path: root/README.vms

# vim: syntax=pod

If you read this file _as_is_, just ignore the equal signs on the left.
This file is written in the POD format (see [.pod]perlpod.pod) which is
specially designed to be readable as is.

=head1 NAME

perlvms - Configuring, building, testing, and installing perl on VMS

=head1 SYNOPSIS

To configure, build, test, and install perl on VMS:

    @configure
    mmk
    mmk test
    mmk install

=head1 DESCRIPTION

=head2 Important safety tip

For best results, make sure you read the "Configuring the Perl Build",
"Building  Perl", and "Installing Perl" sections of this document before
you build or install.  Also please note other changes in the current
release by having a look at L<perldelta/VMS>.

=head2 Introduction to Perl on VMS

The VMS port of Perl is as functionally complete as any other Perl port
(and as complete as the ports on some Unix systems). The Perl binaries
provide all the Perl system calls that are either available under VMS or
reasonably emulated. There are some incompatibilities in process handling
(e.g. the fork/exec model for creating subprocesses doesn't do what you
might expect under Unix), mainly because VMS and Unix handle processes and
sub-processes very differently.

There are still some unimplemented system functions, and of course we
could use modules implementing useful VMS system services, so if you'd like
to lend a hand we'd love to have you.  Join the Perl Porting Team Now!

=head2 Other required software for Compiling Perl on VMS

In addition to VMS and DCL you will need three things:

=over 4

=item 1  A C compiler. 

VSI (formerly DEC/Compaq/HP/HPE) C for VMS (Alpha or Itanium). Various
ancient versions of DEC C had some caveats, so if you're using a version
older than 7.x, you may need to upgrade to get a successful build.

There have been no recent reports of builds using Gnu C, but latent
(and most likely outdated) support for it is still present in various
parts of the sources.

There is rudimentary but not quite complete support for HP C++; to try it out,
configure with C<-"Dusecxx" -"Duser_c_flags=/WARN=INFORMATIONAL=NOCTOBUTCONREFM">.

=item 2  A make tool. 

You will need the free MMS analog MMK (available from
L<http://ftp.endlesssoftware.com.au/mmk/kits/> or 
L<https://github.com/endlesssoftware/mmk>). HP's MMS has not been known to work for
some time as Perl's automatically-generated description files are too complex for it,
but MMS support may return in the future.  Gnu Make might work, but it's been so long
since anyone's tested it that we're not sure.

=item 3  ODS-5 and Extended Parse

All development and testing of Perl on VMS takes place on ODS-5 volumes with
extended parse enabled in the environment via the command C<SET PROCESS/PARSE=EXTENDED>.
Latent support for ODS-2 volumes is still present, but there have been some reports
that it no longer works, and even if it builds, there will be many test failures,
mostly related to the failure to preserve filename case. ODS-2 support may be
explicity disabled in a future release.

=back

=head2 Additional software that is optional for Perl on VMS

You may also want to have on hand:

=over 4

=item 1  gunzip/gzip for VMS 

A de-compressor for *.gz and *.tgz files available from a number 
of web/ftp sites such as:

    L<http://www.antinode.info/dec/sw/gzip.html>
    L<http://vms.process.com/scripts/fileserv/fileserv.com?GZIP>

=item 2  VMS tar 

For reading and writing Unix tape archives (*.tar files).  Vmstar is also 
available from a number of sites such as:

    L<http://www.antinode.info/dec/sw/vmstar.html>
    L<http://vms.process.com/scripts/fileserv/fileserv.com?VMSTAR>

A port of GNU tar is also available as part of the GNV package:

    L<http://h71000.www7.hp.com/opensource/gnv.html>

=item 3  unzip for VMS

A combination decompressor and archive reader/writer for *.zip files.  
Unzip is available from a number of web/ftp sites.

    L<http://www.info-zip.org/UnZip.html>
    L<http://www.hp.com/go/openvms/freeware/>
    L<http://vms.process.com/fileserv-software.html>

=item 5 GNU patch and diffutils for VMS

Patches to Perl are usually distributed as GNU unified or contextual diffs. 
Such patches are created by the GNU diff program (part of the diffutils
distribution) and applied with GNU patch.  VMS ports of these utilities are
available here:

    L<http://www.antinode.info/dec/sw/diffutils.html>
    L<http://vms.pdv-systeme.de/users/martinv/gnupatch.zip>

=back

Please note that unzip and gunzip are not the same thing (they work with
different formats).  Many of the useful files from CPAN (the Comprehensive
Perl Archive Network) are in *.tar.gz or *.tgz format (this includes copies 
of the source code for perl as well as modules and scripts that you may 
wish to add later) hence you probably want to have GUNZIP.EXE and 
VMSTAR.EXE on your VMS machine.

