perlport(1)
NAME
perlport - Writing portable Perl
DESCRIPTION
Perl runs on numerous operating systems. While most of them share much
in common, they also have their own unique features.
This document is meant to help you to find out what constitutes porta-
ble Perl code. That way once you make a decision to write portably,
you know where the lines are drawn, and you can stay within them.
There is a tradeoff between taking full advantage of one particular
type of computer and taking advantage of a full range of them. Natu-
rally, as you broaden your range and become more diverse, the common
factors drop, and you are left with an increasingly smaller area of
common ground in which you can operate to accomplish a particular task.
Thus, when you begin attacking a problem, it is important to consider
under which part of the tradeoff curve you want to operate. Specifi-
cally, you must decide whether it is important that the task that you
are coding have the full generality of being portable, or whether to
just get the job done right now. This is the hardest choice to be
made. The rest is easy, because Perl provides many choices, whichever
way you want to approach your problem.
Looking at it another way, writing portable code is usually about will-
fully limiting your available choices. Naturally, it takes discipline
and sacrifice to do that. The product of portability and convenience
may be a constant. You have been warned.
Be aware of two important points:
Not all Perl programs have to be portable
There is no reason you should not use Perl as a language to glue
Unix tools together, or to prototype a Macintosh application, or to
manage the Windows registry. If it makes no sense to aim for
portability for one reason or another in a given program, then
don't bother.
Nearly all of Perl already is portable
Don't be fooled into thinking that it is hard to create portable
Perl code. It isn't. Perl tries its level-best to bridge the gaps
between what's available on different platforms, and all the means
available to use those features. Thus almost all Perl code runs on
any machine without modification. But there are some significant
issues in writing portable code, and this document is entirely
about those issues.
Here's the general rule: When you approach a task commonly done using a
whole range of platforms, think about writing portable code. That way,
you don't sacrifice much by way of the implementation choices you can
avail yourself of, and at the same time you can give your users lots of
platform choices. On the other hand, when you have to take advantage
of some unique feature of a particular platform, as is often the case
with systems programming (whether for Unix, Windows, Mac OS, VMS,
etc.), consider writing platform-specific code.
When the code will run on only two or three operating systems, you may
need to consider only the differences of those particular systems. The
important thing is to decide where the code will run and to be deliber-
ate in your decision.
The material below is separated into three main sections: main issues
of portability ("ISSUES"), platform-specific issues ("PLATFORMS"), and
built-in perl functions that behave differently on various ports
("FUNCTION IMPLEMENTATIONS").
This information should not be considered complete; it includes possi-
bly transient information about idiosyncrasies of some of the ports,
almost all of which are in a state of constant evolution. Thus, this
material should be considered a perpetual work in progress ("<IMG
SRC="yellow_sign.gif" ALT="Under Construction">").
ISSUES
Newlines
In most operating systems, lines in files are terminated by newlines.
Just what is used as a newline may vary from OS to OS. Unix tradition-
ally uses "\012", one type of DOSish I/O uses "\015\012", and Mac OS
uses "\015".
Perl uses "\n" to represent the "logical" newline, where what is logi-
cal may depend on the platform in use. In MacPerl, "\n" always means
"\015". In DOSish perls, "\n" usually means "\012", but when accessing
a file in "text" mode, STDIO translates it to (or from) "\015\012",
depending on whether you're reading or writing. Unix does the same
thing on ttys in canonical mode. "\015\012" is commonly referred to as
CRLF.
A common cause of unportable programs is the misuse of chop() to trim
newlines:
# XXX UNPORTABLE!
while(<FILE>) {
chop;
@array = split(/:/);
#...
}
You can get away with this on Unix and Mac OS (they have a single char-
acter end-of-line), but the same program will break under DOSish perls
because you're only chop()ing half the end-of-line. Instead, chomp()
should be used to trim newlines. The Dunce::Files module can help
audit your code for misuses of chop().
When dealing with binary files (or text files in binary mode) be sure
to explicitly set $/ to the appropriate value for your file format
before using chomp().
Because of the "text" mode translation, DOSish perls have limitations
in using "seek" and "tell" on a file accessed in "text" mode. Stick to
"seek"-ing to locations you got from "tell" (and no others), and you
are usually free to use "seek" and "tell" even in "text" mode. Using
"seek" or "tell" or other file operations may be non-portable. If you
use "binmode" on a file, however, you can usually "seek" and "tell"
with arbitrary values in safety.
