* copyright (C) 1984-2019 merrill consultants dallas texas usa
Yüklə 28,67 Mb.
|
|
-IBM offset EXO is strange, required +5 to locate the data segment. (-3 is normal, rarely +1, but never +5 before!), but it is now used to correctly align each input. -IBM length EXL is also "wrong", as it is the TOTAL length rather than the expected length of one segment, so it is now divided by EXN to get the per-segment length for each segment's input. -GMT Offset is now kept in existing SM1209CD variable. -SM120STM/SM120ETM are now correctly labeled. Thanks to Paul Volpi, UHC, USA. Thanks to Scott Barry, SBBWorks Inc., USA.
VMXGINIT MXGWARN: SAS 9.3 TS1M0 DETECTED. SAS HOT FIX (SN43828) Apr 14, 2014 MXGWARN: REQUIRED TO PREVENT LET-STATEMENT ERRORS. MXGWARN: BETTER STILL:: 9.3 TS1M1 CONTAINS THE FIX. (This need was documented in CHANGE 29.159 in 2011.)
VMAC102 INPUT and kept in T102S106 dataset: Apr 15, 2014 QWP1DATH QWP1DPSS QWP1DSGN QWP1DSSZ QWP1DXAC QWP1LBIL QWP1LOGM QWP1LOGR QWP1ZPNM QWP4AECR QWP4APCO_VAR QWP4ATRC QWP4CDDC QWP4CDMC QWP4CDSC QWP4CDTSL QWP4CXDC QWP4CXMC QWP4CXSC QWP4DEGD QWP4DLRU QWP4FCPY QWP4HASH QWP4IAST QWP4IMWF QWP4IXCU QWP4KLRU QWP4MIMTS QWP4MIS7 QWP4MIS8 QWP4MIS9 QWP4MISA QWP4MISB QWP4MS4F QWP4MUSE QWP4N0193A QWP4N0193B QWP4PELM QWP4PFLG QWP4PFUP QWP4PLMR QWP4PLSF QWP4PMSC QWP4QRWD QWP4RACK QWP4RDVPR QWP4RMDB QWP4RSLV QWP4S1IL QWP4SFBS QWP4SLRU QWP4WFSAT QWP4WFSST QWP4SEPSD QWP4SECA1_E QWP4SECA1_TYPE QWP4SECADM1 QWP4SECA2_E QWP4SECA2_TYPE QWP4SECADM2 QWP5BPM QWP5APM QWP9TCPVX QWP9MCONQN QWP9MCONQW QWP9DDFC QWP9TCPA Thanks to Wolfgang von Schumann, Finanz Informatik Tech, GERMANY. Thanks to Norbert Philips, Finanz Informatik Technologie, GERMANY
ASUMTALO printed that it had been added to the BY LIES instead of Apr 9, 2014 BY LIST. Thanks to Jorge Fong, NYC Information Technology, USA. Change 32.083 ANALID did not follow MXG conventions with an &Pxxxxxx or ANALID a _L/_Wdddddd tokens. Revised to use the &PDBMXG macro Apr 5, 2014 variable so that if the output destination was overridden by VMXGINIT SMFRECNT will end up in the right place. Change 32.082 These NETEZZA/IDAA PDB.DB2STATS Q8STxxxx variables should VMACDB2 not have been deaccumulated: Apr 5, 2014 Q8STACTV_64 Q8STCCPU_64 Q8STCORS Q8STCPMU Q8STCQL Q8STDSA Q8STDSKA Q8STDSKB Q8STDSKU Q8STMAXA_64 Q8STMAXQ Q8STMNQS Q8STNMDS Q8STNQCS Q8STTATE Q8STWCPU_64 Q8STWNOD_64 IBM doesn't always identify which fields are accumulated, so the DIF() code is put in place, and then data is read to detect negative values (proof of not accumulated so those variables are removed) but I failed to run the test for negative values in Change 32.072. Thanks to Giuseppe Giacomodonato, EPVTECH, ITALY. Change 32.081 A minor exposure in UTILBLDP. If IMACKEEP= was not used UTILBLDP (blank) AND BUILDPDB was YES then the substitution macros Apr 4, 2014 for SPINCNT SPINUOW and TIMEDIF were defined twice which could result in MACRO 0 0 % which is not a valid syntax and produced an error message but did not stop execution. Now only generated if BUILDPDB EQ YES. MACKEEP and MACFILE are now nulled out after execution.
