POS 1 HAS BEEN GIVEN FOR A DISK JOURNAL. THE ERROR
; MESSAGE WHICH WE HAVE SELECTED IS "?MNDJPI JOURNAL
; POSITIONED INCORRECTLY AFTER PROCESSING LAST-RELATIVE
; BOUNDARY." THIS MESSAGE IS NOT EXACTLY APPROPRIATE TO
; THE PROBLEM, BUT IT IS NOT POSSIBLE TO INSTALL NEW ERROR
; MESSAGES INTO THE DBMEND REL FILES.
;
********************************************************************************
EDIT 533 FOR DBMS
;SYMPTOM:
;
; DBMEND WILL PRODUCE THE ERROR MESSAGE "?I/O TO UNASSIGNED
; CHANNEL AT USER PC NNNNNN" WHEN THE /UNLOAD COMMAND IS
; GIVEN WITHOUT PREVIOUSLY SPECIFYING A JOURNAL.
;
;DIAGNOSIS:
;
; DBMEND'S /UNLOAD COMMAND BLYTHELY ASSUMED THAT A JOURNAL
; HAD ALREADY BEEN SPECIFIED.
;
;CURE:
;
; INSTALL DBMS EDIT 533 TO MAKE IT CHECK FOR A JOURNAL. IF
; A JOURNAL HAS NOT BEEN PREVIOUSLY SPECIFIED, /UNLOAD WILL
; ISSUE THE MESSAGE "?MNDJNI JOURNAL NOT YET IDENTIFIED".
;
********************************************************************************
EDIT 534 FOR DBMS
;SYMPTOM:
;
; AFTER DOING A /MERGE OPERATION IN DBMEND-10 TO REMOVE AN
; INCOMPLETE TRANSACTION THEN UNLOADING THE JOURNAL WITH THE
; /UNLOAD COMMAND, ANY SUBSEQUENT RUN-UNIT UNDER DBCS TRYING
; TO RE-USE THE JOURNAL WOULD SOMETIMES FAIL ON AN OPEN WITH
; AN ERROR-STATUS CODE OF 0962.
;
;DIAGNOSIS:
;
; THE 0962 ERROR-STATUS CODE WAS REPORTED IF DBMS-10'S
; JOURNAL-IN-UNDEFINED-STATUS WORD WAS SET IN THE JOURNAL
; LABEL PAGE OF A DISK JOURNAL. THIS WORD WAS SET BY /MERGE
; IN ORDER TO SIGNIFY THAT IT WAS CURRENTLY WORKING WITH THE
; JOURNAL IN ORDER TO RECOVER THE DATA BASE. IT SHOULD BE
; CLEARED BY /UNLOAD ON THE COMPLETION OF A SUCCESSFUL MERGE
; IN ORDER TO INDICATE THAT THIS JOURNAL IS NO LONGER ACTIVELY
; INVOLVED IN RESTORING THE DATA BASE AND IS NOW AVAILABLE
; FOR FURTHER WORK UNDER RUN-UNITS USING DBCS.
; THIS SYMPTOM IS ESPECIALLY TROUBLESOME IN CASES WHERE
; SIMULTANEOUS USERS WISH TO SHARE A PRE-ALLOCATED AND
; TRUNCATED JOURNAL AFTER IT HAS BEEN USED BY DBMEND-10 IN
; ORDER TO BACK OUT A PREVIOUS INCOMPLETE TRANSACTION.
; DBMEND-20 HANDLES THIS PROCEDURE DIFFERENTLY.
;
;CURE:
;
; INSTALL EDIT 534 TO CAUSE DBMEND-10 /UNLOAD TO CLEAR THE
; JOURNAL-IN-UNDEFINED-STATE WORD IN THE EVENT OF A SUCCESS-
; FUL /MERGE. THIS EDIT USES BIT 0 IN THE RIGHT HALF OF
; DBMEND'S STATUS REGISTER IN ORDER TO CONVEY INFORMATION
; ABOUT THE JOURNAL STATE FROM /MERGE TO /UNLOAD. IF THIS
; BIT IS SET COMING INTO THE THE EXECUTION OF /UNLOAD, IT
; WILL CLEAR IT AND GIVE THE MESSAGE "?MNDPIF JOURNAL'S
; LABEL PAGE HAS IMPROPER FORMAT". THIS BIT IS NOT OTHER-
; WISE USED BY DBMEND-10 AND IS NOT DEFINED TO IT. A SUB-
; SEQUENT /UNLOAD WILL UNLOAD THE DISK JOURNAL, CLEARING
; THE JOURNAL-IN-UNDEFINED-STATE WORD. IN THIS WAY THE USER
; WILL BE RELIEVED OF THE DISTRESSING PROBLEM OF HAVING HIS
; JOURNAL LABEL IN AN UNDEFINED STATE WHEN HE EXITS FROM
; DBMEND-10.
;
********************************************************************************
EDIT 535 FOR DBMS
;SYMPTOM:
;
; IN CASES WHERE THE /schema COMMAND IS NOT PRECEDED BY /journal,
; DBMEND-10 DOES NOT UNDERSTAND /schema schnam , BUT IT WILL
; UNDERSTAND /schema SCHNAM.