A common misconception in socket programming is that "\n" eq "\012"
everywhere. When using protocols such as common Internet protocols,
"\012" and "\015" are called for specifically, and the values of the
logical "\n" and "\r" (carriage return) are not reliable.
print SOCKET "Hi there, client!\r\n"; # WRONG
print SOCKET "Hi there, client!\015\012"; # RIGHT
However, using "\015\012" (or "\cM\cJ", or "\x0D\x0A") can be tedious
and unsightly, as well as confusing to those maintaining the code. As
such, the Socket module supplies the Right Thing for those who want it.
use Socket qw(:DEFAULT :crlf);
print SOCKET "Hi there, client!$CRLF" # RIGHT
When reading from a socket, remember that the default input record sep-
arator $/ is "\n", but robust socket code will recognize as either
"\012" or "\015\012" as end of line:
while (<SOCKET>) {
# ...
}
Because both CRLF and LF end in LF, the input record separator can be
set to LF and any CR stripped later. Better to write:
use Socket qw(:DEFAULT :crlf);
local($/) = LF; # not needed if $/ is already \012
while (<SOCKET>) {
s/$CR?$LF/\n/; # not sure if socket uses LF or CRLF, OK
# s/\015?\012/\n/; # same thing
}
This example is preferred over the previous one--even for Unix plat-
forms--because now any "\015"'s ("\cM"'s) are stripped out (and there
was much rejoicing).
Similarly, functions that return text data--such as a function that
fetches a web page--should sometimes translate newlines before return-
ing the data, if they've not yet been translated to the local newline
representation. A single line of code will often suffice:
$data =~ s/\015?\012/\n/g;
return $data;
Some of this may be confusing. Here's a handy reference to the ASCII
CR and LF characters. You can print it out and stick it in your wal-
let.
LF eq \012 eq \x0A eq \cJ eq chr(10) eq ASCII 10
CR eq \015 eq \x0D eq \cM eq chr(13) eq ASCII 13
| Unix | DOS | Mac |
---------------------------
\n | LF | LF | CR |
\r | CR | CR | LF |
\n * | LF | CRLF | CR |
\r * | CR | CR | LF |
---------------------------
* text-mode STDIO
The Unix column assumes that you are not accessing a serial line (like
a tty) in canonical mode. If you are, then CR on input becomes "\n",
and "\n" on output becomes CRLF.
These are just the most common definitions of "\n" and "\r" in Perl.
There may well be others. For example, on an EBCDIC implementation
such as z/OS (OS/390) or OS/400 (using the ILE, the PASE is
ASCII-based) the above material is similar to "Unix" but the code num-
bers change:
LF eq \025 eq \x15 eq \cU eq chr(21) eq CP-1047 21
LF eq \045 eq \x25 eq chr(37) eq CP-0037 37
CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-1047 13
CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-0037 13
| z/OS | OS/400 |
----------------------
\n | LF | LF |
\r | CR | CR |
\n * | LF | LF |
\r * | CR | CR |
----------------------
* text-mode STDIO
Numbers endianness and Width
Different CPUs store integers and floating point numbers in different
orders (called endianness) and widths (32-bit and 64-bit being the most
common today). This affects your programs when they attempt to trans-
fer numbers in binary format from one CPU architecture to another, usu-
ally either "live" via network connection, or by storing the numbers to
secondary storage such as a disk file or tape.
Conflicting storage orders make utter mess out of the numbers. If a
little-endian host (Intel, VAX) stores 0x12345678 (305419896 in deci-
mal), a big-endian host (Motorola, Sparc, PA) reads it as 0x78563412
(2018915346 in decimal). Alpha and MIPS can be either: Digital/Compaq
used/uses them in little-endian mode; SGI/Cray uses them in big-endian
mode. To avoid this problem in network (socket) connections use the
"pack" and "unpack" formats "n" and "N", the "network" orders. These
are guaranteed to be portable.
As of perl 5.8.5, you can also use the ">" and "<" modifiers to force
big- or little-endian byte-order. This is useful if you want to store
signed integers or 64-bit integers, for example.