ANALDB2R to select "Starting With". If you used DB2=DSN to select Apr 3, 2013 only the DSN subsystem, but have subsystems DSN DSNT and Apr 15, 2014 DSNP, all three will be selected. New parameters INCODEACCT and INCODESTAT let you define your own selection criteria. INCODEACCT= applies to the accounting reports and many trace reports, while INCODESTAT= applies to the stats reports. For the For the above single-subsystem selection, you would use: INCODEACCT=IF QWHSSSID='DSN'; -MXGACC03 could fail on zOS with dataset not found error because SYSLAST was overwritten by VGETOBS but the last dataset created by VGETOBS no longer existed, because the TABULATE did not contain a DATA= argument.
VMAC90A 9036 instead of 9035. Apr 3, 2014 Thanks to Chris Weston, SAS ITRM, USA. ====== Changes thru 32.078 were in MXG 32.03 dated Apr 3, 2014========= Change 32.078 Support for OAM SMF 85 record for z/OS 2.1 (INCOMPAT). VMAC85 The record VERSION test required an explicit '2010' value Apr 3, 2014 to prevent INPUT STATEMENT EXCEEDED RECORD error. Thanks to Joachim Sarkoschitz, DATEV, GERMANY. Change 32.077 Support CICS/TS 5.2 OPEN BETA: CICSTRAN COMPAT, Stat NOT: VMAC110 -5.2 CICSTRAN has been SUPPORTED since MXG 31.03 (3/2013), Mar 31, 2014 as NO NEW FIELDS WERE ADDED TO CICSTRAN IN CICS/TS 5.2, Apr 1, 2014 at least not in the OPEN BETA (so GA could be changed!). (MXG 31.03 or later is required for CICS/TS 5.1.) -5.2 STID=62 CICDS Dispatcher CPU Statistics (per-TCB CPU) records read with MXG 32.02 or earlier are wrong/missing values, due to fields inserted in STID=62 segments. MXG does detect and report that STID=62 was CHANGED, but only in obscure ***MXG WARNING.TYPE110 messages that don't tell you the output was trashed. MXG did continue to read the rest of your SMF data after those warnings. WITH THIS CHANGE: CPU times are correct. -STID=62 CICDS dataset has these variables read from the header and kept DSGLXSCN='LAST EXCESS*TCB SCAN' DSGLXSND='LAST EXCESS*TCB SCAN*NO TCB*DETECTED' and 18 sets of these three variables are added, one per TCB, with "DSG" replaced by each TCB's prefix: DSGTMCDQ='QR TCB*CURR TASKS*DISPATCH*QUEUE' DSGTMCDQ='QR TCB*PEAK TASKS*DISPATCH*QUEUE' DSGTMCDQ='QR AVG*PEAK TASKS*DISPATCH*QUEUE' -STID=62 CICSPOOL Pool dataset has these five new varS; DSGLTCB='OPEN POOL*DATETIME*WHEN*POOL LIMIT*REACHED*/ DS2LTCB='JVM POOL *DATETIME*WHEN*POOL LIMIT*REACHED*/ DS3LTCB='XPT POOL *DATETIME*WHEN*POOL LIMIT*REACHED*/ DS4LTCB='SSLD POOL*DATETIME*WHEN*POOL LIMIT*REACHED*/ DS5LTCB='THRD POOL*DATETIME*WHEN*POOL LIMIT*REACHED*/ -5.2 also changed STID 10, 81 and 105, but COMPATIBLY with new data appended rather than inserted, so prior versions MXGWARNed there was new data and skipped with no impact. -5.2 new STIDs 36 and 147 were also safely skipped. -STID=10, CICXMG dataset, XMG DSECT, new variables: XMGATMXT='80X*IF*CURRENTLY*AT MXT' XMGGAMXT='GMTTIME*WHEN*MXT*REACHED' XMGGSMXT='GMT*WHEN*MXT*SET' XMGGTAT ='GMT*WHEN*LAST*ATTACHED' XMGLAMXT='DATETIME*WHEN*MXT*REACHED' XMGLSMXT='DATETIME*WHEN*MXT WAS*SET' XMGLTAT ='DATETIME*WHEN*LAST*ATTACHED' -STID=81, CICM dataset, MNG DSECT, new variables: MNGAUTRT='AVG*USER*RESPONSE*TIME' MNGCAUTA='CURR TASKS*AT LAST*ATTACH' MNGLUCTL='DATETIME*WHEN*LAST*ATTACH' MNGLUCTL='DATETIME*WHEN*LAST*TRANS*ENDED' MNGLUTRT='DATETIME*WHEN*PEAK*RESPONSE' MNGMXUTA='MXT VALUE*AT LAST*ATTACH' MNGPUTRT='PEAK*USER*RESPONSE*TIME' MNGSTNUM='SYSTEM*TRANSACTIONS*ENDED' MNGUTNUM='USER*TRANSACTIONS*ENDED' -STID=105, CICPIR dataset, PIR DSECT, new variables: PIRMSGFO='MSGFORMAT' -New STIDs 36 and 147 are listed as STILDP and STIPGE and listed as PRIVATE LODER and PRIVATE PROGRAMDEF, but the OPEN BETA DATA AREAS does not have LDP nor PGE, yet.