;
;DIAGNOSIS:
;
; FOR THE MODIFIER AFTER THE VERB, DBMEND WASN'T CONVERTING
; LOWER CASE TO UPPER CASE PROPERLY CAUSING COMPARISIONS TO FAIL.
;
;CURE:
;
; APPLY EDIT 535 TO DBMEND-10. EDIT 535 WILL CAUSE LOWER CASE TO
; BE COMPARED PROPERLY TO THE CORRESPONDING UPPER CASE ALPHABETICS
; FOR THIS COMMAND. NOTE, THE SCHEMA NAME WILL BE OUTPUT IN UPPER
; CASE ON YOUR DBMEND LISTINGS.
;
********************************************************************************
EDIT 536 FOR DBMS
;SYMPTOM:
;
; THE DBMEND COMMAND /JOURNAL J*.* GIVES THE ERROR MESSAGE
; "?MNDWCP WILDCARDING IS PROHIBITED", IMMEDIATELY FOLLOWED BY
; THE MESSAGE "?PA1050 ILLEGAL INSTRUCTION AT USER PC 0".
;
;DIAGNOSIS:
;
; AFTER THE ?MNDWCP MESSAGE IS DISPLAYED, THE ERROR-HANDLING
; PROCEDURE DOES A JRST 0 INSTEAD OF A JRST ABORT TO ENTER
; DBMEND'S RECYCLING PROCEDURE AFTER AN ERRONEOUS COMMAND.
;
;CURE:
;
; INSTALL EDIT 536 TO MAKE THE /JOURNAL COMMAND RECYCLE THROUGH
; THE REGULAR RECYCLE ROUTINE AFTER A ?MNDWCP ERROR.
;
********************************************************************************
EDIT 537 FOR DBMS
;SYMPTOM:
;
; IF THE DBMEND COMMAND /MERGE IS ISSUED WITHOUT A "BEFORE" OR
; "AFTER" ARGUMENT, THE RESULTING ?SCNSVR MESSAGE HAS GARBAGE
; AT THE END.
;
;DIAGNOSIS:
;
; THE /MERGE ERROR HANDLING PROCEDURE IS NOT PROVIDING A VALID
; ARGUMENT TO SCAN'S ERROR MESSAGE PROCESSOR.
;
;CURE:
;
; INSTALL EDIT 537.
;
********************************************************************************
EDIT 540 FOR DBMS
;SYMPTOM
;
; IF ORDER IS SORTED DUPLICATES ARE FIRST/LAST/UNSPECIFIED
; ALLOWED IS SPECIFIED IN THE ORDER CLAUSE OF A SET TYPE
; WHICH HAS MEMBERSHIP CONSISTING OF MULTIPLE MEMBER REC-
; ORD TYPES, ENTER MACRO FINS6 WILL NOT NECESSARILY RE-
; TRIEVE THE PROPER MEMBER RECORD TYPE WHEN A MATCH IS
; FOUND ON THE USER-SUPPLIED SORT KEY VALUE.
;
;DIAGNOSIS
;
; FOR A SET OCCURRENCE WITH ORDER SORTED DUPLICATES ...
; AS SPECIFIED ABOVE, FINS6 FAILS TO CHECK THE RECORD
; TYPE OF THE MEMBER RECORD WHEN A USER KEY MATCH IS
; FOUND BECAUSE THE INTERNAL RECORD TYPE ID IS NOT PART
; OF THE SORT KEY FOR THE SET TYPE. THIS PROBLEM APPLIES
; FOR RANGE KEY MATCHES AS WELL AS FOR EXACT KEY VALUE
; MATCHES. FINS6 WORKS APPROPRIATELY FOR ALL OTHER TYPES
; ORDER SORTED WITH EITHER EXACT OR RANGE KEY MATCHES.
;
;CURE
;
; SINCE FINS6 REQUIRES THE USER TO SUPPLY A RECORD NAME
; AS A "USING" PARAMETER AS WELL AS A SET NAME, IT AP-
; PEARS ONLY PROPER THAT IT SHOULD CHECK FOR RECORD TYPE
; AS WELL AS USER KEY VALUE. INSTALL EDIT NO. 540 IN
; ORDER TO MAKE IT DO THIS EXTRA CHECKING FOR ORDER
; SORTED DUPLICATES.