You can explore the endianness of your platform by unpacking a data
structure packed in native format such as:
print unpack("h*", pack("s2", 1, 2)), "\n";
# '10002000' on e.g. Intel x86 or Alpha 21064 in little-endian mode
# '00100020' on e.g. Motorola 68040
If you need to distinguish between endian architectures you could use
either of the variables set like so:
$is_big_endian = unpack("h*", pack("s", 1)) =~ /01/;
$is_little_endian = unpack("h*", pack("s", 1)) =~ /^1/;
Differing widths can cause truncation even between platforms of equal
endianness. The platform of shorter width loses the upper parts of the
number. There is no good solution for this problem except to avoid
transferring or storing raw binary numbers.
One can circumnavigate both these problems in two ways. Either trans-
fer and store numbers always in text format, instead of raw binary, or
else consider using modules like Data::Dumper (included in the standard
distribution as of Perl 5.005) and Storable (included as of perl 5.8).
Keeping all data as text significantly simplifies matters.
The v-strings are portable only up to v2147483647 (0x7FFFFFFF), that's
how far EBCDIC, or more precisely UTF-EBCDIC will go.
Files and Filesystems
Most platforms these days structure files in a hierarchical fashion.
So, it is reasonably safe to assume that all platforms support the
notion of a "path" to uniquely identify a file on the system. How that
path is really written, though, differs considerably.
Although similar, file path specifications differ between Unix, Win-
dows, Mac OS, OS/2, VMS, VOS, RISC OS, and probably others. Unix, for
example, is one of the few OSes that has the elegant idea of a single
root directory.
DOS, OS/2, VMS, VOS, and Windows can work similarly to Unix with "/" as
path separator, or in their own idiosyncratic ways (such as having sev-
eral root directories and various "unrooted" device files such NIL: and
LPT:).
Mac OS uses ":" as a path separator instead of "/".
The filesystem may support neither hard links ("link") nor symbolic
links ("symlink", "readlink", "lstat").
The filesystem may support neither access timestamp nor change time-
stamp (meaning that about the only portable timestamp is the modifica-
tion timestamp), or one second granularity of any timestamps (e.g. the
FAT filesystem limits the time granularity to two seconds).
The "inode change timestamp" (the "-C" filetest) may really be the
"creation timestamp" (which it is not in UNIX).
VOS perl can emulate Unix filenames with "/" as path separator. The
native pathname characters greater-than, less-than, number-sign, and
percent-sign are always accepted.
RISC OS perl can emulate Unix filenames with "/" as path separator, or
go native and use "." for path separator and ":" to signal filesystems
and disk names.
Don't assume UNIX filesystem access semantics: that read, write, and
execute are all the permissions there are, and even if they exist, that
their semantics (for example what do r, w, and x mean on a directory)
are the UNIX ones. The various UNIX/POSIX compatibility layers usually
try to make interfaces like chmod() work, but sometimes there simply is
no good mapping.
If all this is intimidating, have no (well, maybe only a little) fear.
There are modules that can help. The File::Spec modules provide meth-
ods to do the Right Thing on whatever platform happens to be running
the program.
use File::Spec::Functions;
chdir(updir()); # go up one directory
$file = catfile(curdir(), 'temp', 'file.txt');
# on Unix and Win32, './temp/file.txt'
# on Mac OS, ':temp:file.txt'
# on VMS, '[.temp]file.txt'
File::Spec is available in the standard distribution as of version
5.004_05. File::Spec::Functions is only in File::Spec 0.7 and later,
and some versions of perl come with version 0.6. If File::Spec is not
updated to 0.7 or later, you must use the object-oriented interface
from File::Spec (or upgrade File::Spec).
In general, production code should not have file paths hardcoded. Mak-
ing them user-supplied or read from a configuration file is better,
keeping in mind that file path syntax varies on different machines.
This is especially noticeable in scripts like Makefiles and test
suites, which often assume "/" as a path separator for subdirectories.
Also of use is File::Basename from the standard distribution, which
splits a pathname into pieces (base filename, full path to directory,
and file suffix).
Even when on a single platform (if you can call Unix a single plat-
form), remember not to count on the existence or the contents of par-
ticular system-specific files or directories, like /etc/passwd,
/etc/sendmail.conf, /etc/resolv.conf, or even /tmp/. For example,
/etc/passwd may exist but not contain the encrypted passwords, because
the system is using some form of enhanced security. Or it may not con-
tain all the accounts, because the system is using NIS. If code does
need to rely on such a file, include a description of the file and its
format in the code's documentation, then make it easy for the user to
override the default location of the file.
