floating point mangling warnings for README.os390 and README.posix-bc

[p5sagit/p5-mst-13.2.git] / README.os390

This document is written in pod format hence there are punctuation 
characters in odd places.  Do not worry, you've apparently got 
the ASCII->EBCDIC translation worked out correctly.  You can read 
more about pod in pod/perlpod.pod or the short summary in the 
INSTALL file.

=head1 NAME

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

=head1 SYNOPSIS

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

=head1 DESCRIPTION

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.  

=head2 Unpacking

Gunzip/gzip for OS/390 is discussed at:

   http://www.s390.ibm.com/products/oe/bpxqp1.html

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

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

=head2 Setup and utilities

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 required for the build of perl (as well as
building CPAN modules and extensions), is available from:

  http://www.mks.com/s390/gnu/index.htm

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.

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).

=head2 Configure

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:

=over 4

=item *

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.

=item *

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.  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.  If for some reason you do not
want to do that then, after ensuring that /etc/yyparse.c is properly in place 
run the following commands from within the Perl build directory:

    rm -f y.tab.c y.tab.h
    yacc -d perly.y 
    mv -f y.tab.c perly.c
    chmod u+w perly.c
    sed -e '/^#include "perl\.h"/a\
 \
 #define yydebug    PL_yydebug\
 #define yynerrs    PL_yynerrs\
 #define yyerrflag  PL_yyerrflag\
 #define yychar     PL_yychar\
 #define yyval      PL_yyval\
 #define yylval     PL_yylval'                           \
            -e '/YYSTYPE *yyval;/D'                     \
            -e '/YYSTYPE *yylval;/D'                    \
            -e '/int  yychar,/,/yynerrs;/D'             \
            -e 's/int yydebug = 0;/yydebug = 0;/'       \
            -e 's/[^_]realloc(/PerlMem_realloc(/g'      \
            -e 's/fprintf *( *stderr *,/PerlIO_printf(Perl_debug_log,/g' \
            -e 's/y\.tab/perly/g' perly.c >perly.tmp
    mv -f perly.tmp perly.c
    mv -f y.tab.h perly.h
    cd x2p
    rm -f y.tab.c y.tab.h
    yacc  a2p.y
    mv -f y.tab.c a2p.c
    chmod u+w a2p.c
    sed -e 's/fprintf *( *stderr *,/PerlIO_printf(Perl_debug_log,/g' \
                -e 's/y\.tab/a2p/g' a2p.c >a2p.tmp
    mv -f a2p.tmp a2p.c
    mv -f y.tab.h a2p.h
    cd ..

There, easy huh?  If you find typing all that in difficult then perhaps
you should reconsider the rm -rf of the perl build directory and 
re extraction of the source tar ball.

=item *

This port doesn't support dynamic loading.  Although OS/390 has support 
for DLLs via dllload(), there are some differences that cause problems 
for Perl.  (We need a volunteer to write a ext/DynaLoader/dl_dllload.xs 
file).

=item *

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.

=item *

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).

=back

=head2 Build, test, install

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.

=head2 build anomalies

"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 $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.

=head2 testing anomalies

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:

=over 4

=item *

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.

=item *

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.

=item *

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.

=back

=head2 Usage Hints

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.

=head2 Floating point anomalies

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 $y and $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) */
    }

=head2 Modules and Extensions

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

    perl Makefile.PL
    make
    make test
    make install

You can also build xs based extensions to Perl for OS/390 but 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.

=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.

=head1 SEE ALSO

L<INSTALL>, L<perlport>, L<perlebcdic>, L<ExtUtils::MakeMaker>.

    http://www.mks.com/s390/gnu/index.htm

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

    http://www.s390.ibm.com/products/oe/bpxa1ty1.html#opensrc

    http://www.s390.ibm.com/products/oe/portbk/bpxacenv.html

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

=head2 Mailing list

The Perl Institute (http://www.perl.org/) maintains a perl-mvs 
mailing list of interest to all folks building and/or
using perl on all EBCDIC platforms (not just OS/390).  
To subscribe, send a message of:

    subscribe perl-mvs

to majordomo@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/

=head1 HISTORY

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 12 November 2000 for the 5.7.1 release of Perl.

Updated 15 January 2001 for the 5.7.1 release of Perl.

=cut