VMAC80A 'UID' segment with 6 bytes of data but TOK80LN2 only 3; Mar 31, 2014 this is detected, an MXG error message printed of the first two instances, and two due to the increased lengths of TOKQUEST2 and TOKANS2 fields (where MXG has to pick a max length, but I guessed wrong). -Variable RACF31DS, RVARY DSN added to TYPE8025 dataset. -TOKDANAM values 'FTEL' 'MTEL' 'EMAIL' are decoded into new TOKFTEK/TOKMTEL/TOKEMAIL variables and 'ADDRESS1' and 'ADDRESS2' values are output in existing TOKAD1/TOKAD2 variables in TYPE80TK dataset. Thanks to Phil Grasser, Norfolk Southern, USA. Change 32.075 Retrieves current ODS destination into a global variable VGETDEST VGETDEST. If no ODS destination has been set it will get Mar 31, 2014 the ODSDEST option value (normally AUTO), which on zOS is LISTING and on ASCII is HTML. Used internally in MXG in new SGPLOT reports.
VMAC102 previously only the first segment was output. But obs Mar 27, 2014 are only output after deaccumulation if QW0402ACTIVE, Mar 31, 2014 QW0402ACTIVE=SUM(QW0402TE,QW0402TQ,QW0402TF,QW0402TW, QW0402CE,QW0402CW,QW0402OE,QW0402OW); shows there was activity during this 1-minute interval. -The BY list _B102402 now includes QW0402PI, Profile ID, which is the unique token in each segment. Thanks to Paul Walters, Navy Federal Credit Union, USA.
VMACSMF //SMF DD to read dumped SMF VBS data, or the //LOGGER DD VMXGINIT to directly read current SMF logstream data, or to Mar 27, 2014 first read //SMF DD (which can be a concatenation of SMF VBS data, and then read the //LOGGER DD. The default of SMF is unchanged, to read the //SMF DD, but you can use %LET MXGREADSMF=LOGGER; to ONLY read the LOGGER file, or you can use %LET MXGREADSMF=BOTH; to read both. -For LOGGER read, OFFSMF=0 is now set in VMACSMF, so the macro variable READLSMF (Change 31.235) was removed from VMACSMF as no longer used by this change, which also removed READLSMF from VMXGINIT (in MXG 32.03). Apr 22: I had to send this 32.03 VMXGINIT to a site with the INSTREAM error; using the new VMXGINIT with the old VMACSMF did NOT work, and generated spurious errors of INVALID SMFTIME; removing READLSMF from VMXGINIT caused UNRESOLVED MACRO READLSMF condition, which caused MXG's test for &READLSMF EQ 0 to fail, so VMACSMF fell thru and attempted to read the VBS //SMF file with the (correct for Logger) attributes of RECFM=VB,LRECL=32756, which broke logical records and caused INVALID SMFTIME errors. This won't likely occur again, but I put READLSMF back in VMXGINIT's %GLOBAL statement. -Variable READSMF contains 'SMF' or 'LOGGER' to identify the source of each record during processing, variables NDUMPREC and NLOGREC count the records read from //SMF and //LOGGER respectively. You could hex-dump the first record from each infile with %LET MACFILE= %QUOTE( IF NDUMPREC=1 THEN LIST; IF NLOGREC=1 THEN LIST; ); Thanks to Jim Horne, LOWE's, USA. Change 32.072 -DB2ACCT variables QWACAACC and QWACAACW (IDAA statistics) VMACDB2 are now INPUT and kept. Mar 28, 2014 -These DB2STATS IDAA/NETEZZA variables added in DB2 V11 are now INPUT and kept: Q8STACTV_64 Q8STCQL Q8STCRL Q8STCSS Q8STDSA Q8STMAXA_64 Q8STMAXQ_64 Q8STMNQS Q8STNDS Q8STNBA Q8STNBS Q8STNDA Q8STNIA Q8STNIS Q8STNLRA Q8STNLRS Q8STNLTA Q8STNLTS Q8STNQCS Q8STNQFA Q8STNQSA Q8STNUA Q8STNUS Q8STTART Q8STTATC Q8STTCCA Q8STTCCS Q8STTCMA Q8STTCMS Q8STTCQA Q8STTCQS Q8STTLSC Q8STWCPU_64 Q8STWNOD_64 PROBLEMS: -The data needs validation; many values are zero making determination difficult; most values do look reasonable, but there are 8 extra bytes not in the DSECT that are skipped at the top of the segment, which might be the wrong place, but there is consistency with the "ALL DB2" variables with "S" suffix that all have values of -1 (INPUT as IB8) while their "THIS DB2" variables with "A" suffix are populated. Thanks to Giuseppe Giacomodonato, EPVTECH, ITALY. Change 32.071 Connect-Direct NDMCPU='CPU TIME*OF STEP' is now added to VMACNDM the NDMPT dataset, along with these new variables: Mar 26, 2014 NDMECP0 ='CPU TIME*ON CP' NDMECP1 ='CPU TIME*ON ZIIP' NDMECP2 ='ZIIP*QUALIFIED*PART OF*NDMECP0' NDMHQSTA ='QUEUE*STATUS*IF*PTQHOLD' NDMSUBJID='SUBMITTING*JCTJOBID' NDMSUBJOB='SUBMITTING*JOB' And these variables are added to NDMCT dataset: NDMECP0S='SEND*CPU TIME*ON CP' NDMECP1S='SEND*CPU TIME*ON ZIIP' NDMECP2S='SEND*ZIIP*QUALIFIED*PART OF*NDMECP0S' NDMECP0R='RECV*CPU time*on CP' NDMECP1R='RECV*CPU time*on ZIIP' NDMECP2R='RECV*ZIIP*QUALIFIED*PART OF*NDMECP0R' Thanks to Michael Oujesky, DTCC, USA. Change 32.070 Support for ASG-TMON CICS TS V3.4 (NO MXG UPDATE NEEDED). TYPETMO2 TMON CICS v3.4 is the toleration support for CICS/TS 5.2, Mar 24, 2014 which is now in Open Beta, but there were NO changes to their data records so there is no associated update. Change 32.069 When trying to reduce the line size of the code being VMXGPARS generated by UTILBLDP to 65 bytes to fit on neat and Mar 24, 2014 pretty lines in INSTREAM, VMXGPARS parsing went backwards from BYTE 65 to 50 looking for a blank but if it did not find one it arbitrarily stopped, which could be in the middle of a quoted string causing broken syntax and many strange errors. Now, the search continues until a blank is found. This error was NOT reported. It was exposed when testing UTILBLDP with new syntax in MACFILEX=.
ANALDB2R Anaheim Session 14610, Robert F. Catterall, IBM, "Key Mar 23, 2014 Metrics for DB2 for z/OS Subsystem and Application Performance Monitoring", using the DB2STATB dataset. Based on the total read I/O rate, it can suggest if the Buffer Pool is too small, or if perhaps it is too large. With SAS 9.3+, charts are produced, followed by a report identifying BPs these problem criteria: Avg read IO rate GE 1000/sec Max read IO rate GE 5000/sec Total BP size exceeds 50% of available CSTORE Max used buffers LT 25% of BP If no BPs fit the exception criteria a message is produced telling you that. This is report MXGDB2B1 in ANALDB2R.
VMAC64 variable is not populated; TYPE64 observations with Mar 21, 2014 SITUATN='NO SPACE AVAILABLE' or 'CATALOG UPDATE' do not populate OPENTIME; OPENTM calculation is now protected. Some of these obs have large negative value in UNUSEDCI.
