README.os390 - building and installing Perl for OS/390 and z/OS

This document will help you Configure, build, test and install Perl on OS/390 (aka z/OS) Unix System Services.

This is a fully ported Perl for OS/390 Version 2 Release 3, 5, 6, 7, 8, and 9. It may work on other versions or releases, but those are the ones we've tested it on.

You may need to carry out some system configuration tasks before running the Configure script for Perl.

The z/OS Unix Tools and Toys list may prove helpful and contains links to ports of much of the software helpful for building Perl. http://www-1.ibm.com/servers/eserver/zseries/zos/unix/bpxa1toy.html

If using ftp remember to transfer the distribution in binary format.

Gunzip/gzip for OS/390 is discussed at:

  http://www-1.ibm.com/servers/eserver/zseries/zos/unix/faq/bpxqp1.html

to extract an ASCII tar archive on OS/390, try this:

   pax -o to=IBM-1047,from=ISO8859-1 -r < latest.tar

or

   zcat latest.tar.Z | pax -o to=IBM-1047,from=ISO8859-1 -r

If you get lots of errors of the form

  tar: FSUM7171 ...: cannot set uid/gid: EDC5139I Operation not permitted.

you didn't read the above and tried to use tar instead of pax, you'll first have to remove the (now corrupt) perl directory

   rm -rf perl-...

and then use pax.

Be sure that your yacc installation is in place including any necessary parser template files. If you have not already done so then be sure to:

  cp /samples/yyparse.c /etc

This may also be a good time to ensure that your /etc/protocol file and either your /etc/resolv.conf or /etc/hosts files are in place. The IBM document that described such USS system setup issues was SC28-1890-07 OS/390 UNIX System Services Planning, in particular Chapter 6 on customizing the OE shell.

GNU make for OS/390, which is recommended for the build of perl (as well as building CPAN modules and extensions), is available from the Tools.

Some people have reported encountering Out of memory! errors while trying to build Perl using GNU make binaries. If you encounter such trouble then try to download the source code kit and build GNU make from source to eliminate any such trouble. You might also find GNU make (as well as Perl and Apache) in the red-piece/book Open Source Software for OS/390 UNIX, SG24-5944-00 from IBM.

If instead of the recommended GNU make you would like to use the system supplied make program then be sure to install the default rules file properly via the shell command:

    cp /samples/startup.mk /etc

and be sure to also set the environment variable _C89_CCMODE=1 (exporting _C89_CCMODE=1 is also a good idea for users of GNU make).

You might also want to have GNU groff for OS/390 installed before running the `make install` step for Perl.

There is a syntax error in the /usr/include/sys/socket.h header file that IBM supplies with USS V2R7, V2R8, and possibly V2R9. The problem with the header file is that near the definition of the SO_REUSEPORT constant there is a spurious extra '/' character outside of a comment like so:

 #define SO_REUSEPORT    0x0200    /* allow local address & port
                                      reuse */                    /

You could edit that header yourself to remove that last '/', or you might note that Language Environment (LE) APAR PQ39997 describes the problem and PTF's UQ46272 and UQ46271 are the (R8 at least) fixes and apply them. If left unattended that syntax error will turn up as an inability for Perl to build its Socket extension.

For successful testing you may need to turn on the sticky bit for your world readable /tmp directory if you have not already done so (see man chmod).

Once you've unpacked the distribution, run sh Configure (see INSTALL for a full discussion of the Configure options). There is a hints file for os390 that specifies the correct values for most things. Some things to watch out for include:

•

A message of the form:

 (I see you are using the Korn shell.  Some ksh's blow up on Configure,
 mainly on older exotic systems.  If yours does, try the Bourne shell instead.)

is nothing to worry about at all.

•

Some of the parser default template files in /samples are needed in /etc. In particular be sure that you at least copy /samples/yyparse.c to /etc before running Perl's Configure. This step ensures successful extraction of EBCDIC versions of parser files such as perly.c, perly.h, and x2p/a2p.c. This has to be done before running Configure the first time. If you failed to do so then the easiest way to re-Configure Perl is to delete your misconfigured build root and re-extract the source from the tar ball. Then you must ensure that /etc/yyparse.c is properly in place before attempting to re-run Configure.

•

This port will support dynamic loading, but it is not selected by default. If you would like to experiment with dynamic loading then be sure to specify -Dusedl in the arguments to the Configure script. See the comments in hints/os390.sh for more information on dynamic loading. If you build with dynamic loading then you will need to add the CW$archlibexp/CORE directory to your LIBPATH environment variable in order for perl to work. See the config.sh file for the value of CW$archlibexp. If in trying to use Perl you see an error message similar to:

 CEE3501S The module libperl.dll was not found.
         From entry point __dllstaticinit at compile unit offset +00000194 at

then your LIBPATH does not have the location of libperl.x and either libperl.dll or libperl.so in it. Add that directory to your LIBPATH and proceed.