;
; THIS LARGE AND COMPLEX PATCH AFFECTS RETRIEVAL THRU
; FINS6 ON THE FOURTH OF THE FOUR MAJOR TYPES OF SORTED
; SETS, WHICH MAY NOT NECESSARILY BE USED AT ALL SITES. IT
; IS COMPLEX BECAUSE CHECKING MUST BE DONE FOR RANGE KEY
; MATCHES AS WELL AS FOR EXACT KEY MATCHES. FOR RANGE KEY
; MATCHES, WE HAVE TAKEN THE INTERPRETATION THAT A RANGE
; MATCH EXISTS ONLY IF TWO CONDITIONS EXIST. THE FIRST IS
; AS SPECIFIED IN THE DBMS VER 5 ADMINISTRATOR'S MANUALS
; FOR RANGE KEY MATCHES ON PAGE 4-21 TECHNICAL NOTES 8 AND
; 9 AND ON PAGE 4-23 TECHNICAL NOTE 9. HOWEVER, THE LIST
; OF CONDITIONS GIVEN IN NOTE 9 ON PAGE 4-23 APPLIES TO
; BOTH MODES OF SET OCCURRENCE SELECTION -- CURRENT OF SET
; AS WELL AS LOCATION MODE OF OWNER. THE OTHER FORMS OF
; ORDER SORTED WHERE DUPLICATES ARE ALLOWED CURRENTLY
; ADHERE TO THIS INTERPRETATION. THE SECOND IS THAT
; THE INTERNAL TYPE ID OF THE FOUND RECORD MUST MATCH THE
; RECORD TYPE SPECIFIED BY THE USER AS THE "RECORD NAME"
; PARAMETER SUPPLIED TO FINS6 UNDER ORDER SORTED DUPLICATES.
;
;
;
********************************************************************************
EDIT 542 FOR DBMS
;SYMPTOM
;
; WHEN USING FORDML IN /UNFLAGGED MODE, LOWER-CASE STRINGS
; ARE REPLACED BY UPPER-CASE STRINGS EVEN IN FORMAT AND DATA
; STATEMENTS.
;
;DIAGNOSIS
;
; WITH UNFLAGGED MODE, FORDML MUST CHECK EVERY LINE OF THE
; INPUT TEXT TO SEE IF IT TERMINATES WITH A PERIOD (.), AND
; IS THUS A DML STATEMENT. WHEN DOING SO, IT ALTERS
; LOWER-CASE INPUT ALPHABETICS TO UPPER-CASE BEFORE IT
; SCANS ITS TABLE OF LEXICAL ITEMS FOR A MATCH. AS A RESULT
; ALL LOWER-CASE ALPHABETICS ARE CHANGED TO UPPER-CASE
; BEFORE FORDML GOES ON TO BUILD ITS OUTPUT TEXTS.
;
;CURE
;
; INSTALL EDIT 542 TO GIVE FORDML AN EXTRA WORK AREA IN
; WHICH TO SAVE INPUT TEXT ASIDE AS IS WHEN IT IS WORKING IN
; UNFLAGGED MODE.
;
;WARNING
;
; IN UNFLAGGED MODE, WHEN A LINE OF DML IS SPLIT ACROSS TWO
; BUFFERS, PART OR ALL OF THE LINE MAY APPEAR IN LOWER-CASE
; IN THE OUTPUT TEXT. THIS IS BECAUSE A GREAT DEAL OF TEXT
; MAY BE INTERPOLATED IN ORDER TO CREATE THE EXTERNAL CALLS.
; AS A RESULT, THE INPUT AND OUTPUT BUFFERS ARE NOT IN SYNC
; AND TEXT WHICH WE MAY WISH TO CHANGE FROM LOWER-CASE TO
; UPPER-CASE IN THE OUTPUT BUFFER MAY HAVE ALREADY BEEN
; WRITTEN OUT TO DISK. SINCE THE ONLY TEXT IN THIS CATEGORY
; ERRONEOUSLY IS DML WHICH HAS BEEN COMMENTED OUT WITH AN
; ASTERISK (*) IN THE OUTPUT TEXT, WE FELT THAT THIS CONSTI-
; TUTED NO PROBLEM WHATEVER.
;
;
********************************************************************************
EDIT 543 FOR DBMS
;SYMPTOM:
;
; WITH EDIT 420 INSTALLED AN ERROR-STATUS CODE OF 961 AND
; THE MESSAGE "?DBSXWX EXCEPTION WHILE PROCESSING AN EXCEP-
; TION ARE RECEIVED WHILE OPENING WITH A PRE-ALLOCATED AND
; ZEROED JOURNAL UNDER AN UPDATING USAGE MODE IN DBMS-10.
;
;DIAGNOSIS:
;
; IF A DBMS JOURNAL FILE ALREADY EXISTS ON THE USER'S
; [P,PN], DBCS EXPECTS THAT IT ALREADY CONTAINS VALID
; JOURNAL LABEL AND LABEL PAGE HEADER INFORMATION, IN-
; CLUDING A VALID LABEL PAGE NUMBER. WITH A PRE-ALLOCATED
; AND ZEROED JOURNAL THIS IS NOT THE CASE, AS IN FACT THE
; JOURNAL LABEL PAGE NUMBER IS 0. WHEN DBCS THEN SUBTRACTS
; 1 FROM THE 0 IT ARRIVES AT A NEGATIVE JOURNAL PAGE NUMBER,
; WHICH IT THEN SENDS TO THE FOLLOWING FILOP UUO, WHICH
; DOES A USETO. EDIT 420 NEVER ADDRESSED THE PROBLEM OF
; THE JOURNAL LABEL PAGE NUMBER OF 0. HOWEVER, IT IS A
; VALID EDIT AND SHOULD BE RETAINED.