Don't assume a text file will end with a newline. They should, but
people forget.
Do not have two files or directories of the same name with different
case, like test.pl and Test.pl, as many platforms have case-insensitive
(or at least case-forgiving) filenames. Also, try not to have non-word
characters (except for ".") in the names, and keep them to the 8.3 con-
vention, for maximum portability, onerous a burden though this may
appear.
Likewise, when using the AutoSplit module, try to keep your functions
to 8.3 naming and case-insensitive conventions; or, at the least, make
it so the resulting files have a unique (case-insensitively) first 8
characters.
Whitespace in filenames is tolerated on most systems, but not all, and
even on systems where it might be tolerated, some utilities might
become confused by such whitespace.
Many systems (DOS, VMS) cannot have more than one "." in their file-
names.
Don't assume ">" won't be the first character of a filename. Always
use "<" explicitly to open a file for reading, or even better, use the
three-arg version of open, unless you want the user to be able to spec-
ify a pipe open.
open(FILE, '<', $existing_file) or die $!;
If filenames might use strange characters, it is safest to open it with
"sysopen" instead of "open". "open" is magic and can translate charac-
ters like ">", "<", and "|", which may be the wrong thing to do.
(Sometimes, though, it's the right thing.) Three-arg open can also
help protect against this translation in cases where it is undesirable.
Don't use ":" as a part of a filename since many systems use that for
their own semantics (Mac OS Classic for separating pathname components,
many networking schemes and utilities for separating the nodename and
the pathname, and so on). For the same reasons, avoid "@", ";" and
"|".
Don't assume that in pathnames you can collapse two leading slashes
"//" into one: some networking and clustering filesystems have special
semantics for that. Let the operating system to sort it out.
The portable filename characters as defined by ANSI C are
a b c d e f g h i j k l m n o p q r t u v w x y z
A B C D E F G H I J K L M N O P Q R T U V W X Y Z
0 1 2 3 4 5 6 7 8 9
. _ -
and the "-" shouldn't be the first character. If you want to be hyper-
correct, stay case-insensitive and within the 8.3 naming convention
(all the files and directories have to be unique within one directory
if their names are lowercased and truncated to eight characters before
the ".", if any, and to three characters after the ".", if any). (And
do not use "."s in directory names.)
System Interaction
Not all platforms provide a command line. These are usually platforms
that rely primarily on a Graphical User Interface (GUI) for user inter-
action. A program requiring a command line interface might not work
everywhere. This is probably for the user of the program to deal with,
so don't stay up late worrying about it.
Some platforms can't delete or rename files held open by the system,
this limitation may also apply to changing filesystem metainformation
like file permissions or owners. Remember to "close" files when you
are done with them. Don't "unlink" or "rename" an open file. Don't
"tie" or "open" a file already tied or opened; "untie" or "close" it
first.
Don't open the same file more than once at a time for writing, as some
operating systems put mandatory locks on such files.
Don't assume that write/modify permission on a directory gives the
right to add or delete files/directories in that directory. That is
filesystem specific: in some filesystems you need write/modify permis-
sion also (or even just) in the file/directory itself. In some
filesystems (AFS, DFS) the permission to add/delete directory entries
is a completely separate permission.
Don't assume that a single "unlink" completely gets rid of the file:
some filesystems (most notably the ones in VMS) have versioned filesys-
tems, and unlink() removes only the most recent one (it doesn't remove
all the versions because by default the native tools on those platforms
remove just the most recent version, too). The portable idiom to
remove all the versions of a file is
1 while unlink "file";
This will terminate if the file is undeleteable for some reason (pro-
tected, not there, and so on).
Don't count on a specific environment variable existing in %ENV. Don't
count on %ENV entries being case-sensitive, or even case-preserving.
Don't try to clear %ENV by saying "%ENV = ();", or, if you really have
to, make it conditional on "$^O ne 'VMS'" since in VMS the %ENV table
is much more than a per-process key-value string table.
Don't count on signals or %SIG for anything.
Don't count on filename globbing. Use "opendir", "readdir", and
"closedir" instead.
Don't count on per-program environment variables, or per-program cur-
rent directories.