=head1 Unpacking the Perl source code

You may need to set up a foreign symbol for the unpacking utility of
choice.  Once you have done so, use a command like the following to
unpack the archive:

    vmstar -xvf perl-5^.37^.12.tar

Then set default to the top-level source directory like so:

    set default [.perl-5^.37^.12]

and proceed with configuration as described in the next section.


=head1 Configuring the Perl build

To configure perl (a necessary first step), issue the command

   @configure.com

from the top of an unpacked perl source directory.  You will be asked a 
series of questions, and the answers to them (along with the capabilities 
of your C compiler and network stack) will determine how perl is custom-
built for your machine.

If you have any symbols or logical names in your environment that may 
interfere with the build or regression testing of perl then F<configure.com> 
will try to warn you about them.  If a logical name is causing
you trouble but is in an LNM table that you do not have write access to
then try defining your own to a harmless equivalence string in a table 
such that it is resolved before the other (e.g. if TMP is defined in the
SYSTEM table then try DEFINE TMP "NL:" or somesuch in your process table) 
otherwise simply deassign the dangerous logical names.  The potentially 
troublesome logicals and symbols include:

    COMP    "LOGICAL"
    EXT     "LOGICAL"
    FOO     "LOGICAL"
    LIB     "LOGICAL"
    LIST    "LOGICAL"
    MIME    "LOGICAL"
    POSIX   "LOGICAL"
    SYS     "LOGICAL"
    T       "LOGICAL"
    THREAD  "LOGICAL"
    THREADS "LOGICAL"
    TIME    "LOGICAL"
    TMP     "LOGICAL"
    UNICODE "LOGICAL"
    UTIL    "LOGICAL"
    TEST    "SYMBOL"

As a handy shortcut, the command:

    @configure "-des"

(note the quotation marks and case) will choose reasonable defaults 
automatically.  Some options can be given explicitly on the command line;
the following example specifies a non-default location for where Perl
will be installed:

    @configure "-d" "-Dprefix=dka100:[utils.perl5.]"

Note that the installation location would be by default where you unpacked 
the source with a "_ROOT." appended.  For example if you unpacked the perl 
source into:

   F<DKA200:[PERL-5^.18^.0...]>

Then the F<PERL_SETUP.COM> that gets written out by F<configure.com> will
try to DEFINE your installation PERL_ROOT to be:

   F<DKA200:[PERL-5^.18^.0_ROOT.]>

More help with configure.com is available from:

    @configure "-h"

If you find yourself reconfiguring and rebuilding  then be sure to also follow
the advice in the "Cleaning up and starting fresh (optional)" and the checklist
of items in the "CAVEATS" sections below.

=head2 Changing compile-time options (optional) for Perl on VMS

Most of the user-definable features of Perl are enabled or disabled in
configure.com, which processes the hints file config_h.SH.  There is
code in there to Do The Right Thing, but that  may end up being the
wrong thing for you.  Make sure you understand what you are doing since
inappropriate changes to configure.com or config_h.SH can render perl 
unbuildable; odds are that there's nothing in there you'll need to
change. Note also that non-default options are tested less than default
options, so you may end up being more of a pioneer than you intend to be.

=head1 Building Perl

The configuration script will print out, at the very end, the MMS or MMK
command you need to compile perl.  Issue it (exactly as printed) to start
the build.  

Once you issue your MMS or MMK command, sit back and wait.  Perl should 
compile and link without a problem.  If a problem does occur check the 
"CAVEATS" section of this document.  If that does not help send some 
mail to the VMSPERL mailing list.  Instructions are in the L</"Mailing Lists"> 
section of this document.

=head1 Testing Perl

Once Perl has built cleanly you need to test it to make sure things work.
This step is very important since there are always things that can go wrong
somehow and yield a dysfunctional Perl for you.

Testing is very easy, though, as there's a full test suite in the perl
distribution.  To run the tests, enter the I<exact> MMS line you used to
compile Perl and add the word "test" to the end, like this:

If the compile command was:

    MMK

then the test command ought to be:

    MMK test

MMK (or MMS) will run all the tests.  This may take some time, as there are 
a lot of tests.  If any tests fail, there will be a note made on-screen. 
At the end of all the tests, a summary of the tests, the number passed and 
failed, and the time taken will be displayed.