IMACDSCK contain a "DSNAME" value with the new &MACDSCK facility. Mar 21, 2014 See the documentation examples in IMACDSCK, but this will find all observations in ISM SMF record with DSNAME that starts with SYS1." %LET MACDSCK= %QUOTE ( LENGTH DSNAME ENTRNAME $44; IF INDEX(DSNAME,'SYS1.') OR INDEX(ENTRNAME,'SYS1.') THEN ); %UTILBLDP(BUILDPDB=NO,OUTFILE=INSTREAM, USERADD=1415 1718 42 60 6156 62 6 %INCLUDE INSTREAM; RUN; Thanks to Scott Barry, SBBWorks, Inc., USA. Change 32.065 Support for APAR OA44798 enhancement to SMF ID=22 record, VMAC22 added the new Subtype 12 Storage Element Extension, adds Mar 21, 2014 two variables to the existing TYPE22_3 dataset: SMF22XGL='BLOCK*NUMBER OF*FIRST*FRAME' SMF22XPG='ONLINE*FRAMES'
ANALID prevented ODS output and are now removed. Mar 20, 2014 Thanks to Sabrina Mandelatz, Provincial, GERMANY. Change 32.063 Hydra TS7700 IBM reports are replicated in ANALBVIR. ANALBVIR These reports are completed: JCLBVIR REPORT DESCRIPTION Mar 18, 2014 H20VIRT VNODE VIRTUAL DEVICE HISTORICAL RECORDS Apr 6, 2014 H21ADP00 VNODE VIRTUAL DEVICE HISTORICAL RECORDS H21ADPXX VNODE VIRTUAL DEVICE HISTORICAL COMBINDED H21ADPSU VNODE ADAPTOR HISTORICAL ACTIVITY COMBINDED H21ADPTP VNODE ADAPTOR THROUGHPUT DISTRIBUTION H30TVC1 HNODE HSM HISTORICAL CACHE PARTITION H30TVC2 HNODE HSM HISTORICAL CACHE PARTITION (1) H32TDU12 HNODE LIBRARY HISTORICAL DRIVE ACTIVITY H32CSP HNODE LIBRARY HIST SCRTCH POOL ACTIVITY H32GUP&NUM HNODE LIBRARY HIST GUP/POOLING ACTIVITY Thanks to Joe Babcock, JPMorgan Chase, USA. Change 32.062 Variables R791PHTA/PHTI/FLG3/FLG3A/R791FLG3S & similarly VMAC79 variables R792PHTA/PHTI/FLG3/FLG3A/R792FLG3S are kept in Mar 17, 2014 datasets TYPE791 and TYPE792 respectively. Added in z/OS 2.1, but the 2.1 SMF manual had no vertical bars to mark the new fields. (IBM acknowledges the error and will fix for the 2.2 SMF manual.) Change 32.061 Graphs from two SHARE 2014 Anaheim presentations, Session GRAFCIMP 15214, "WLM Update for z/OS 2.1 and 1.13" by IBM's Horst Mar 16, 2014 Sinram, and Session 14745, "WLM - Performing a Quick WLM Performance Checkup", by Peter Enrico are replicated with with graphs of CPU resource by importance level and service class level, and then at the service class level within importance, plus a scatter graph of performance index at the importance level with additional scatter graphs at the service class level within importance. As a side benefit you can also see how many service class periods are active at each importance level. The graphs are produced by PROC SGPLOT; SAS/GRAPH is NOT used. Example report output can be at viewed at: Http://www.mxg.com/download/Change 32.061 Change 32.060 %INCLUDE of VMXGUOW was needed in these examples to get JCLUOW _SUOW and _BSUOW values, but that could cause errors due JCLUOWP to a conflict executing VMXGUOWC. This change inserts the JCLUOWV _SUOWxxx and _BSUUOW macro definitions in these members Mar 16, 2014 to eliminate the %INCLUDE and associated exposure.
