diff options
Diffstat (limited to 'cmsmvs/README.MVS.LE')
| -rw-r--r-- | cmsmvs/README.MVS.LE | 286 |
1 files changed, 286 insertions, 0 deletions
diff --git a/cmsmvs/README.MVS.LE b/cmsmvs/README.MVS.LE new file mode 100644 index 0000000..f7dcb8a --- /dev/null +++ b/cmsmvs/README.MVS.LE @@ -0,0 +1,286 @@ +Notes on Zip under MVS Language Environment (LE). + +First see README.MVS. This note describes just one beta test on OS/390 +V2R5 using IBM's C compiler (5647A01), V2R4. The major difference is +the use of LE on the beta site, together with some MVS native mode +fixes. Changes have not been tested on CMS. + +Some of the notes are to clarify things that were not clear from the +MANUAL or README.MVS. + +1. By default, IBM C generates the same csect name for each input + source. The prelink stage does not rename them and the linkage + editor throws away all but the first occurrence of each duplicate. + Oops, my code just disappeared :(. + + To get around this "feature", compile with the CSECT option to + force sensible names on the code and data sections of each csect. + The name of the static data csect defaults to the source name in + lower case, the code csect defaults to the source name in upper + case. These csect names still have to be unique, they cannot be + the same as function names. Of course, several csects have a + function which is the same name as the source in lower case, not + exactly an unusual occurrence. Therefore to make the csect name + unique, some of the sources have + + #ifdef MVS + # pragma csect(STATIC,xxxx_s) + #endif + + Where xxxx is an abbreviation of the source name. There has to be + a better way! + +2. The prelink step always gets cond code 4. It complains about + unresolved references, ignore it unless the linker also complains. + Prelink also complains about duplicate @@PPA2 sections and so does + the linker, but it seems to do no harm. Compile and link steps + should get 0, just prelink gets 4. See JCL at the bottom. + +3. Under MVS native mode (not Open Edition), tmpnam() returns a quoted + name of 5 qualifiers. The first is a HLQ chosen according to the + MVS LE algorithm (see below), the other qualifiers are time stamps. + If running on MVS and tmpnam() returns a quoted name with at leat + one '.', it is only safe to let the user change the high level + qualifier. Therefore -b insists on a single qualifier without '.' + in the MVS native environment. + +4. In Open Edition (OE) mode, the manual says that tmpnam() returns a + fully qualified name in directory TMPDIR or /tmp if TMPDIR is not + set. There is no point in zip trying to override that name so -b + is ignored in MVS OE mode (untested). The user should specify + environment variable TMPDIR instead. + +5. The MVS LE algorithm for choosing the high level qualifier for + native filenames is interesting, as in "May you live in interesting + times". The HLQ varies according to the environment the program is + running in, sometimes it is userid, sometimes it is TSO prefix. + See OS/390 C/C++ Programming Guide, Using a Data Set Name, + somewhere around section 2.9. + + If in doubt, use fully qualified and quoted names. Instead of + archive.zip, use 'prefix.archive.zip'. For input files, instead of + filename, use 'prefix.filename'. For PARM= in JCL, double up the + quotes. You even have to quote filenames in stdin. + +6. If your PARM includes any '/', make sure the PARM starts with '/'. + LE assumes anything before the first '/' is LE run time parameters. + It does no harm to always code a leading '/' for LE parms. + +7. JCL limits a PARM= to 100 characters total with approx. 65 on a + single line. Alas the syntax for continuing PARM= always embeds an + extra ',' somewhere in the parameters that the program finally + gets. No workaround, limit your PARM to a single line. With the + extra quotes around filenames, that does not leave much room. In + most cases, you will have to use '-@' to read the list of filenames + from SYSIN (stdin), it will not fit on a single PARM line. + +8. Filenames can be dataset names or you can refer to a ddname with + 'DD:name', case insensitive for external files, case sensitive for + OE files. You can even specify 'dd:name(mem)'. No wildcards, to + zip a complete pds you have to specify each member individually. + Directory recursion in OE does not appear to work at the moment. + +9. Zip attempts to map MVS filenames to Unix style names. It did not + work correctly for quoted names, fixed. Although you can pick up + an external (non-OE) file with a name using any case, be aware that + the mapping to a Unix style name faithfully follows the case you + supply. + +10. The archive file was being created with recfm=V and lrecl=32760. + 32760 is not valid for recfm=V under MVS, I originally changed it + to lrecl=32756. Then zip broke trying to fseek() over a record + boundary, I do not know whether this was a zip or LE bug. Trial + and error showed that recfm=U with byteseek seems to work on MVS. + No BDW or RDW, just a byte stream. The blocksize is always 6144. + + NOTE: This is an incompatible change from the previous beta, + archive files used to be recfm=V. That should not matter + because we just transfer the data, ignoring BDW and RDW + anyway. + +11. Zip used to complain about preallocated but empty archives, wrong + length records, no signature etc. The usual IBM/360 problem of no + end of file marker in a new, unopened dataset. Fixed, see routine + readzipfile in zipfile.c for the gory details. PARM= works fine. + +12. Several source files have records