COPYRIGHT (C) 1984-2023 MERRILL CONSULTANTS DALLAS TEXAS USA
MXG CHANGES 01.01
=========================member=CHANGE01================================
/* COPYRIGHT (C) 1985 BY MERRILL CONSULTANTS dallas texas */
This member contains the complete set of changes made during the
Early Shipment Program (November 1983 to August 1984). It is kept
only for historical purposes. All of these changes were made before
the MXG production software was shipped, starting in August, 1984.
/* COPYRIGHT (C) 1984 BY SAS INSTITUTE Inc, CARY N.C. AND
BY MERRILL CONSULTANTS, DALLAS, TEXAS */
MXG Software Status as of August 9, 1984:
No Temporary Changes have been made to this source library.
No Support Subscription Updates have been issued.
This is the MXG production software.
This member will contain the change status (i.e., last change number)
when the library was shipped to you. It will be updated (by complete
replacement) when you receive MXG Support Subscription Updates (SSU)
from Merrill Consultants, 10717 Cromwell Drive, Dallas Tx 75229,
214-351-1966. Between SSU's, fixes will be provided as "Temporary
Changes" in printed form. You should update this member after you
install any temporary changes.
EARLY SHIPMENT CHANGES
Listed below are all of the changes were applied to the early shipment
software (Nov 1, 1983 to Aug 15, 1984). These changes have already been
made to this library, but are included here for information.
Change 00.50 PERFGRP was removed from keep list for macro _PDB30_1.
BUILDPDB For Started Tasks and TSO Users, the value of PERFGRP
BUILDPD3 in the subtype 1 type 30 record is zero, which caused
Aug 8, 1984 PERFGRP in STEPS data set to be wrong if the step was
TYPETASK=STC or TSU and a subtype 1 record was found.
This is a circumvention, until the value of PERFGRP in
the subtype 1 (Job Init) is corrected by IBM.
Thanks to Georg Simon, Prudential, USA.
Change 00.49 Advised that variable RECORDS can have a negative value.
VMAC64 Altered the input format from PIB to IB to handle these
Aug 6, 1984 potential negative values, but have not yet determined
the intrepretation of these negative values.
Thanks to Geoff Neal, Health Insurance Comm., AUSTRALIA
Change 00.48 Created member to process CICS Monitor Facility records
TYPE110J which were written to a journal file, as described in
Aug 6, 1984 the Chapter on CICS.
Thanks to Debby Blackey, Fidelity Savings, USA.
Change 00.47 Corrected spelling errors and inconsistencies in labels
DOC which the super proofreaders at SAS found in reading the
Aug 6, 1984 text at the galley stage.
Thanks to ???, SAS proofreaders, USA.
Change 00.46 Corrected spelling errors in comments in describing the
BUILDPDB SPIN logic and added a comment line where default number
Aug 6, 1984 of days to spin is set.
Thanks to Patricia McKenzie, B.C. Systems, CANADA.
Change 00.45 Test to define DEVICE variable expanded to include JES3
VMAC6 printers (which have PRTnnnnn format for OUTDEVCE value.
Aug 6, 1984
Thanks to Paul Walker, IBM, USA.
Change 00.44 CPU busy is now zeroed and CPUWAITM set to the duration
VMAC7072 of the interval if either the CPU was offline at the end
Aug 6, 1984 of interval, or for an interval during which the CPU was
varied on or off line. Early shipment code only tested
for offline at end of interval.
Thanks to Patricia McKenzie, B.C. Systems, CANADA.
Change 00.43 A nit. IMACUCB in the comments was changed to VMACUCB.
VMACUCB Even nits require a change, and thus are documented here.
Aug 6, 1984 Thanks for my primary nitpicker!
Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.
Change 00.42 The page fault rate variables (SYSPFR for system areas,
VMAC71 and LCLPFR for local area) were slightly in error. SYSPFR
Aug 6, 1984 included LPAGINS twice (because PVTCAIN already includes
LPAGINS), which caused LCLPFR to be smaller by a LPAGINS.