;
;CURE:
;
; INSTALL EDIT 543 TO TEST FOR A JOURNAL LABEL PAGE NUMBER
; OF 0 WHILE OPENING. IF IT IS 0, ALSO CHECK THE REST OF
; THE JOURNAL LABEL AND ITS HEADER TO SEE IF THEY CONTAIN
; ALL ZEROES. IF THESE ARE NOT ZERO THEN GIVE AN ERROR-
; STATUS CODE OF 961 ON THE OPEN. IN OTHER WORDS, YOU ARE
; NOT SUPPOSED TO START OFF WITH JUNK IN YOUR PRE-ALLOCATED
; AND ZEROED JOURNAL. THE USER SHOULD TERMINATE THE RUN-UNIT
; IN THE EVENT OF A 961 ERROR-STATUS CODE AND CHECK OUT HIS
; RUN ENVIRONMENT.
;
********************************************************************************
EDIT 545 FOR DBMS
;SYMPTOM
;
; SCHEMA LOOPS, APPENDING PAGES FULL OF BINARY ZEROES TO
; THE OUTPUT SCH FILE FOR THE FIRST DDL FILE ENTERED BY
; THE USER WHEN A 01 RECORD-NAME STATEMENT IN A SUB-SCHEMA
; IS NOT TERMINATED BY A PERIOD (.).
;
; NOTE: ORIGINAL SYMPTOM REPORT -- SCHEMA TRIES TO DO A
; USETO -3(1) WHEN AC1 CONTAINS A 0.
;
;DIAGNOSIS
;
; SCHEMA MUST CHECK EVERY SO OFTEN ON WHETHER IT MUST DO
; OUTPUT TO THE SCH FILE, SO THE SYMPTOM IS INTERMITTENT.
; HOWEVER, BEFORE THE SYMPTOM OCCURS SCHEMA HAS ALREADY
; FLAGGED THE OFFENDING SYNTAX. THE USETO WHICH GETS CON-
; FUSED IS AT PUTIO+5 IN SCHIO1, AND WE MUST BYPASS IT IF
; IT HAS PICKED UP AN INVALID REFERENCE TO AN ADDRESS
; BLOCK. HOWEVER, WHEN THIS CONDITION OCCURS IT MEANS THAT
; SCHEMA HAS ENCOUNTERED A SERIOUS INCONSISTENCY IN ITS
; PROCESSING, AND SO WE SHOULD CAUSE IT TO GIVE THE ?DDLSAF
; MESSAGE, WHICH SCHEMA FOR DBMS-20 ALREADY DOES.
;
;CURE
;
; INSTALL EDIT 545 TO BYPASS THE USETO AND FORCE THE ?DDLSAF
; MESSAGE IN THE HIGHER-LEVEL CODE.
;
********************************************************************************
EDIT 546 FOR DBMS
;SYMPTOM:
;
; (1) MAKLIB WILL NOT INSTALL EDIT 417 INTO DDLANY.REL FOR
; THE DBMS-10 SCHEMA PROCESSOR.
;
; (2) THE SUB-SCHEMA SET SECTION IS STILL NOT FULLY OPTIONAL
;
;DIAGNOSIS:
;
; (1) EXISTING CODE DOES NOT MATCH ORIGINAL CODE AT THREE
; LOCATIONS -- TCO.DA+13, TCO.TX+21, AND R01END+5
;
; (2) TYPOGRAPHICAL ERROR FOLLOWING .INSERT TCO.DA+13.
; WITHIN THE LITERAL THE TEXT "AOJA 13,SETSEC" OCCURS TWICE
; BUT IT SHOULD OCCUR ONLY ONCE.
;
;CURE:
;
; INSTALL EDIT 546 TO MAKE THE SUB-SCHEMA SET SECTION FULLY
; OPTIONAL.
;
********************************************************************************
EDIT 550 FOR DBMS
;SYMPTOM
;
; AFTER A FAILURE ON AN OPEN OF AN AREA, THE SPECIAL REGISTER
; ERROR-AREA MAY BE FILLED WITH AN INCORRECT AREA NAME.
;
;DIAGNOSIS
;
; WHEN AN OPEN VERB FAILS, DBCS IMMEDIATELY USES THE NORMAL CLOSE
; PROCEDURES IN ORDER TO UN-DO THE OPEN. OPEN HAS DEPOSITED THE
; ADDRESS OF THE OFFENDING AREA BLOCK IN LAS.AB(SS). WHEN THE
; CLOSE PROCEDURES ARE GOING THROUGH THE AREAS IN THE OPEN LIST,
; THEY ALSO DEPOSIT THE ADDRESSES OF THE AREA BLOCKS WHICH THEY SEE
; INTO LAS.AB(SS). AS A RESULT, THE ADDRESS OF THE OFFENDING AREA
; BLOCK CAN BE LOST.
;
;CURE
;
; SAVE ASIDE THE ADDRESS OF THE OFFENDING AREA BLOCK WHILE CLOSE
; PROCEDURES ARE DOING THEIR THING, THEN WHEN CLOSE IS FINISHED
; RESTORE THIS ADDRESS TO LAS.AB(SS). INSTALL EDIT 550.