Don't count on specific values of $!, neither numeric nor especially
the strings values-- users may switch their locales causing error mes-
sages to be translated into their languages. If you can trust a POSIX-
ish environment, you can portably use the symbols defined by the Errno
module, like ENOENT. And don't trust on the values of $! at all
except immediately after a failed system call.
Command names versus file pathnames
Don't assume that the name used to invoke a command or program with
"system" or "exec" can also be used to test for the existence of the
file that holds the executable code for that command or program.
First, many systems have "internal" commands that are built-in to the
shell or OS and while these commands can be invoked, there is no corre-
sponding file. Second, some operating systems (e.g., Cygwin, DJGPP,
OS/2, and VOS) have required suffixes for executable files; these suf-
fixes are generally permitted on the command name but are not required.
Thus, a command like "perl" might exist in a file named "perl",
"perl.exe", or "perl.pm", depending on the operating system. The vari-
able "_exe" in the Config module holds the executable suffix, if any.
Third, the VMS port carefully sets up $^X and $Config{perlpath} so that
no further processing is required. This is just as well, because the
matching regular expression used below would then have to deal with a
possible trailing version number in the VMS file name.
To convert $^X to a file pathname, taking account of the requirements
of the various operating system possibilities, say:
use Config;
$thisperl = $^X;
if ($^O ne 'VMS')
{$thisperl .= $Config{_exe} unless $thisperl =~ m/$Config{_exe}$/i;}
To convert $Config{perlpath} to a file pathname, say:
use Config;
$thisperl = $Config{perlpath};
if ($^O ne 'VMS')
{$thisperl .= $Config{_exe} unless $thisperl =~ m/$Config{_exe}$/i;}
Networking
Don't assume that you can reach the public Internet.
Don't assume that there is only one way to get through firewalls to the
public Internet.
Don't assume that you can reach outside world through any other port
than 80, or some web proxy. ftp is blocked by many firewalls.
Don't assume that you can send email by connecting to the local SMTP
port.
Don't assume that you can reach yourself or any node by the name
'localhost'. The same goes for '127.0.0.1'. You will have to try
both.
Don't assume that the host has only one network card, or that it can't
bind to many virtual IP addresses.
Don't assume a particular network device name.
Don't assume a particular set of ioctl()s will work.
Don't assume that you can ping hosts and get replies.
Don't assume that any particular port (service) will respond.
Don't assume that Sys::Hostname (or any other API or command) returns
either a fully qualified hostname or a non-qualified hostname: it all
depends on how the system had been configured. Also remember things
like DHCP and NAT-- the hostname you get back might not be very useful.
All the above "don't":s may look daunting, and they are -- but the key
is to degrade gracefully if one cannot reach the particular network
service one wants. Croaking or hanging do not look very professional.
Interprocess Communication (IPC)
In general, don't directly access the system in code meant to be porta-
ble. That means, no "system", "exec", "fork", "pipe", ``, "qx//",
"open" with a "|", nor any of the other things that makes being a perl
hacker worth being.
Commands that launch external processes are generally supported on most
platforms (though many of them do not support any type of forking).
The problem with using them arises from what you invoke them on.
External tools are often named differently on different platforms, may
not be available in the same location, might accept different argu-
ments, can behave differently, and often present their results in a
platform-dependent way. Thus, you should seldom depend on them to pro-
duce consistent results. (Then again, if you're calling netstat -a, you
probably don't expect it to run on both Unix and CP/M.)
One especially common bit of Perl code is opening a pipe to sendmail:
open(MAIL, '|/usr/lib/sendmail -t')
or die "cannot fork sendmail: $!";
This is fine for systems programming when sendmail is known to be
available. But it is not fine for many non-Unix systems, and even some
Unix systems that may not have sendmail installed. If a portable solu-
tion is needed, see the various distributions on CPAN that deal with
it. Mail::Mailer and Mail::Send in the MailTools distribution are com-
monly used, and provide several mailing methods, including mail, send-
mail, and direct SMTP (via Net::SMTP) if a mail transfer agent is not
available. Mail::Sendmail is a standalone module that provides simple,
platform-independent mailing.
The Unix System V IPC ("msg*(), sem*(), shm*()") is not available even
on all Unix platforms.