that are more than 80 bytes long. + It works if you transfer to mainframe datasets with a larger lrecl, + I used recfm=fb,lrecl=120 for the .C and .H files. To compile with + anything longer than 72 bytes, you need MVS C options NOMARGINS and + NOSEQUENCE (NOMAR,NOSEQ). + +13. cmsmvs was still using zname instead of name for open. Fixed. + +14. zip has to jump through a lot of hoops to see if an existing + zipfile actually contains data. A side effect of this is that + creating a zipfile with the RLSE parameter is a waste of time. + +Keith Owens <kaos@ocs.com.au>. Not a maintainer, just a beta tester. +Mon Sep 14 19:31:30 EST 1998 + + +Sample JCL to compile Zip under MVS LE. You might need a large region, +I used REGION=128M on the job card. Also watch the output lines, +75,000 with OPT(2), 100,000+ with OPT(2) replaced with DEF(DEBUG). You +need to allocate prefix.ZIP.C.OBJ (recfm=FB, lrecl=80) and +prefix.ZIP.LOAD (recfm=U, blksize is site defined). + +//CBC JCLLIB ORDER=CBC.SCBCPRC +//ZIP EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(ZIP)', +// OUTFILE='prefix.ZIP.C.OBJ(ZIP),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//CRYPT EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(CRYPT)', +// OUTFILE='prefix.ZIP.C.OBJ(CRYPT),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//TTYIO EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(TTYIO)', +// OUTFILE='prefix.ZIP.C.OBJ(TTYIO),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//TREES EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(TREES)', +// OUTFILE='prefix.ZIP.C.OBJ(TREES),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//DEFLATE EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(DEFLATE)', +// OUTFILE='prefix.ZIP.C.OBJ(DEFLATE),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//FILEIO EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(FILEIO)', +// OUTFILE='prefix.ZIP.C.OBJ(FILEIO),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//GLOBALS EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(GLOBALS)', +// OUTFILE='prefix.ZIP.C.OBJ(GLOBALS),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//UTIL EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(UTIL)', +// OUTFILE='prefix.ZIP.C.OBJ(UTIL),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//CRC32 EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(CRC32)', +// OUTFILE='prefix.ZIP.C.OBJ(CRC32),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//CRCTAB EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(CRCTAB)', +// OUTFILE='prefix.ZIP.C.OBJ(CRCTAB),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//ZIPFILE EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(ZIPFILE)', +// OUTFILE='prefix.ZIP.C.OBJ(ZIPFILE),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//ZIPUP EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(ZIPUP)', +// OUTFILE='prefix.ZIP.C.OBJ(ZIPUP),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//CMSMVS EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(CMSMVS)', +// OUTFILE='prefix.ZIP.C.OBJ(CMSMVS),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//MVS EXEC EDCC,COND=(0,NE),CREGSIZ='4M', +// INFILE='prefix.ZIP.C(MVS)', +// OUTFILE='prefix.ZIP.C.OBJ(MVS),DISP=SHR', +// CPARM='LONG,NOTERM,LIST,XREF,SOURCE', +// CPARM2='OPT(2),DEF(MVS),NOMAR,NOSEQ,CSECT' +//COMPILE.USERLIB DD DSN=prefix.ZIP.H,DISP=SHR +//PLINK EXEC PROC=EDCPL, +// OUTFILE='prefix.ZIP.LOAD(ZIP),DISP=SHR', +// PREGSIZ=6M, +// PPARM='NONCAL,MAP,MEMORY', +// LPARM='LIST,MAP,XREF' +//PLKED.SYSIN DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(ZIP) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(CRYPT) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(TREES) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(DEFLATE) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(FILEIO) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(GLOBALS) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(UTIL) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(CRC32) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(CRCTAB) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(ZIPFILE) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(ZIPUP) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(MVS) +// DD DISP=SHR,DSN=prefix.ZIP.C.OBJ(CMSMVS) +//LKED.SYSLIB DD DISP=SHR,DSN=CEE.SCEELKED +//SYSUT1 DD UNIT=SYSDA,SPACE=(CYL,(2,2)) +// + +Sample JCL to zip the mainframe .C and .H files as ASCII (-a). Delete +any existing archive first, point the temporary file at a particular +prefix (-b), use 'prefix.ARCHIVE.ZIP' for the archive file, read the +list of files to zip from stdin (SYSIN). + +//DELETE EXEC PGM=IDCAMS +//SYSPRINT DD SYSOUT=* +//SYSIN DD * + DELETE prefix.ARCHIVE.ZIP + SET MAXCC = 0 +//ZIP EXEC PGM=ZIP, +// PARM='/-a -v -b temppref ''prefix.ARCHIVE.ZIP'' -@' +//STEPLIB DD DSN=prefix.ZIP.LOAD,DISP=SHR +//SYSPRINT DD SYSOUT=* +//SYSOUT DD SYSOUT=* +//CEEDUMP DD SYSOUT=* +//ZIPC DD DISP=SHR,DSN=prefix.ZIP.C +//ZIPH DD DISP=SHR,DSN=prefix.ZIP.H +//SYSIN DD * +dd:zipc(api) +dd:zipc(cms) +dd:zipc(cmsmvs) +dd:zipc(crctab) +dd:zipc(crc32) +dd:zipc(crypt) +dd:zipc(deflate) +dd:zipc(fileio) +dd:zipc(globals) +dd:zipc(mktime) +dd:zipc(mvs) +dd:zipc(trees) +dd:zipc(ttyio) +dd:zipc(util) +dd:zipc(zip) +dd:zipc(zipcloak) +dd:zipc(zipfile) +dd:zipc(zipnote) +dd:zipc(zipsplit) +dd:zipc(zipup) +dd:ziph(api) +dd:ziph(cmsmvs) +dd:ziph(crypt) +dd:ziph(cstat) +dd:ziph(ebcdic) +dd:ziph(mvs) +dd:ziph(revision) +dd:ziph(stat) +dd:ziph(tailor) +dd:ziph(ttyio) +dd:ziph(zip) +dd:ziph(ziperr) +dd:ziph(zipup) |