;
;
********************************************************************************
EDIT 551 FOR DBMS
;SYMPTOM:
;
; WHEN THE CALC CHAIN POINTER IN A CALC RECORD POINTS TO ITSELF,
; WITH DUPLICATES NOT ALLOWED THE USER PROGRAM HANGS, BUT WITH
; WITH DUPLICATES ALLOWED THE USER PROGRAM LOOPS.
;
;DIAGNOSIS:
;
; FIND RSE 5 IS NOT EXPECTING A CALC-CHAIN POINTER TO POINT TO
; ITSELF, AND IT ACCEPTS THIS CORRUPT POINTER AS THE POINTER
; TO THE NEXT RECORD ON THE CALC CHAIN. THIS KIND OF CORRUPTION
; CAN OCCUR IF AN ERROR-STATUS CODE IS ISSUED ON A STORE OF A
; CALC RECORD AND THE FAILED STORE IS NOT PROPERLY BACKED OUT.
; FIND RSE 5 CHECKS FOR OTHER KINDS OF INCONSISTENCIES ON CALC-
; CHAIN POINTERS, HOWEVER.
;
;CURE:
;
; INSTALL EDIT 551 FOR DBMS-10/20 V5A, WHICH WILL SPOT-CHECK
; CALC-CHAIN POINTERS TO SEE IF THEY POINT BACK TO THEMSELVES.
; IF IT FINDS THIS CONDITION, IT WILL ISSUE AN ERROR-STATUS CODE
; OF 356.
;
;
********************************************************************************
EDIT 552 FOR DBMS
;SYMPTOM:
;
; AN ABSTRACT OF A JOURNAL WITH TEXT ENTRIES DOES NOT CORRECTLY PRINT
; ANY TEXT ENTRY WHICH IS LONGER THAN 80 ASCII CHARACTERS.
;
;DIAGNOSIS:
;
; THE ROUTINE WHICH PRODUCES THE MESSAGES FOR SUCCEEDING BLOCKS OF THE
; SAME TEXT ENTRY FORGETS THAT THE RUN-UNIT-ID, WHICH IS NOT PRINTED
; FOR BLOCKS OTHER THAN THE FIRST BLOCK IN THE TEXT, IS PASSED ALONG
; WITH THE TEXT, AND ATTEMTPS TO USE IT INSTEAD OF THE TEXT.
;
;CURE:
;
; INSTALL EDIT 457. THIS WILL CAUSE TEXT BLOCKS OVER 80 ASCII CHARACTERS
; TO BE PRINTED AS MULTIPLE TEXT BLOCKS AND IS A TEMPORARY SOLUTION TO
; THIS PROBLEM.
;
********************************************************************************
EDIT 553 FOR DBMS
;SYMPTOM:
;
; IF THE DBMEND COMMANDS /ABSTRACT, /BUILD, /EXCLUDE, /FORCE
; /JOURNAL, /OPEN, OR /SCHEMA ARE ISSUED WITHOUT AN
; ARGUMENT, THE RESULTING ?SCNSVR MESSAGE HAS GARBAGE
; AT THE END.
;
;DIAGNOSIS:
;
; THE COMMAND ERROR HANDLING PROCEDURES ARE NOT PROVIDING A VALID
; ARGUMENT TO SCAN'S ERROR MESSAGE PROCESSOR.
;
;CURE:
;
; INSTALL EDIT 553.
;
********************************************************************************
EDIT 555 FOR DBMS
;SYMPTOM:
;
; IN DBMEND-10,
; WHEN JOURNALLING IS DONE BE COMMAND, THE DBMEND /START AND /END
; COMMANDS DO NOT ALLOW THE USER TO EFFECTIVELY SPECIFY A PARAMETER
; FOR RUN-UNIT ID.
;
;DIAGNOSIS:
;
; FOR COMMAND JOURNALLING, THE BOUNDARY-CHECKING ROUTINE WAS BY-
; PASSING THE CHECK FOR RUN-UNIT ID. HOWEVER, IN THE CASES WHERE
; THESE COMMANDS ARE GIVEN IN THE FORM "/START LAST" OR "/END LAST"
; THE USER SHOULD NOT BE ALLOWED TO SUPPLY A RUN-UNIT ID, ACCORDING
; TO THE DBMS-10/20 ADMINISTRATORS MANUALS.
;
;CURE:
;
; INSTALL EDIT 555 IN ORDER TO DO THE APPROPRIATE CHECKING.
; DBMS-10 TRANSACTION RECOVERY IS UNAFFECTED BY THIS EDIT.
;
********************************************************************************
EDIT 556 FOR DBMS
;SYMPTOM:
;
; FORTRAN SPEUDONYMS DO NOT WORK ON ALIASES, DIRECT IDENTIFIERS,
; OR AREA-IDS.