Do not use either the bare result of "pack("N", 10, 20, 30, 40)" or
bare v-strings (such as "v10.20.30.40") to represent IPv4 addresses:
both forms just pack the four bytes into network order. That this
would be equal to the C language "in_addr" struct (which is what the
socket code internally uses) is not guaranteed. To be portable use the
routines of the Socket extension, such as "inet_aton()", "inet_ntoa()",
and "sockaddr_in()".
The rule of thumb for portable code is: Do it all in portable Perl, or
use a module (that may internally implement it with platform-specific
code, but expose a common interface).
External Subroutines (XS)
XS code can usually be made to work with any platform, but dependent
libraries, header files, etc., might not be readily available or porta-
ble, or the XS code itself might be platform-specific, just as Perl
code might be. If the libraries and headers are portable, then it is
normally reasonable to make sure the XS code is portable, too.
A different type of portability issue arises when writing XS code:
availability of a C compiler on the end-user's system. C brings with
it its own portability issues, and writing XS code will expose you to
some of those. Writing purely in Perl is an easier way to achieve
portability.
Standard Modules
In general, the standard modules work across platforms. Notable excep-
tions are the CPAN module (which currently makes connections to exter-
nal programs that may not be available), platform-specific modules
(like ExtUtils::MM_VMS), and DBM modules.
There is no one DBM module available on all platforms. SDBM_File and
the others are generally available on all Unix and DOSish ports, but
not in MacPerl, where only NBDM_File and DB_File are available.
The good news is that at least some DBM module should be available, and
AnyDBM_File will use whichever module it can find. Of course, then the
code needs to be fairly strict, dropping to the greatest common factor
(e.g., not exceeding 1K for each record), so that it will work with any
DBM module. See AnyDBM_File for more details.
Time and Date
The system's notion of time of day and calendar date is controlled in
widely different ways. Don't assume the timezone is stored in
$ENV{TZ}, and even if it is, don't assume that you can control the
timezone through that variable. Don't assume anything about the three-
letter timezone abbreviations (for example that MST would be the Moun-
tain Standard Time, it's been known to stand for Moscow Standard Time).
If you need to use timezones, express them in some unambiguous format
like the exact number of minutes offset from UTC, or the POSIX timezone
format.
Don't assume that the epoch starts at 00:00:00, January 1, 1970,
because that is OS- and implementation-specific. It is better to store
a date in an unambiguous representation. The ISO 8601 standard defines
YYYY-MM-DD as the date format, or YYYY-MM-DDTHH-MM-SS (that's a literal
"T" separating the date from the time). Please do use the ISO 8601
instead of making us to guess what date 02/03/04 might be. ISO 8601
even sorts nicely as-is. A text representation (like "1987-12-18") can
be easily converted into an OS-specific value using a module like
Date::Parse. An array of values, such as those returned by "local-
time", can be converted to an OS-specific representation using
Time::Local.
When calculating specific times, such as for tests in time or date mod-
ules, it may be appropriate to calculate an offset for the epoch.
require Time::Local;
$offset = Time::Local::timegm(0, 0, 0, 1, 0, 70);
The value for $offset in Unix will be 0, but in Mac OS will be some
large number. $offset can then be added to a Unix time value to get
what should be the proper value on any system.
On Windows (at least), you shouldn't pass a negative value to "gmtime"
or "localtime".
Character sets and character encoding
Assume very little about character sets.
Assume nothing about numerical values ("ord", "chr") of characters. Do
not use explicit code point ranges (like \xHH-\xHH); use for example
symbolic character classes like "[:print:]".
Do not assume that the alphabetic characters are encoded contiguously
(in the numeric sense). There may be gaps.
Do not assume anything about the ordering of the characters. The low-
ercase letters may come before or after the uppercase letters; the low-
ercase and uppercase may be interlaced so that both "a" and "A" come
before "b"; the accented and other international characters may be
interlaced so that ae comes before "b".
Internationalisation
If you may assume POSIX (a rather large assumption), you may read more
about the POSIX locale system from perllocale. The locale system at
least attempts to make things a little bit more portable, or at least
more convenient and native-friendly for non-English users. The system
affects character sets and encoding, and date and time format-
ting--amongst other things.
If you really want to be international, you should consider Unicode.
See perluniintro and perlunicode for more information.