•

Do not turn on the compiler optimization flag -O. There is a bug in either the optimizer or perl that causes perl to not work correctly when the optimizer is on.

•

Some of the configuration files in /etc used by the networking APIs are either missing or have the wrong names. In particular, make sure that there's either an /etc/resolv.conf or an /etc/hosts, so that gethostbyname() works, and make sure that the file /etc/proto has been renamed to /etc/protocol (NOT /etc/protocols, as used by other Unix systems). You may have to look for things like HOSTNAME and DOMAINORIGIN in the //'SYS1.TCPPARMS(TCPDATA)' PDS member in order to properly set up your /etc networking files.

Simply put:

    sh Configure
    make
    make test

if everything looks ok (see the next section for test/IVP diagnosis) then:

    make install

this last step may or may not require UID=0 privileges depending on how you answered the questions that Configure asked and whether or not you have write access to the directories you specified.

Out of memory! messages during the build of Perl are most often fixed by re building the GNU make utility for OS/390 from a source code kit.

Another memory limiting item to check is your MAXASSIZE parameter in your 'SYS1.PARMLIB(BPXPRMxx)' data set (note too that as of V2R8 address space limits can be set on a per user ID basis in the USS segment of a RACF profile). People have reported successful builds of Perl with MAXASSIZE parameters as small as 503316480 (and it may be possible to build Perl with a MAXASSIZE smaller than that).

Within USS your /etc/profile or CW$HOME/.profile may limit your ulimit settings. Check that the following command returns reasonable values:

    ulimit -a

To conserve memory you should have your compiler modules loaded into the Link Pack Area (LPA/ELPA) rather than in a link list or step lib.

If the c89 compiler complains of syntax errors during the build of the Socket extension then be sure to fix the syntax error in the system header /usr/include/sys/socket.h.

The `make test` step runs a Perl Verification Procedure, usually before installation. You might encounter STDERR messages even during a successful run of `make test`. Here is a guide to some of the more commonly seen anomalies:

•

A message of the form:

 comp/cpp.............ERROR CBC3191 ./.301989890.c:1     The character $ is not a
  valid C source character.
 FSUM3065 The COMPILE step ended with return code 12.
 FSUM3017 Could not compile .301989890.c. Correct the errors and try again.
 ok

indicates that the t/comp/cpp.t test of Perl's -P command line switch has passed but that the particular invocation of c89 -E in the cpp script does not suppress the C compiler check of source code validity.

•

A message of the form:

 io/openpid...........CEE5210S The signal SIGHUP was received.
 CEE5210S The signal SIGHUP was received.
 CEE5210S The signal SIGHUP was received.
 ok

indicates that the t/io/openpid.t test of Perl has passed but done so with extraneous messages on stderr from CEE.

•

A message of the form:

 lib/ftmp-security....File::Temp::_gettemp: Parent directory (/tmp/) is not safe
 (sticky bit not set when world writable?) at lib/ftmp-security.t line 100
 File::Temp::_gettemp: Parent directory (/tmp/) is not safe (sticky bit not
 set when world writable?) at lib/ftmp-security.t line 100
 ok

indicates a problem with the permissions on your /tmp directory within the HFS. To correct that problem issue the command:

     chmod a+t /tmp

from an account with write access to the directory entry for /tmp.

•

Out of Memory! Recent perl test suite is quite memory hunrgy. In addition to the comments above on memory limitations it is also worth checking for _CEE_RUNOPTS in your environment. Perl now has (in miniperlmain.c) a C #pragma to set CEE run options, but the environment variable wins. The C code asks for:

 #pragma runopts(HEAP(2M,500K,ANYWHERE,KEEP,8K,4K) STACK(,,ANY,) ALL31(ON))

The important parts of that are the second argument (the increment) to HEAP, and allowing the stack to be Above the (16M) line. If the heap increment is too small then when perl (for example loading unicode/Name.pl) tries to create a big (400K+) string it cannot fit in a single segment and you get Out of Memory! - even if there is still plenty of memory available. A related issue is use with perl's malloc. Perl's malloc uses CWsbrk() to get memory, and CWsbrk() is limited to the first allocation so in this case something like:

  HEAP(8M,500K,ANYWHERE,KEEP,8K,4K)

is needed to get through the test suite.

The installman script will try to run on OS/390. There will be fewer errors if you have a roff utility installed. You can obtain GNU groff from the Redbook SG24-5944-00 ftp site.