;
;DIAGNOSIS:
;
; THE ROUTINE WHICH COPIED THE PSEUDONYM INTO THE ITEM LINE FOR
; THE SCHEMA FILE WAS BEING CALLED INCORRECTLY, CAUSING THE LENGTH OF
; THE PSEUDONYM TO BE SET TO ZERO, THE ACTUAL LENGTH TO APPEAR WHERE
; THE ASCII CHARACTERS OF THE PSEUDONYM SHOULD BE, AND THE ASCII
; CHARACTERS TO BE LOST.
;
;CURE:
;
; INSTALL EDIT 556 TO DO THE APPROPRIATE CALL TO COPSTR.
;
********************************************************************************
EDIT 560 FOR DBMS
;SYMPTOM:
;
; IF THE DBINFO COMMAND /DISPLAY IS ISSUED WITHOUT AN ARGUMENT
; DBINFO-20 FAILS WITH THE MESSAGE "?PA1050 ILLEGAL INSTRUCTION"
; AND DBINFO-10 FAILS WITH THE MESSAGE "?HALT AT USER PC ...".
;
;DIAGNOSIS:
;
; /DISPLAY IS NOT CHECKING FOR AN ARGUMENT FOLLOWING.
;
;CURE:
;
; INSTALL EDIT 560 TO MAKE /DISPLAY CHECK FOR A FOLLOWING ARGUMENT
; AND IF THE ARGUMENT IS NOT FOUND, CALL SCAN'S ERROR PROCESSING
; PROCEDURE.
;
********************************************************************************
EDIT 562 FOR DBMS
;SYMPTOM:
;
; A DML PROGRAM USING ENTER MACRO JRDATA, WHICH OMITS THE SECOND
; PARAMETER INDICATING THE NUMBER OF WORDS OF NON-PRINTING DATA
; TO BE PASSED TO THE JOURNAL, FAILS WITH A ?MEMORY PROTECTION
; VIOLATION.
;
;DIAGNOSIS:
;
; JRDATA NEVER CHECKS TO SEE IF THIS REQUIRED PARAMETER IS PRESENT
; AND MERELY ASSUMES THAT IT IS THERE.
;
;CURE:
;
; INSTALL EDIT 562 FOR DBMS-10/20 V5A TO MAKE JRDATA CHECK FOR THE
; PRESENCE OF THIS REQUIRED PARAMETER. IT IT IS NOT PRESENT, MAKE
; JRDATA GIVE AN ERROR-STATUS CODE OF 1658.
;
********************************************************************************
EDIT 564 FOR DBMS
;SYMPTOM:
;
; FORDML PROBLEMS WITH EDITS 542 AND 545 INSTALLED
;
;DIAGNOSIS:
;
; EDIT 545 IS NOT NEEDED IN DBMS-10 V.5A
;
;CURE:
;
; REMOVE EDIT 545 FROM DBMS-20 V.5A
;
********************************************************************************
EDIT 565 FOR DBMS
;SYMPTOM
;
; WITH EDIT 540 INSTALLED, FINS6 APPEARS TO LOOP INDEFINITELY IN
; CASES WHERE THERE ARE NO OWNER POINTERS IN THE MEMBER RECORDS
; OF A SORTED SET OCCURRENCE.
;
;DIAGNOSIS
;
; IN EDIT 540 THE CALL TO LIN.BF ASSUMES THAT REGISTER AP IS STILL
; SET UP FROM THE PREVIOUS CALL, BUT IN THE CASE DESCRIBED IN THE
; SYMPTOM THIS REGISTER HAS BEEN CLOBBERED.
;
;CURE
;
; INSTALL EDIT 565 TO TEACH EDIT 540 HOW TO PRESERVE THE REGISTER.
;
; NOTE: THIS EDIT REMOVES EDIT 540. IF YOU DO NOT HAVE EDIT 540
; INSTALLED, OMIT THE TWO OCCURRENCES OF THE MAKLIB INSTRUCTION
; ".REMOVE 540" WHICH ARE SHOWN IN THE TEXT OF THE EDIT BELOW.
;
********************************************************************************
EDIT 566 FOR DBMS
;SYMPTOM
;
; WHEN USING THE FIND identifier-2 FORMAT OF FIND RSE 3 AND IF
; identifier-2 IS NOT DECLARED AS A DATA-NAME WITH PICTURE S9(10)
; OR LESS AND USAGE COMP (IN OTHER WORDS ONE-WORD BINARY), THIS
; FORM OF THE RSE DOES NOT NECESSARILY RETRIEVE THE CORRECT RECORD
; OCCURRENCE OR RETURN AN APPROPRIATE ERROR-STATUS CODE.
;
;DIAGNOSIS
;
; THE RSE EXPECTS THAT THE identifier-2 WHICH IS PASSED TO IT MUST
; BE ONE-WORD BINARY BUT IT DOES NOT CHECK TO SEE IF IT IS.
;
;CURE
;
; INSTALL EDIT 566 TO MAKE FIND RSE 3 CHECK THE DATA TYPE OF THE
; PASSED identifier-2 AND IF IT IS NOT COMPUTATIONAL, TO REJECT
; IT WITH AN ERROR-STATUS CODE OF 358.