If you want to use non-ASCII bytes (outside the bytes 0x00..0x7f) in
the "source code" of your code, to be portable you have to be explicit
about what bytes they are. Someone might for example be using your
code under a UTF-8 locale, in which case random native bytes might be
illegal ("Malformed UTF-8 ...") This means that for example embedding
ISO 8859-1 bytes beyond 0x7f into your strings might cause trouble
later. If the bytes are native 8-bit bytes, you can use the "bytes"
pragma. If the bytes are in a string (regular expression being a curi-
ous string), you can often also use the "\xHH" notation instead of
embedding the bytes as-is. If they are in some particular legacy
encoding (ether single-byte or something more complicated), you can use
the "encoding" pragma. (If you want to write your code in UTF-8, you
can use either the "utf8" pragma, or the "encoding" pragma.) The
"bytes" and "utf8" pragmata are available since Perl 5.6.0, and the
"encoding" pragma since Perl 5.8.0.
System Resources
If your code is destined for systems with severely constrained (or
missing!) virtual memory systems then you want to be especially mindful
of avoiding wasteful constructs such as:
# NOTE: this is no longer "bad" in perl5.005
for (0..10000000) {} # bad
for (my $x = 0; $x <= 10000000; ++$x) {} # good
@lines = <VERY_LARGE_FILE>; # bad
while (<FILE>) {$file .= $_} # sometimes bad
$file = join('', <FILE>); # better
The last two constructs may appear unintuitive to most people. The
first repeatedly grows a string, whereas the second allocates a large
chunk of memory in one go. On some systems, the second is more effi-
cient that the first.
Security
Most multi-user platforms provide basic levels of security, usually
implemented at the filesystem level. Some, however, do not-- unfortu-
nately. Thus the notion of user id, or "home" directory, or even the
state of being logged-in, may be unrecognizable on many platforms. If
you write programs that are security-conscious, it is usually best to
know what type of system you will be running under so that you can
write code explicitly for that platform (or class of platforms).
Don't assume the UNIX filesystem access semantics: the operating system
or the filesystem may be using some ACL systems, which are richer lan-
guages than the usual rwx. Even if the rwx exist, their semantics
might be different.
(From security viewpoint testing for permissions before attempting to
do something is silly anyway: if one tries this, there is potential for
race conditions-- someone or something might change the permissions
between the permissions check and the actual operation. Just try the
operation.)
Don't assume the UNIX user and group semantics: especially, don't
expect the $< and $> (or the $( and $)) to work for switching identi-
ties (or memberships).
Don't assume set-uid and set-gid semantics. (And even if you do, think
twice: set-uid and set-gid are a known can of security worms.)
Style
For those times when it is necessary to have platform-specific code,
consider keeping the platform-specific code in one place, making port-
ing to other platforms easier. Use the Config module and the special
variable $^O to differentiate platforms, as described in "PLATFORMS".
Be careful in the tests you supply with your module or programs. Mod-
ule code may be fully portable, but its tests might not be. This often
happens when tests spawn off other processes or call external programs
to aid in the testing, or when (as noted above) the tests assume cer-
tain things about the filesystem and paths. Be careful not to depend
on a specific output style for errors, such as when checking $! after a
failed system call. Using $! for anything else than displaying it as
output is doubtful (though see the Errno module for testing reasonably
portably for error value). Some platforms expect a certain output for-
mat, and Perl on those platforms may have been adjusted accordingly.
Most specifically, don't anchor a regex when testing an error value.
CPAN Testers
Modules uploaded to CPAN are tested by a variety of volunteers on dif-
ferent platforms. These CPAN testers are notified by mail of each new
upload, and reply to the list with PASS, FAIL, NA (not applicable to
this platform), or UNKNOWN (unknown), along with any relevant nota-
tions.
The purpose of the testing is twofold: one, to help developers fix any
problems in their code that crop up because of lack of testing on other
platforms; two, to provide users with information about whether a given
module works on a given platform.
Also see:
o Mailing list: cpan-testers@perl.org
o Testing results: http://testers.cpan.org/
PLATFORMS
As of version 5.002, Perl is built with a $^O variable that indicates
the operating system it was built on. This was implemented to help
speed up code that would otherwise have to "use Config" and use the
value of $Config{osname}. Of course, to get more detailed information
about the system, looking into %Config is certainly recommended.
%Config cannot always be trusted, however, because it was built at com-
pile time. If perl was built in one place, then transferred elsewhere,
some values may be wrong. The values may even have been edited after
the fact.