Variables PFRATE and PAGING are replicated as variables
PTR and DPR for report program compatability. PFRATE and
PAGING existed before MVS/SE1 introduced PTR and DPR as
"Happy Value" SRM measures. Note there is a significant
change in TYPE71 in MXG when compared with the original
MG code. In MXG the paging variables are now calculated
as rates per second, whereas the original code variables
contained total pages during the RMF interval. You will
have to check report programs which use TYPE71 data and
will likely need to modify your SAS program.
Thanks to Wayne Lauck, Central Data Processing, USA.
Change 00.41 The variable names starting with LOG.... were changed to
ANALNPA begin with NPA.... to agree with the names in TYPE38.
Aug 6, 1984
Change 00.40 Format for timestamps in subtype 1 and 2 records was not
VMAC90 correct in SMF manual; documentation APAR OZ81250 will be
Aug 6, 1984 issued by IBM to correct the manual. MXG was corrected to
decode fields as data actually exists. Structure of code
was altered to prevent possible errors when no active SMF
data set is given in a type 90 record, and variable BITS
was assigned format $HEX64.
Thanks to Paul Walker, IBM, USA.
Change 00.39 Variable IPLTIME, with length 8, was added to the LENGTH
VMAC0 statement. Timestamp variable must be length 8 to prevent
May 31, 1984 loss of minutes and seconds of precision.
Thanks to Paul Walker, IBM, USA.
Change 00.38 The error handling of CICS CMF records introduced in ES23
VMAC110 creates *ERROR.VMAC110.1 for all dictionary entries. The
May 28, 1984 dictionary entry is identifiable by the value of zero for
variable MCTSSDID in the second line after the "SECTION
HEADER DESCRIPTOR." The error message has no effect on
the data, but is eliminated by changing the test for this
error message from
OR SECTLLBB LT TEMPBYTE
to
OR (SECTLLBB LT TEMPBYTE AND MCTSSDID NE 0)
Additionally, CICS 1.6.1 has some initial data values in
error which cause IRESPTM, the internal response time, to
be very negative. Variable WTTCIOCN is 65535 and WTTCIOTM
is completely invalid. In all cases observed so far, the
value of TASERRFG was non zero for these values, and thus
the calculation of IRESPTM is now only calculated as:
IF TASERRFG EQ 0 THEN IRESPTM=ELAPSTM-WTTCIOTM;
There may in fact be PTFs which have not been applied at
my CICS 1.6.1 test site which cause these data values. It
is still wise to calculate the response only when it is
valid to do so!
Change 00.37 TWO NEW VALUES WERE ADDED TO THE MG110ER FORMAT, DUE TO
FORMATS new error flag values appearing in CICS 1.6.1:
May 28, 1984 00X=' '
50X='50X:INVALID CLOCK AND INTERNAL ERROR'
OTHER='OTHER ERROR'
Change 00.36 VARIABLE ALLOCATN IS REPLACED BY THREE ONE-BYTE VARIABLES
VMAC25 describing how this allocation was made: ALOCBYDD ('Y' by
May 25, 1984 DD, 'N' for dynamic), ALOCATLG ('Y' if Catalog used, 'N'
if not), and ALOCAUTO ('Y' if automatic, 'N' if manual).
Thanks to Paul Walker, IBM, USA.
Change 00.35 STATEMENT SWAPRATE=SWAPRATE*INVTIME; WAS REPEATED IN BOTH
VMAC71 the MVS/370 and MVS/XA segments. The second occurrence was
May 25, 1984 deleted in both segments.
Thanks to Jim Trenkle, Safeway, USA.
Change 00.34 VARIABLE TYPETASK UNDER JES3 CONTAINS A ZERO RATHER THAN
VMAC26J3 blank for the fourth position. A blank is now forced into
May 25, 1984 the fourth position. It was not shortened to three-bytes
to maintain consistency with TYPETASK in other data sets.
Thanks to Jim Trenkle, Safeway, USA.
Change 00.33 WORKLOAD VARIABLES WHICH WERE INCORRECTLY CALCULATED WERE
RMFINTRV corrected as described in NEWSLETTER Volume I Number Two.