;
********************************************************************************
EDIT 570 FOR DBMS
;SYMPTOM
;
; IN FIND RSE 3 THE SYNTAX FOR SET OR FOR AREA CAN BE USED INDIS-
; CRIMINATELY WITH A USER-DEFINED SET-NAME OR AREA-NAME.
;
;DIAGNOSIS
;
; FIND RSE 3 DOES NOT CHECK WHETHER IT HAS BEEN PASSED THE AREA
; PARAMETER OR THE SET PARAMETER.
;
;CURE
;
; INSTALL EDIT 570 TO MAKE FIND RSE 3 CHECK WHETHER THE APPRO-
; PRIATE PARAMETER HAS BEEN PASSED WITH THE USER-DEFINED NAME.
;
********************************************************************************
EDIT 571 FOR DBMS
;SYMPTOM:
;
; WHEN USING DBMEND TO DO THE FIRST OPEN ON A SYSTEM AREA WHICH
; JUST BEEN CREATED BY THE SCHEMA PROCESSOR, DBMEND DOES NOT
; CREATE THE SYSTEM RECORD AND INITIALIZE THE SETS WHICH IT OWNS.
;
;DIAGNOSIS:
;
; DBMEND IN VERSION 5A DOES NOT KNOW ABOUT RECORDS AND SETS, EXCEPT
; FOR THE AREA STATUS RECORD, AS IT ONLY KNOWS ABOUT DATA BASE
; AREAS AND PAGES. SO, WHEN IT OPENS A NEWLY CREATED AREA, IT DOES
; NOT KNOW THAT IT IS BEING EXPECTED TO CREATE THE SYSTEM RECORD
; WHEN IT OPENS THE SYSTEM AREA. IT WOULD BE EXTREMELY DIFFICULT
; TO TEACH DBMEND ABOUT THE SYSTEM RECORD, AND BESIDES DBMEND
; SHOULD NOT BE USED TO OPEN OR FORCEOPEN A NEWLY CREATED AREA.
; RATHER, A USER-PROGRAMMED RUN-UNIT SHOULD BE DOING THIS INSTEAD.
; SO, THE BEST APPROACH IS FOR DBMEND TO GIVE AN ERROR MESSAGE IF
; IT IS USED IN THIS CIRCUMSTANCE. SINCE DBMEND DOES NOT KNOW ABOUT
; RECORDS, WE CAN USE THIS FACT AS A BASIS FOR TESTING FOR THIS
; ERRONEOUS USAGE.
;
;
; HOWEVER, THE AREA STATUS RECORD HAS ALREADY BEEN FORMATTED IN THE
; BUFFER AND THE BUFFER HAS ALREADY BEEN FLAGGED AS UPDATED BEFORE
; THIS TEST IS MADE. AS A RESULT, THIS EDIT WILL NOT UN-DO THE
; UPDATE BEING DONE BY DBMEND. IT WILL ONLY AVISE THE USER THAT HE
; HAS INCURRED A PROBLEM. AS A RESULT, HE WILL HAVE TO GO BACK AND
; RE-CREATE THE AREA AS INDICATED IN THE EXPLANATION OF ERROR-STATUS
; CODE XX43 IN THE DBMS-10/20 PROGRAMMERS MANUALS.
;
;CURE:
;
; INSTALL EDIT 571 IN ORDER TO DO THE APPROPRIATE CHECKING AND TO
; HAVE DBMEND GIVE THE ?MNDACS ERROR MESSAGE IN THIS SITUATION.
;
; NOTE 1: THIS EDIT IS INSTALLED INTO THE DBSANY CODE WHICH IS USED
; BY DBCS AND DBINFO, AS WELL AS DBMEND, FOR OPENING DATA BASE
; AREAS. IT MAY BE INSTALLED INTO DBCS AND DBINFO WITH NO ILL
; EFFECT BECAUSE DBCS AND DBINFO KNOW ABOUT RECORDS. BESIDES,
; DBINFO ONLY OPENS THE DATA BASE IN USAGE-MODE EXCLUSIVE RETRIEVAL
; AND DOES NOT UPDATE THE DATA BASE. ALSO, WHEN DBINFO ATTEMPTS
; TO OPEN A NEWLY CREATED AREA, IT ALREADY GIVES THE ?INFAOE ERROR
; MESSAGE.
;
; NOTE 2: THIS EDIT ONLY CAUSES DBMEND TO PRODUCE THE ?MNDACS
; ERROR MESSAGE WHEN IT GETS INVOLVED IN THE INITIALIZATION OF A
; DATA BASE AREA. IT DOES NOT INHIBIT FURTHER ATTEMPTS TO OPEN
; THE AREA, EVEN WITHIN THE SAME RUN OF DBMEND. THUS, IT IS
; EXPECTED THAT THE USER WILL TAKE DIAGNOSTIC ACTION ON THE BASIS
; OF THE APPEARANCE OF THE ERROR MESSAGE ITSELF.
;
********************************************************************************
EDIT 572 FOR DBMS
;SYMPTOM
;
; FORDML MAY NOT ACCEPT PRIVACY KEYS FOR INVOKE AND ACCESS
; STATEMENTS WHICH ARE LESS THAN FIVE CHARACTERS LONG.