EXZOSAJB because existing subtype 32 header had inserts). EXZOSAPO Three new datasets are created: EXZOSASU dddddd Dataset Description IMACZOSA ZOSASU ZOSASUMM CSM Summary VMACZOSA ZOSAPO ZOSAPOOL CSM Per Pool VMXGINIT ZOSAJB ZOSAJOB CSM per JOB Mar 14, 2014 -Outstanding issues with first iteration: SYSTEM variable is blank. Intervals drift 2 seconds per hour, and are not SYNC with SMF. JOB dataset does not have READTIME nor JESNR. Change 32.058 Thruput Manager INVALID ARGUMENT- TPMPI=INPUT(TPMPICH,1.) VMACTPMX because the previously expected $EBCDIC1 numeric value of Mar 13, 2014 importance is now a binary ('03'x) requiring a new test and conditional input of TPMPI, in dataset TPM16J. Thanks to Scott Barry, SBBWorks, Inc., USA. Thanks to Al Sherkow, I/S Management Strategies, Ltd. Change 32.057 z/VM data files for Velocity zXPS, MONWRITE, etc, MUST be VMACXAM FTP'd with TYPE E specified, which is NOT the default. Mar 13, 2014 Worse, when the default IS used its value is NOT shown on the ftp log. The ftp default is TYPE A, ASCII text, and I was unaware that when TYPE A FTP transfer goes from one EBCDIC system to another EBCDIC system (e.g. z/VM to z/OS), it changes from EBCDIC to ASCII and back to EBCDIC, which leaves the UPPERCASE TEXT unchanged, but the binary data is mangled. One specific mangle: '64'x original value becomes '3F'x, so the 100 byte length of SYTSYP became 63 which caused MXG ERROR messages. This change tests for a length of not-100 and issues a USER ABEND 666 to report the wrong FTP TYPE was used. Using MODE BLOCK is also recommended for these ftp's. Thanks to Wang Zahn, AIG, USA. Thanks to Rob van der Heij, Velocity Software, EUROPE. Change 32.056 CANNOT CLONE BUFFSIZE warnings during PROC COPY might be VGETOBS bad for execution EXCPs and run time, and this could have Mar 13, 2014 been caused by VGETOBS, under unusual circumstances, when it made a bad decision and issued a LIBNAME statement for a sequential-format-on-disk instead of normal disk. This could happen on z/OS only, only if the dataset had never been opened for output, and VGETOBS was issued against the DDNAME, AND, there are tape datasets allocated to the job, and this can only happen on the very first invoke of VGETOBS when it reads all LIBNAMEs to identify those on tape/sequential format. Now, VGETOBS recognizes that the DDNAME is not SAS or at least has not been opened and it does NOT issue a LIBNAME statement. Note: A SAS data library in sequential format on disk does not have a directory, so SAS must read all records from the start to find a dataset. Thanks to Paul Volpi, UHC, USA.
ADOCRMFV -A new generic RMF Monitor III table selection parameter ASMRMFV BASIC is supported. This is a convenient shorthand Mar 12, 2014 method to select RMF III tables of most common general Mar 14, 2014 interest and usage. There is no alias for BASIC. Mar 30, 2014 BASIC selects these RMF III tables: ASI CAT CPC CPU DVT ENC GEI OPD RCD -A new generic RMF Monitor III table selection parameter MOST is supported. This is a convenient shorthand method to select all possible tables of any practical value. There is no alias for MOST. MOST includes all the tables selected by BASIC and in addition selects these RMF III tables: CFI CPD CSR ENT SPG SVP. -NOTE: All other RMF III tables may still be selected or filtered even when BASIC or MOST is specified to achieve other useful table selection combinations. For example, these PARM values in JCL (or alternatively in the SYSIN DD file) are all valid: PARM='BASIC,CPD,CSR' Add CPD and CSR to BASIC set PARM='BASIC,NODVT' Exclude DVT from BASIC set PARM='MOST,NOSPG' Exclude SPG from MOST set -Recent information obtained indicates that the Invalid Resource bit may be set for Processor entries in the RED table in a MINTIME interval due to normal processor change events. So these are not true RED table errors. Operation is changed for RED Processor entries only so that when the RED table is selected and the Invalid Resource bit is on, then the entry is still skipped, but no error messages are now issued, and the Return Code is left unchanged. Prior to this change error messages were generated and Return Code 0016 would be set. Error handling for other types of RED entries is unchanged. -The behavior for the 'ALL' table selection parameter was changed in Change 32.037 so that it overrode any NOxxx table filter options regardless of input order. However, Yüklə 28,67 Mb. Dostları ilə paylaş:
|