May 25, 1984
Change 00.32 VARIABLE JOBID WAS REMOVED FROM KEEP LIST. IT HAD BEEN
VMAC26J3 replaced by variables TYPETASK and JESNR and should never
May 20, 1984 have appeared in the KEEP statement.
Thanks to Paul Walker, IBM, USA.
Change 00.31 NEW MODULE, FOR TYPE 110 CICS RECORD WHICH SETS THE SIZE
IMACICUS of the USERCHAR variable in CICSTRAN. If you want account
May 20, 1984 data on CICS transactions, you must use the optional user
character field. Default is for a one-byte field. See the
discussion in the member, and more in Newsletter #2.
Change 00.30 NEW MODULE, FOR JES3. IT BUILDS THE PDB JUST AS BUILDPDB
BUILDPD3 for JES2, but this module also gets the account fields in
May 20, 1984 TYPE26J3 and uses them if there was no type 30 record for
the job (as happens with JCL error).
Change 00.29 CLEANUP CHANGE. EXCP AND IOTM FIELDS ADDED TO KEEP LIST
ANALRMFI for TYPE30_5. LABEL statement was too long for ONLINE in
VMAC30 TYPE73 and TYPE78, and for AVGDSOPN in TYPE74. One title
VMAC73 was incorrect in ANALRMFI.
VMAC74
VMAC78
May 20, 1984
Change 00.28 CORRECTED VALUE FOR MG090CM FORMAT (TYPE90 DATA SET) TO
FORMATS 6:SWITCH SMF.
May 10, 1984
Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.
Change 00.27 FOR IPL PROMPT, THE OPERATOR CAN REPLY WITH THE DURATION
VMAC90 of the outage, or "U". MXG raised "invalid function call"
May 10, 1984 error if "U" was response. Code now checks for "U" before
attempting to convert string to duration. Also changed
TOD to EVENTIME for several events so that time of event
would actually be kept.
Thanks to Paul Walker, IBM, USA.
Change 00.26 VARIABLE SMF49L CORRECTED TO SMF49LN IN THREE PLACES TO
VMAC4789 correct error. JES3 Type 48 records for SNA were found to
May 10, 1984 be incorrectly documented in the SMF manual, resulting in
a STOPOVER Abend. After reaching Level II, the code was
fixed and a document APAR from IBM can be expected. At the
same time another of my errors was corrected.
Thanks to Paul Walker, IBM, USA.
Change 00.25 INVALID TYPE 30 RECORDS HAVE BEEN CREATED BY MVS WITH THE
VMAC30 'triplets' (the three values of offset, length, and number
Apr 26, 1984 of relocatable sections) overlaid with blanks. This caused
a STOPOVER abend when the blanks were read in as PIB4! The
offset value plus length of the relocatable section is now
compared with the LENGTH of the record, and the invalid
records will be printed in hex on the log.
Change 00.24 OCCASIONALLY, THE NUMBER OF SAMPLES NRSAMPLE WILL BE ZERO
VMAC7072 in a type 70 record. It appears this can happen if the
Apr 26, 1984 operator issues a Modify command at startup, although it
may be due to other phenomena. The records have a very
short duration, and have little other effect, but the
division by zero caused an abend. This change tests for
nonzero value of NRSAMPLE before division. Additionally,
the divisions by CPUCOEFF and SRBCOEFF are also now tested
for nonzero for CPUTCBTM and CPUSRBTM in TYPE72 data set.
Finally, the error messages for a missing value of SU_SEC,
(the service units per CPU second) were cleaned up.
Thanks to Sue Rosansky, Metropolitan Life, USA.
Change 00.23 CHANGES TO SUPPORT CICS VERSION 1.6.1
VMAC110 Seven new variables in data set CICSTRAN:
Apr 25, 1984 MAXTASK Maxtask condition occurred?
PRIINCHR Primary input character count
PRIOUCHR Primary output character count
PROGSTOR Program storage used
SECINCHR Secondary input character count
SECOUCHR Secondary output character count
SHRTSTOR Short on storage condition occurred?
One new variable in data set CICSYSTM:
PROGCOMP Amount of storage deleted by CICS program
compression dynamically because system storage
was overloaded.
