usr.bin/perl/README.os390

This document is written in pod format hence there are punctuation
characters in odd places. You can read more
about pod in pod/perlpod.pod or the short summary in the INSTALL file.

=head1 NAME

perlos390 - building and installing Perl for z/OS (previously called OS/390)

=head1 SYNOPSIS

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

=head1 DESCRIPTION

This is a ported Perl for z/OS. It has been tested on z/OS 2.4 and
should work fine with z/OS 2.5.
It may work on other versions or releases, but those are
the ones it has been tested on.

The native character set for z/OS is EBCDIC, but it can also run in ASCII mode.
Perl can support either, but you have to compile it explicitly for one or the
other.  You could have both an ASCII perl, and an EBCDIC perl on the same
machine.  If you use ASCII mode and an ASCII perl, the Encode module shipped
with perl can be used to translate files from various EBCDIC code pages for
handling by perl, and then back on output

This document describes how to build a 64-bit Dynamic Perl, either ASCII or
EBCDIC.  You can interactively choose other configurations, as well as many
other options in the Configure script that is run as part of the build
process.  You may need to carry out some system configuration tasks before
running Configure, as detailed below.

=head2 Tools

You will want to get GNU make 4.1 or later. GNU make can be downloaded from a
port that Rocket Software provides.  You will need the z/OS c99 compiler from
IBM (though xlc in c99 mode without optimization turned on works in EBCDIC).

If you want the latest development version of Perl, you will need git.
You can use git on another platform and transfer the result via sftp or ftp to
z/OS.  But there is a z/OS native git client port available through Rocket
Software.

You may also need the gunzip client port that Rocket Software provides to unzip
any zipped tarball you upload to z/OS.

=head2 Building a 64-bit Dynamic ASCII Perl

For building from an official stable release of Perl, go to
L<https://www.perl.org/get.html> and choose any one of the
"Download latest stable source" buttons.  This will get you a tarball.  The
name of that tarball will be something like 'perl-V.R.M,tar,gz', where V.R.M is
the version/release/modification of the perl you are downloading. Do

  gunzip perl-V.R.M.tar.gz

Then one of:

  tar -xvf perl-V.R.M.tar

  pax -r -f perl-V.R.M.tar

Either of these will create the source directory.  You can rename it to
whatever you like; for these instructions, 'perl' is assumed to be the name.

If instead you want the latest unstable development release, using the native
git on z/OS, clone Perl:

  git clone https://github.com/Perl/perl5.git perl

Either way, once you have a 'perl' directory containing the source, cd into it,
and tag all the code as ASCII:

  cd perl
  chtag -R -h -t -cISO8859-1 *

Configure the build environment as 64-bit, Dynamic, ASCII, development,
deploying it to F</usr/local/perl/ascii>:

  export PATH=$PWD:$PATH
  export LIBPATH=$PWD:$PATH
  ./Configure -Dprefix=/usr/local/perl/ascii -des -Dusedevel \
        -Duse64bitall -Dusedl

If you are building from a stable source, you don't need "-Dusedevel".
(If you run Configure without options, it will interactively ask you about
every possible option based on its probing of what's available on your
particular machine, so you can choose as you go along.)

Run GNU make to build Perl

  make

Run tests to ensure Perl is working correctly. Currently, there are about a
dozen failing tests out of nearly 2500

  make test_harness

Install Perl into F</usr/local/perl/ascii>:

  make install

=head2 Building a 64-bit Dynamic EBCDIC Perl

You will need a working perl on some box with connectivity to the destination
machine.  On z/OS, it could be an ASCII perl, or a previous EBCDIC one.
Many machines will already have a pre-built perl already running, or one can
easily be downloaded from L<https://www.perl.org/get.html>.

Follow the directions above in "Building a 64-bit Dynamic ASCII Perl" as far as
getting a populated 'perl' directory.  Then come back here to proceed.

The downloaded perl will need to be converted to 1047 EBCDIC.  To do this:

  cd perl
  Porting/makerel -e

If the Porting/makerel step fails with an error that it can not issue the tar
command, proceed to issue the command interactively, where V.R.M is the
version/release/modification of Perl you are uploading:

  cd ../
  tar cf -  --format=ustar perl-V.R.M | gzip --best > perl-V.R.M.tar.gz

Use sftp to upload the zipped tar file to z/OS:

  sftp <your system>
  cd /tmp
  put perl-V.R.M.tar.gz

Unzip and untar the zipped tar file on z/OS:

  cd /tmp
  gunzip perl-V.R.M.tar.gz

Then one of:

  tar -xvf perl-V.R.M.tar

  pax -r -f perl-V.R.M.tar

You now have the source code for the EBCDIC Perl on z/OS and can proceed to
build it. This is analagous to how you would build the code for ASCII, but
note: you B<should not> tag the code but instead leave it untagged.

Configure the build environment as 64-bit, Dynamic, native, development,
deploying it to F</usr/local/perl/ebcdic>:

  export PATH=$PWD:$PATH
  export LIBPATH=$PWD:$PATH
  ./Configure -Dprefix=/usr/local/perl/ebcdic -des -Dusedevel \
        -Duse64bitall -Dusedl

If you are building from a stable source, you don't need "-Dusedevel".
(If you run Configure without options, it will interactively ask you about
every possible option based on its probing of what's available on your
particular machine, so you can choose as you go along.)