;
;DIAGNOSIS
;
; THE CODE CHECKS THE ENTIRE SECOND WORD OF THE STRING ARGUMENT
; TO SEE IF THE LENGTH EXCEEDS FIVE AND SETS IT TO FIVE IF IT
; DOES. HOWEVER, THE LEFT HALF OF THE WORD IS USED TO KEEP TRACK
; OF OTHER ATTRIBUTES, SUCH AS WHETHER OR NOT THE STRING IS NUMERIC.
; THE LENGTH OF THE STRING IS STORED IN THE RIGHT HALF ONLY.
;
;CURE
;
; INSTALL EDIT 572 TO FORDML TO CAUSE ONLY THE RIGHT HALF TO BE USED.
;
;
********************************************************************************
EDIT 576 FOR DBMS
;SYMPTOM:
;
; EDIT 557 TO MODULE DBJOUR CHECKS A FLAG FOR THE JOURNAL-IN-
; UNDEFINED-STATE WORD, AND IF AN ERROR RETURN WAS MADE FROM
; A PREVIOUS DBCS CALL, INCLUDING FIND VERBS, IT REFUSES TO
; PROCESS THE DBCS CALL, RETURNING A XX62 ERROR.
;
;DIAGNOSIS:
;
; THE INTENT OF EDIT 557 WAS TO STOP SHARING UPDATERS TO A
; DBSM-10/20 DATA BASE FROM DOING FURTHER WORK IN THE EVENT THAT
; THE JOURNAL-IN-UNDEFINED-STATE WORD WAS FOUND TO BE SET WHEN
; ENTERING A DML VERB. IT USES THE VALUE LEFT IN THE INTERNAL
; DCBS EXCEPTION ID FIELD FROM THE PREVIOUS DML VERB AS THE
; FLAG THAT NOT ALL IS WELL WITH THE JOURNAL AND THE DATA BASE.
; HOWEVER, EDIT 557 IS TOO RESTRICTIVE BECAUSE IT FLAGS RETRIEVALS
; AS BEING DANGEROUS AS WELL AS UPDATES. HOWEVER, THE CONTROL OVER
; THE CONSISTENCY OF THE DATA BASE WHICH IS OFFERED BY THAT EDIT
; SHOULD BE RETAINED IN THE CASE OF ALL DML UPDATES. THUS, THE
; TWO DML VERBS "FIND" AND "GET" SHOULD BE MADE TO CLEAR THE
; INTERNAL EXCEPTION ID FIELD WHEN THEY ARE ABOUT TO LEAVE DBCS,
; EXCEPT IN CASES WHERE THE SUS BIT HAS BEEN PREVIOUSLY SET BY
; A FAILED DML UPDATING VERB. IN SUCH CASES, THE RUN-UNIT ITSELF
; MAY HAVE ALREADY CREATED SOME CORRUPTION AND IT SHOULD BE FORCED
; TO TERMINATE. ACCORDINGLY, ONLY RUN-UNITS WHICH HAVE OPENED THE
; DATA BASE FOR SOME FORM OF RETRIEVAL WILL BE ABLE TO CONTINUE
; TO RUN.
;
;CURE:
;
; INSTALL EDIT 576 FOR DBMS-10/20 V5A IN ORDER TO MAKE THEM CLEAR
; THE INTERNAL EXCEPTION ID FIELD AS THEY ARE ABOUT TO LEAVE DBCS
; IF THE SUS BIT IS NOT SET.
;
********************************************************************************
EDIT 577 FOR DBMS
;SYMPTOM
;
; WITH EDIT 565 INSTALLED, FINS6 INCORRECTLY EXCLUDES RECORDS FROM
; ITS SEARCH PROCEDURE, RETURNING AN ERROR-STATUS CODE OF 326 WHEN
; THE DESIRED RECORD EXISTS ON THE SORTED SET OCCURRENCE. IN OTHER
; CASES, IT WAS REPORTING A SUCCESSFUL RETRIEVAL PREMATURELY.
;
;DIAGNOSIS
;
; EDIT 565 FAILED TO UPDATE TWO "JRST" INSTRUCTIONS WHEN IT ADDED
; AN INSTRUCTION IN THE FOURTH INSERT OF EDIT 540.
;
;CURE
;
; INSTALL EDIT 577 TO CAUSE EDIT 565 TO "JRST" TO THE PROPER
; INSTRUCTION IN BOTH CASES.
;
; NOTE: THIS EDIT REMOVES EDIT 565. EDIT 565 REMOVED EDIT 540.
; IF YOU DO NOT HAVE EITHER EDIT INSTALLED, OMIT THE TWO OCCUR-
; RENCES OF THE MAKLIB INSTRUCTION ".REMOVE 565" SHOWN IN THE
; TEXT OF THE EDIT BELOW. IF YOU HAVE EDIT 540 INSTALLED BUT NOT
; EDIT 565, CHANGE THE TWO OCCURRENCES OF THE MAKLIB INSTRUCTION
; FROM ".REMOVE 565" TO ".REMOVE 540".
;
********************************************************************************
END OF DBMS-10-V5A