Much better detection of invalid segments. Prior error
handling deleted remainder of physical record when an
invalid header was encountered. Now, the invalid section
header is skipped, but subsequent valid sections are
processed.
Change 00.22 CHANGES TO SUPPORT MVS/SP VERSION 2 RELEASE 1.2:
VMAC434 REGREQST created from four byte field, replacing old
Apr 20, 1984 two byte field, PVTAREA:
PVTAREA - Now zero (see REGREQST)
REGREQST - REGION requested (from JCL)
VMAC6 Nine new variables to support 3800-3 printer are added:
Apr 20, 1984 DOCLENFT - Length of document printed, in feet.
FONTLOAD - Number of Fonts Loaded.
FONTUSED - Number of Fonts Used.
FORMDEFS - Number of FORMDEFs Used.
OVLYLOAD - Number of Overlays Loaded.
OVLYUSED - Number of Overlays Used.
PAGEDEFS - Number of PAGEDEFs Used.
PGSGLOAD - Number of Page Segments Loaded.
PGSGUSED - Number of Page Segments Used.
Additionally, FORM is now character length eight (it used
to be only four).
VMAC426J 2 OUTFORM is now eight bytes (increased from four).
Apr 20, 1984
VMAC30 Six overlooked variables and the new one added (REGREQST)
Apr 20, 1984 are added to data sets TYPE30_V, TYPE30_4, TYPE30_5 and
TYPE30_6:
LSQSZHI - LSQA+SWA Subpools above 16MB
LSQSZLOW - LSQA+SWA Subpools below 16MB
PVTSZHI - Private Area size above 16MB
PVTSZLOW - Private Area size below 16MB
USRSZHI - User sub pools above 16MB
USRSZLOW - User sub pools below 16MB
PVTAREA - Now zero (see REGREQST)
REGREQST - REGION requested (from JCL)
Change 00.21 IBM PTF WILL BE ISSUED TO CHANGE END OF FILE TEXT IN THE
VMACSMF VSAM SMF data from "SMF EOF MA" to "SMFEOFMARK". Test was
Apr 14, 1984 altered to support either string.
Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.
Change 00.20 TYPE 30 RECORD STOPOVER DUE TO INVALID EXCP SEGMENT
VMAC30 description. There should have been 251 22-byte segments,
Apr 14, 1984 but record was only 1800 butes. A new check was added to
verify the record length could be read before reading EXCP
data, and an error message put to the log, with dump of
record, if same type of invalid record is encountered.
The problem will be persued with IBM to uncover their
cause of the bad records. At same time, error messages in
VMACEXC2 were numbered and enhanced.
Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.
Change 00.19 OFFSET VALUE FOR THREE VARIABLES WAS WRONG FOR TYPE 77
VMAC77 MVS/XA segment.
Apr 12, 1984 Variable @was @change to
OFFEDSS 33 41
LENEDSS 37 45
NRSECT 39 47
Change 00.18 BUILDPDB FAILED AFTER CHANGE ES11 BECAUSE VARIABLES IN BY
BUILDPDB list were not in the _NULL_ data set (which replaced the
Apr 12, 1984 SPIN.SPINnn data sets due to OPTIONS NODSNFERR). Fix
required minor redesign of the interleave of today's and
yesterday's (i.e., SPIN) data sets. If you do not ever
use the SPIN data sets, a short fix is to precede OPTIONS
NODSNFERR statement with an /* and to follow the
subsequent OPTIONS DSNFERR statement with an */, which
comments out the interleave of TYPEnnnn and SPIN.SPINnnnn
data sets.
Thanks to Bob Warren, Wachovia Bank, USA.
Change 00.17 VARIABLE CPUCPTM WAS NOT IN KEEP LIST FOR VMSESSN.
TYPEVM
Apr 10, 1984
Change 00.16 DIVISION BY ZERO WHEN SRBCOEFF IS ZERO. CORRECTED BY TEST
VMAC7072 for zero before division.
Apr 10, 1984
Thanks to Bob Warren, Wachovia Bank, USA.
Change 00.15 CREATION OF CHARACTER VARIABLE TIME FROM STRTTIME RESULTS
ANALCICS in invalid sort sequence when month boundary was crossed.
