Add more known sprintf failures.

[p5sagit/p5-mst-13.2.git] / README.bs2000

This document is written in pod format hence there are punctuation
characters in 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.posix-bc - building and installing Perl for BS2000 POSIX.

=head1 SYNOPSIS

This document will help you Configure, build, test and install Perl
on BS2000 in the POSIX subsystem.

=head1 DESCRIPTION

This is a ported perl for the POSIX subsystem in BS2000 VERSION OSD
V3.1A.  It may work on other versions, but that's the one we've tested
it on.

You may need the following GNU programs in order to install perl:

=head2 gzip

We used version 1.2.4, which could be installed out of the box with
one failure during 'make check'.

=head2 bison

The yacc coming with BS2000 POSIX didn't work for us.  So we had to
use bison.  We had to make a few changes to perl in order to use the
pure (reentrant) parser of bison.  We used version 1.25, but we had to
add a few changes due to EBCDIC.

=head2 Unpacking

To extract an ASCII tar archive on BS2000 POSIX you need an ASCII
filesystem (we used the mountpoint /usr/local/ascii for this).  Now
you extract the archive in the ASCII filesystem without
I/O-conversion:

cd /usr/local/ascii
export IO_CONVERSION=NO
gunzip < /usr/local/src/perl.tar.gz | pax -r

You may ignore the error message for the first element of the archive
(this doesn't look like a tar archive / skipping to next file...),
it's only the directory which will be created automatically anyway.

After extracting the archive you copy the whole directory tree to your
EBCDIC filesystem.  B<This time you use I/O-conversion>:

cd /usr/local/src
IO_CONVERSION=YES
cp -r /usr/local/ascii/perl5.005_02 ./

=head2 Compiling

There is a "hints" file for posix-bc that specifies the correct values
for most things.  The major problem is (of course) the EBCDIC character
set.  We have german EBCDIC version.

Because of our problems with the native yacc we used GNU bison to
generate a pure (=reentrant) parser for perly.y.  So our yacc is
really the following script:

-----8<-----/usr/local/bin/yacc-----8<-----
#! /usr/bin/sh

# Bison as a reentrant yacc:

# save parameters:
params=""
while [[ $# -gt 1 ]]; do
    params="$params $1"
    shift
done

# add flag %pure_parser:

tmpfile=/tmp/bison.$$.y
echo %pure_parser > $tmpfile
cat $1 >> $tmpfile

# call bison:

echo "/usr/local/bin/bison --yacc $params $1\t\t\t(Pure Parser)"
/usr/local/bin/bison --yacc $params $tmpfile

# cleanup:

rm -f $tmpfile
-----8<----------8<-----

We still use the normal yacc for a2p.y though!!!  We made a softlink
called byacc to distinguish between the two versions:

ln -s /usr/bin/yacc /usr/local/bin/byacc

We build perl using GNU make.  We tried the native make once and it
worked too.

=head2 Testing

We still got a few errors during C<make test>.  Some of them are the
result of using bison.  Bison prints I<parser error> instead of I<syntax
error>, so we may ignore them.  The following list shows
our errors, your results may differ:

op/numconvert.......FAILED tests 1409-1440
op/regexp...........FAILED tests 483, 496
op/regexp_noamp.....FAILED tests 483, 496
pragma/overload.....FAILED tests 152-153, 170-171
pragma/warnings.....FAILED tests 14, 82, 129, 155, 192, 205, 207
lib/bigfloat........FAILED tests 351-352, 355
lib/bigfltpm........FAILED tests 354-355, 358
lib/complex.........FAILED tests 267, 487
lib/dumper..........FAILED tests 43, 45
Failed 11/231 test scripts, 95.24% okay. 57/10595 subtests failed, 99.46% okay.

=head2 Install

We have no nroff on BS2000 POSIX (yet), so we ignored any errors while
installing the documentation.


=head2 Using Perl

BS2000 POSIX doesn't support the shebang notation
(C<#!/usr/local/bin/perl>), so you have to use the following lines
instead:

: # use perl
    eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}'
        if $running_under_some_shell;

=head2 Floating point anomalies

There appears to be a bug in the floating point implementation on BS2000 POSIX
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.

=head1 AUTHORS

Thomas Dorner

=head1 SEE ALSO

L<INSTALL>, L<perlport>.

=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 EBCDIC
platforms.  To subscribe, send a message of:

    subscribe perl-mvs

to majordomo@perl.org.

=head1 HISTORY

This document was originally written by Thomas Dorner for the 5.005
release of Perl.

This document was podified for the 5.6 release of perl 11 July 2000.

=cut