Run GNU make to build Perl

  make

Run tests to ensure Perl is working correctly.

  make test_harness

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

Install Perl into F</usr/local/perl/ebcdic>:

  make install

EBCDIC Perl is still a work in progress.  All the core code works as far as we
know, but various modules you might want to download from CPAN do not.  The
failures range from very minor to catastrophic.  Many of them are simply bugs
in the tests, with the module actually working properly.  This happens because,
for example, the test is coded to expect a certain character ASCII code point;
when it gets the EBCDIC value back instead, it complains.  But the code
actually worked.  Other potential failures that aren't really failures stem
from checksums coming out differently, since C<A>, for example, has a different
bit representation between the character sets.  A test that is expecting the
ASCII value will show failure, even if the module is working perfectly.  Also
in sorting, uppercase letters come before lowercase letters on ASCII systems;
the reverse on EBCDIC.

Some CPAN modules come bundled with the downloaded perl.  And a few of those
have yet to be fixed to pass on EBCDIC platforms.  As a result they are skipped
when you run 'make test'.  The current list is:

 Archive::Tar
 Config::Perl::V
 CPAN::Meta
 CPAN::Meta::YAML
 Digest::MD5
 Digest::SHA
 Encode
 ExtUtils::MakeMaker
 ExtUtils::Manifest
 HTTP::Tiny
 IO::Compress
 IPC::Cmd
 JSON::PP
 libnet
 MIME::Base64
 Module::Metadata
 PerlIO::via-QuotedPrint
 Pod::Checker
 podlators
 Pod::Simple
 Socket
 Test::Harness

See also F<hints/os390.sh> for other potential gotchas.

=head2 Setup and utilities for Perl on OS/390

This may also be a good time to ensure that your F</etc/protocol> file
and either your F</etc/resolv.conf> or F</etc/hosts> files are in place.
The IBM document that describes such USS system setup issues is
"z/OS UNIX System Services Planning"

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).

=head2 Useful files for trouble-shooting

If your configuration is failing, read hints/os390.sh
This file provides z/OS specific options to direct the build process.

=head3 Shell

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.

=head3 Dynamic loading

Dynamic loading is required if you want to use XS modules from CPAN (like
DBI (and DBD's), JSON::XS, and Text::CSV_XS) or update CORE modules from
CPAN with newer versions (like Encode) without rebuilding all of the perl
binary.

The instructions above will create a dynamic Perl. If you do not want to
use dynamic loading, remove the -Dusedl option.
See the comments in hints/os390.sh for more information on dynamic loading.

=head3 Optimizing

Optimization has not been turned on yet. There may be issues if Perl
is optimized.

=head2 Build Anomalies with Perl on OS/390

"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.

Within USS your F</etc/profile> or F<$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 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.

=head2 Testing Anomalies with Perl on OS/390

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:

=head3 Out of Memory (31-bit only)

Out of memory problems should not be an issue, unless you are attempting to build
a 31-bit Perl.

If you _are_ building a 31-bit Perl, the constrained environment may mean you
need to change memory options for Perl.
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 for 31-bit only
to set CEE run options, but the environment variable wins.

The 31-bit 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 C<sbrk()>
to get memory, and C<sbrk()> 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.

=head2 Usage Hints for Perl on z/OS

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

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.

=head2 Modules and Extensions for Perl on z/OS (Static Only)

Pure Perl (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
static linking you can still build XS based extensions for z/OS
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

=head2 Running Perl on z/OS

To run the 64-bit Dynamic Perl environment, update your PATH and LIBPATH
to include the location you installed Perl into, and then run the perl you
installed as perlV.R.M where V/R/M is the Version/Release/Modification level
of the current development level.
If you are running the ASCII/EBCDIC Bi-Modal Perl environment, you also need to
set up your ASCII/EBCDIC Bi-Modal environment variables, and ensure any Perl
source code you run is tagged appropriately as ASCII or EBCDIC using
"chtag -t -c<CCSID>":

=over

=item For ASCII Only:

 export _BPXK_AUTOCVT=ON
 export _CEE_RUNOPTS="FILETAG(AUTOCVT,AUTOTAG),POSIX(ON)"
 export _TAG_REDIR_ERR="txt"
 export _TAG_REDIR_IN="txt"
 export _TAG_REDIR_OUT="txt"

=item For ASCII or EBCDIC:

 export PATH=/usr/local/perl/ascii:$PATH
 export LIBPATH=/usr/local/perl/ascii/lib:$LIBPATH
 perlV.R.M args

=back

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

=head1 AUTHORS

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.

Mike Fulton and Karl Williamson have provided updates for UTF8, DLL, 64-bit and
ASCII/EBCDIC Bi-Modal support

=head1 OTHER SITES

L<https://github.com/ZOSOpenTools/perlport/> provides documentation and tools
for building various z/OS Perl configurations and has some useful tools in the
'bin' directory you may want to use for building z/OS Perl yourself.

=head1 HISTORY

Updated 24 December 2021 to enable initial ASCII support

Updated 03 October  2019 for perl-5.33.3+

Updated 28 November 2001 for broken URLs.

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

Updated 24 January  2001 to mention dynamic loading.

Updated 15 January  2001 for the 5.7.1 release of Perl.

Updated 12 November 2000 for the 5.7.1 release of Perl.

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

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

=cut