Apr 1, 1984 Reconstructed as a DATETIME value.
Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.
Change 00.14 CPU FACTORS FOR THE GX,KX,BX,EX, AND QX 308X PROCESSORS
VMAC7072 were added and comments added to the MVS/370 section Minor
Mar 31, 1984 corrections to some values, based on IBM publication TNL
GN28-0912 (Jan 31,1984) to GC28-1149-1 were also made.
Change 00.13 FORMATS FOR CONTROL/IMS ADDED TO MEMBER FORMATS.
FORMATS
Mar 31, 1984
Change 00.12 TYPETASK='TSO' CORRECTED TO TYPETASK='TSU'.
VMAC30
Mar 31, 1984
Thanks to Chuck Hopf, Computer Language Research (FASTAX), USA.
Change 00.11 IN INTERLEAVING TYPE30_1 AND SPIN.SPIN30_1, THE BY
BUILDPDB statement had JINTTIME misspelled as JINITIME. At the
Mar 31, 1984 same time, MACRO _SECOND and its percent sign were
replaced by OPTIONS NODSNFERR and OPTIONS DSNFERR as
described in Chapter 32.
Thanks to Steve Glick, NTSU, USA.
See Change ES18.
Change 00.10 MANY LABELS WERE RESPELLED (CORRECTLY THIS TIME), AND ALL
VMAC.all occurrences of the following labels were made consistent:
Mar 31, 1984 DATASET to DATA SET
TIMESTAMP to TIME STAMP
JOBNAME to JOB NAME
Thanks to ???, SAS proofreaders, USA.
Change 00.09 THE LENGTH VALUE USED TO READ IN THE SIGNON VARIABLE
VMAC4789 includes the two bytes of length itself, causing a
Mar 31, 1984 STOPOVER abend. Code restructured to subtract two from
length before executing the $VARYING60. INPUT statement.
Additionally, the offset @41 in ID=49 code was typographic
and should be @51.
Thanks to Allan Russell, SAS Institute Europe, GERMANY.
Change 00.08 RMV VERSION 3.2.1 CHANGED RMF VERSION FROM 2 BYTE EBCDIC
VMAC7072 to a 2 byte packed decimal. Now version can print as 321
VMAC73 whereas before it was limited to 32.
VMAC74
VMAC75
VMAC76
VMAC77
VMAC78
Feb 4, 1984
Change 00.07 SUPPORT FOR RMF VERSION 3 RELEASE 2 MODIFICATION LEVEL 1
VMAC78 installed. RMF 3.2 added TYPE78VS, TYPE78SP and TYPE78
Feb 4, 1984 data sets and 3.2.1 added support for 4381 processor
types. Only TYPE78VS data set has actually been
validated.
Change 00.06 IMPROVED ERROR HANDLING FOR INVALID RECORDS NOW PROVIDES
VMAC110 diagnostic messages and a dump of the type 110 record when
Feb 4, 1984 invalid data is encounterd. This will eliminate STOPOVER
abends in type 110.
Change 00.05 AVGWKSET VARIABLE REPLACED AVGMEMSZ WITH NEW DESCRIPTION
VMAC7072 and discussion in TYPE72 section of Chapter Forty.
Jan 15, 1984
Change 00.04 AVGWKSET LABELS WERE CORRECTED TO UNITS OF (K-BYTES) AND
VMAC30 the calculations made consistent in these data sets.
VMAC434
VMAC48PC
Jan 15, 1984
Change 00.03 REPEAT (LOGICALLY) CHANGE ES2 FOR SIO74CNT, SSCHSAMP AND
VMAC74 IORATE.
Dec 23, 1983
Change 00.02 SEVERAL VARIABLES WERE NOT SET TO MISSING WHEN NRSTCPS,
VMAC73 the number of samples, was zero, and thus had the value of
Dec 23, 1983 value of the previous calculation.
Change 00.01 DELETE OBSERVATION FROM SET PDB.TYPE72 WHEN IT WAS FROM A
RMFINTRV report performance group. To correctly calculate overhead
Dec 23, 1983 CPU time, the sum of only the control performance groups
is used.
END OF Early Shipment Changes, December 23, 1983.