The test driver invoked via MMK TEST has a DCL wrapper ([.VMS]TEST.COM) that
downgrades privileges to NETMBX, TMPMBX for the duration of the test run,
and then restores them to their prior state upon completion of testing. 
This is done to ensure that the tests run in a private sandbox and can do no
harm to your system even in the unlikely event something goes badly wrong in
one of the test scripts while running the tests from a privileged account. 
A side effect of this safety precaution is that the account used to run the
test suite must be the owner of the directory tree in which Perl has been
built; otherwise the manipulations of temporary files and directories
attempted by some of the tests will fail.

If any tests fail, it means something is wrong with Perl, or at least
with the particular module or feature that reported failure. If the test suite
hangs (some tests can take upwards of two or three minutes, or more if
you're on an especially slow machine, depending on your machine speed, so
don't be hasty), then the test I<after> the last one displayed failed. Don't
install Perl unless you're confident that you're OK. Regardless of how
confident you are, make a bug report to the VMSPerl mailing list.

If one or more tests fail, you can get more information on the failure by 
issuing this command sequence:

    @[.vms]test .typ "" "-v" [.subdir]test.t

where ".typ" is the file type of the Perl images you just built (if you
didn't do anything special, use .EXE), and "[.subdir]test.t" is the test
that failed. For example, with a normal Perl build, if the test indicated
that t/op/time failed, then you'd do this:

    @ .vms]test .EXE "" "-v" [.op]time.t

Note that test names are reported in UNIX syntax and relative to the
top-level build directory.  When supplying them individually to the test
driver, you must specify them in Unix format if they are outside of the [.t]
directory; otherwise VMS syntax is ok. Note that you must also give the path
relative to the [.t] directory and you must also add the .t extension to the
filename.  So, for example if the test lib/warnings.t fails, you would run:

    @[.vms]test .EXE "" -"v" "../lib/warnings.t"

When you send in a bug report for failed tests, please include the output
from this command, which is run from the main source directory:

    MCR []MINIPERL "-Ilib" "-V"

Note that -"V" really is a capital V in double quotes. This will dump out a
couple of screens worth of configuration information, and can help us 
diagnose the problem.  If (and only if) that did not work then try enclosing 
the output of:

    MMK printconfig

If (and only if) that did not work then try enclosing the output of:

    @[.vms]myconfig

You may also be asked to provide your C compiler version ("CC/VERSION NL:" 
with DEC C, "gcc --version" with GNU CC).  To obtain the version of MMS or 
MMK you are running try "MMS/ident" or "MMK /ident".  The GNU make version 
can be identified with "make --version".

=head2 Cleaning up and starting fresh (optional) installing Perl on VMS

If you need to recompile from scratch, you have to make sure you clean up
first.  There is a procedure to do it--enter the I<exact> MMK line you used 
to compile and add "realclean" at the end, like this:

if the compile command was:

    MMK

then the cleanup command ought to be:

    MMK realclean

If you do not do this things may behave erratically during the subsequent 
rebuild attempt.  They might not, too, so it is best to be sure and do it.

=head1 Installing Perl

There are several steps you need to take to get Perl installed and
running.

=over 4

=item 1

Check your default file protections with

     SHOW PROTECTION /DEFAULT

and adjust if necessary with C<SET PROTECTION=(code)/DEFAULT>.

=item 2

Decide where you want Perl to be installed (unless you have already done so
by using the "prefix" configuration parameter -- see the example in the
"Configuring the Perl build" section).

The DCL script PERL_SETUP.COM that is written by configure.com will help you
with the definition of the PERL_ROOT and PERLSHR logical names and the PERL
foreign command  symbol.  Take a look at PERL_SETUP.COM and modify it if you
want to.  The installation process will execute PERL_SETUP.COM and copy
files to the directory tree pointed to by the PERL_ROOT logical name defined
there, so make sure that you have write access to the parent directory of
what will become the root of your Perl installation.

=item 3

Run the install script via:

    MMK install

If for some reason it complains about target INSTALL being up to date,
throw a /FORCE switch on the MMS or MMK command.