When using perl on OS/390 please keep in mind that the EBCDIC and ASCII character sets are different. See perlebcdic.pod for more on such character set issues. Perl builtin functions that may behave differently under EBCDIC are also mentioned in the perlport.pod document.

Open Edition (UNIX System Services) from V2R8 onward does support #!/path/to/perl script invocation. There is a PTF available from IBM for V2R7 that will allow shell/kernel support for #!. USS releases prior to V2R7 did not support the #! means of script invocation. If you are running V2R6 or earlier then see:

    head `whence perldoc`

for an example of how to use the eval exec trick to ask the shell to have Perl run your scripts on those older releases of Unix System Services.

If you are having trouble with square brackets then consider switching your rlogin or telnet client. Try to avoid older 3270 emulators and ISHELL for working with Perl on USS.

There appears to be a bug in the floating point implementation on S/390 systems such that calling int() on the product of a number and a small magnitude number is not the same as calling int() on the quotient of that number and a large magnitude number. For example, in the following Perl code:

    my $x = 100000.0;
    my $y = int($x * 1e-5) * 1e5; # '0'
    my $z = int($x / 1e+5) * 1e5;  # '100000'
    print "\$y is $y and \$z is $z\n"; # $y is 0 and $z is 100000

Although one would expect the quantities CW$y and CW$z to be the same and equal to 100000 they will differ and instead will be 0 and 100000 respectively.

The problem can be further examined in a roughly equivalent C program:

    #include <stdio.h>
    #include <math.h>
    main()
    {
    double r1,r2;
    double x = 100000.0;
    double y = 0.0;
    double z = 0.0;
    x = 100000.0 * 1e-5;
    r1 = modf (x,&y);
    x = 100000.0 / 1e+5;
    r2 = modf (x,&z);
    printf("y is %e and z is %e\n",y*1e5,z*1e5);
    /* y is 0.000000e+00 and z is 1.000000e+05 (with c89) */
    }

Pure pure (that is non xs) modules may be installed via the usual:

    perl Makefile.PL
    make
    make test
    make install

If you built perl with dynamic loading capability then that would also be the way to build xs based extensions. However, if you built perl with the default static linking you can still build xs based extensions for OS/390 but you will need to follow the instructions in ExtUtils::MakeMaker for building statically linked perl binaries. In the simplest configurations building a static perl + xs extension boils down to:

    perl Makefile.PL
    make
    make perl
    make test
    make install
    make -f Makefile.aperl inst_perl MAP_TARGET=perl

In most cases people have reported better results with GNU make rather than the system's /bin/make program, whether for plain modules or for xs based extensions.

If the make process encounters trouble with either compilation or linking then try setting the _C89_CCMODE to 1. Assuming sh is your login shell then run:

    export _C89_CCMODE=1

If tcsh is your login shell then use the setenv command.

David Fiander and Peter Prymmer with thanks to Dennis Longnecker and William Raffloer for valuable reports, LPAR and PTF feedback. Thanks to Mike MacIsaac and Egon Terwedow for SG24-5944-00. Thanks to Ignasi Roca for pointing out the floating point problems. Thanks to John Goodyear for dynamic loading help.

INSTALL, perlport, perlebcdic, ExtUtils::MakeMaker.

    http://www-1.ibm.com/servers/eserver/zseries/zos/unix/bpxa1toy.html

    http://www.redbooks.ibm.com/abstracts/sg245944.html

    http://www-1.ibm.com/servers/eserver/zseries/zos/unix/bpxa1ty1.html#opensrc

    http://www.xray.mpe.mpg.de/mailing-lists/perl-mvs/

    http://publibz.boulder.ibm.com:80/cgi-bin/bookmgr_OS390/BOOKS/ceea3030/

    http://publibz.boulder.ibm.com:80/cgi-bin/bookmgr_OS390/BOOKS/CBCUG030/

If you are interested in the VM/ESA, z/OS (formerly known as OS/390) and POSIX-BC (BS2000) ports of Perl then see the perl-mvs mailing list. To subscribe, send an empty message to perl-mvs-subscribe@perl.org.

See also:

    http://lists.perl.org/showlist.cgi?name=perl-mvs

There are web archives of the mailing list at:

    http://www.xray.mpe.mpg.de/mailing-lists/perl-mvs/
    http://archive.develooper.com/perl-mvs@perl.org/

This document was originally written by David Fiander for the 5.005 release of Perl.

This document was podified for the 5.005_03 release of Perl 11 March 1999.

Updated 28 November 2001 for broken URLs.

Updated 12 November 2000 for the 5.7.1 release of Perl.

Updated 15 January 2001 for the 5.7.1 release of Perl.

Updated 24 January 2001 to mention dynamic loading.

Updated 12 March 2001 to mention //'SYS1.TCPPARMS(TCPDATA)'.