=back

Installation will copy F<PERL_SETUP.COM> to the root of your installation
tree.  If you want to give everyone on the system  access to Perl (and you
have, for example, installed to F<dsa0:[utils.perl_root]>) then add a line
that reads:

    $ @dsa0:[utils.perl_root]perl_setup

to F<SYS$MANAGER:SYLOGIN.COM>.  Or for your own use only, simply place
that line in F<SYS$LOGIN:LOGIN.COM>.

Two alternatives to the foreign symbol would be to install PERL into 
DCLTABLES.EXE (Check out the section "Installing Perl into DCLTABLES 
(optional)" for more information), or put the image in a 
directory that's in your DCL$PATH.

See also the "INSTALLing images (optional)" section.

=head2 Installing Perl into DCLTABLES (optional) on VMS

Execute the following command file to define PERL as a DCL command.
You'll need CMKRNL privilege to install the new dcltables.exe.

    $ create perl.cld
    !
    ! modify to reflect location of your perl.exe
    !
    define verb perl
      image perl_root:[000000]perl.exe
      cliflags (foreign)
    $!
    $ set command perl /table=sys$common:[syslib]dcltables.exe -
     /output=sys$common:[syslib]dcltables.exe
    $ install replace sys$common:[syslib]dcltables.exe
    $ exit

=head2 INSTALLing Perl images (optional) on VMS

On systems that are using perl quite a bit, and particularly those with 
minimal RAM, you can boost the performance of perl by INSTALLing it as
a known image.  PERLSHR.EXE is typically larger than 3000 blocks
and that is a reasonably large amount of IO to load each time perl is 
invoked. 

   INSTALL ADD PERLSHR/SHARE
   INSTALL ADD PERL/HEADER

should be enough for F<PERLSHR.EXE> (/share implies /header and /open), 
while /HEADER should do for FPERL.EXE> (perl.exe is not a shared image).

If your code 'use's modules, check to see if there is a shareable image for
them, too.  In the base perl build, POSIX, IO, Fcntl, Opcode, SDBM_File,
DCLsym, and Stdio, and other extensions all have shared images that can be
installed /SHARE.

How much of a win depends on your memory situation, but if you are firing
off perl with any regularity (like more than once every 20 seconds or so)
it is probably beneficial to INSTALL at least portions of perl.

While there is code in perl to remove privileges as it runs you are advised
to NOT INSTALL F<PERL.EXE> with PRIVs!

=head2 Running h2ph to create perl header files (optional) on VMS

If using HP C, ensure that you have extracted loose versions of your 
compiler's header or *.H files.  Be sure to check the contents of:

    SYS$LIBRARY:DECC$RTLDEF.TLB
    SYS$LIBRARY:SYS$LIB_C.TLB
    SYS$LIBRARY:SYS$STARLET_C.TLB

etcetera.

If using GNU cc then also check your GNU_CC:[000000...] tree for the locations
of the GNU cc headers.

=head1 Reporting Bugs

If you come across what you think might be a bug in Perl, please report
it. The issue tracker at L<https://github.com/Perl/perl5/issues> walks you
through the process of creating a bug report and including details of your
installation.

=head1 CAVEATS

Probably the single biggest gotcha in compiling Perl is giving the wrong
switches to MMS/MMK when you build. Use I<exactly> what the configure.com 
script prints!

Be sure that the process that you use to build Perl has a PGFLQUO of at
least 400000.  Be sure to have a correct local time zone to UTC offset
defined (in seconds) in the logical name SYS$TIMEZONE_DIFFERENTIAL before
running the regression test suite.  The SYS$MANAGER:UTC$CONFIGURE_TDF.COM 
procedure will help you set that logical for your system but may require 
system privileges.  For example, a location 5 hours west of UTC (such as 
the US East coast while not on daylight savings time) would have:

    DEFINE SYS$TIMEZONE_DIFFERENTIAL "-18000"

A final thing that causes trouble is leftover pieces from a failed
build.  If things go wrong make sure you do a "(MMK|MMS|make) realclean"
before you rebuild.

=head2 Floating Point Considerations

Prior to 5.8.0, Perl simply accepted the default floating point options of the
C compiler, namely representing doubles with G_FLOAT on Alpha.  Single
precision floating point values are represented in F_FLOAT format when either
D_FLOAT or G_FLOAT is in use for doubles.  Beginning with 5.8.0, Alpha builds
now use IEEE floating point formats by default, which in VMS parlance are S_FLOAT
for singles and T_FLOAT for doubles.  Itanium builds have always used IEEE by
default. The  available non-default options are D_FLOAT or G_FLOAT on Alpha
or Itanium.

The use of IEEE introduces NaN, infinity, and denormalization capabilities not
available with D_FLOAT and G_FLOAT.  When using one of those non-IEEE formats,
silent underflow and overflow are emulated in the conversion of strings to
numbers, but it is preferable to get the real thing by using IEEE where possible.
You are likely to see quite a few test failures when not using IEEE floating point.

Regardless of what floating point format you consider preferable, be aware
that the choice may have an impact on compatibility with external libraries,
such as database interfaces, and with existing data, such as data created with
the C<pack> function and written to disk, or data stored via the Storable
extension.  For example, a C<pack("d", $foo)")> will create a D_FLOAT,
G_FLOAT, or T_FLOAT depending on what your Perl was configured with.  When
written to disk, the value can only be retrieved later by a Perl configured
with the same floating point option that was in effect when it was created.

To obtain a non-IEEE build, simply answer no to the "Use IEEE math?" question
during the configuration or specify -"Uuseieee" as a parameter to configure.com
on the command line.

=head1 Mailing Lists

There are several mailing lists available to the Perl porter.  For VMS
specific issues (including both Perl questions and installation problems)
there is the VMSPERL mailing list.  It is usually a low-volume (10-12
messages a week) mailing list.

To subscribe, send a mail message to VMSPERL-SUBSCRIBE@PERL.ORG. The VMSPERL
mailing list address is VMSPERL@PERL.ORG.  Any mail sent there gets echoed
to all subscribers of the list.  There is an archive of the list
on the web at:

    L<https://www.nntp.perl.org/group/perl.vmsperl/>

To unsubscribe from VMSPERL send a message to VMSPERL-UNSUBSCRIBE@PERL.ORG.
Be sure to do so from the subscribed account that you are canceling.

=head2 Web sites for Perl on VMS

Vmsperl pages on the web include:

    L<http://www.sidhe.org/vmsperl/index.html>
    L<https://www.cpan.org/modules/by-module/VMS/>
    L<https://www.nntp.perl.org/group/perl.vmsperl/>
    L<https://sourceforge.net/projects/vmsperlkit/>

=head1 SEE ALSO

Perl information for users and programmers about the port of perl to VMS is
available from the [.pod]perlvms.pod file that gets installed as L<perlvms>.
For administrators the perlvms document also includes a detailed discussion 
of extending vmsperl with CPAN modules after Perl has been installed.

=head1 AUTHORS

Originally by Charles Bailey bailey@newman.upenn.edu.  See the git repository
for history.

=head1 ACKNOWLEDGEMENTS

A real big thanks needs to go to Charles Bailey
bailey@newman.upenn.edu, who is ultimately responsible for Perl 5.004
running on VMS. Without him, nothing the rest of us have done would be at
all important.

There are, of course, far too many people involved in the porting and testing
of Perl to mention everyone who deserves it, so please forgive us if we've
missed someone.  That said, special thanks are due to the following:

  Tim Adye T.J.Adye@rl.ac.uk
     for the VMS emulations of getpw*()
  David Denholm denholm@conmat.phys.soton.ac.uk
     for extensive testing and provision of pipe and SocketShr code,
  Mark Pizzolato mark@infocomm.com
     for the getredirection() code
  Rich Salz rsalz@bbn.com
     for readdir() and related routines
  Peter Prymmer pvhp@best.com 
     for extensive testing, as well as development work on
     configuration and documentation for VMS Perl,
  Dan Sugalski dan@sidhe.org
     for extensive contributions to recent version support,
     development of VMS-specific extensions, and dissemination
     of information about VMS Perl,
  the Stanford Synchrotron Radiation Laboratory and the
     Laboratory of Nuclear Studies at Cornell University for
     the opportunity to test and develop for the AXP,
  John Hasstedt John.Hasstedt@sunysb.edu
     for VAX VMS V7.2 support
  John Malmberg wb8tyw@qsl.net
     for ODS-5 filename handling and other modernizations

and to the entire VMSperl group for useful advice and suggestions.  In
addition the perl5-porters deserve credit for their creativity and
willingness to work with the VMS newcomers.  Finally, the greatest debt of
gratitude is due to Larry Wall larry@wall.org, for having the ideas which
have made our sleepless nights possible.

Thanks,
The VMSperl group

=cut