Unix
Perl works on a bewildering variety of Unix and Unix-like platforms
(see e.g. most of the files in the hints/ directory in the source code
kit). On most of these systems, the value of $^O (hence $Con-
fig{'osname'}, too) is determined either by lowercasing and stripping
punctuation from the first field of the string returned by typing
"uname -a" (or a similar command) at the shell prompt or by testing the
file system for the presence of uniquely named files such as a kernel
or header file. Here, for example, are a few of the more popular Unix
flavors:
uname $^O $Config{'archname'}
--------------------------------------------
AIX aix aix
BSD/OS bsdos i386-bsdos
Darwin darwin darwin
dgux dgux AViiON-dgux
DYNIX/ptx dynixptx i386-dynixptx
FreeBSD freebsd freebsd-i386
Linux linux arm-linux
Linux linux i386-linux
Linux linux i586-linux
Linux linux ppc-linux
HP-UX hpux PA-RISC1.1
IRIX irix irix
Mac OS X darwin darwin
MachTen PPC machten powerpc-machten
NeXT 3 next next-fat
NeXT 4 next OPENSTEP-Mach
openbsd openbsd i386-openbsd
OSF1 dec_osf alpha-dec_osf
reliantunix-n svr4 RM400-svr4
SCO_SV sco_sv i386-sco_sv
SINIX-N svr4 RM400-svr4
sn4609 unicos CRAY_C90-unicos
sn6521 unicosmk t3e-unicosmk
sn9617 unicos CRAY_J90-unicos
SunOS solaris sun4-solaris
SunOS solaris i86pc-solaris
SunOS4 sunos sun4-sunos
Because the value of $Config{archname} may depend on the hardware
architecture, it can vary more than the value of $^O.
DOS and Derivatives
Perl has long been ported to Intel-style microcomputers running under
systems like PC-DOS, MS-DOS, OS/2, and most Windows platforms you can
bring yourself to mention (except for Windows CE, if you count that).
Users familiar with COMMAND.COM or CMD.EXE style shells should be aware
that each of these file specifications may have subtle differences:
$filespec0 = "c:/foo/bar/file.txt";
$filespec1 = "c:\\foo\\bar\\file.txt";
$filespec2 = 'c:\foo\bar\file.txt';
$filespec3 = 'c:\\foo\\bar\\file.txt';
System calls accept either "/" or "\" as the path separator. However,
many command-line utilities of DOS vintage treat "/" as the option pre-
fix, so may get confused by filenames containing "/". Aside from call-
ing any external programs, "/" will work just fine, and probably bet-
ter, as it is more consistent with popular usage, and avoids the prob-
lem of remembering what to backwhack and what not to.
The DOS FAT filesystem can accommodate only "8.3" style filenames.
Under the "case-insensitive, but case-preserving" HPFS (OS/2) and NTFS
(NT) filesystems you may have to be careful about case returned with
functions like "readdir" or used with functions like "open" or
"opendir".
DOS also treats several filenames as special, such as AUX, PRN, NUL,
CON, COM1, LPT1, LPT2, etc. Unfortunately, sometimes these filenames
won't even work if you include an explicit directory prefix. It is
best to avoid such filenames, if you want your code to be portable to
DOS and its derivatives. It's hard to know what these all are, unfor-
tunately.
Users of these operating systems may also wish to make use of scripts
such as pl2bat.bat or pl2cmd to put wrappers around your scripts.
Newline ("\n") is translated as "\015\012" by STDIO when reading from
and writing to files (see "Newlines"). "binmode(FILEHANDLE)" will keep
"\n" translated as "\012" for that filehandle. Since it is a no-op on
other systems, "binmode" should be used for cross-platform code that
deals with binary data. That's assuming you realize in advance that
your data is in binary. General-purpose programs should often assume
nothing about their data.
The $^O variable and the $Config{archname} values for various DOSish
perls are as follows:
OS $^O $Config{archname} ID Version
--------------------------------------------------------
MS-DOS dos ?
PC-DOS dos ?
OS/2 os2 ?
Windows 3.1 ? ? 0 3 01
Windows 95 MSWin32 MSWin32-x86 1 4 00
Windows 98 MSWin32 MSWin32-x86 1 4 10
Windows ME MSWin32 MSWin32-x86 1 ?
Windows NT MSWin32 MSWin32-x86 2 4 xx
Windows NT MSWin32 MSWin32-ALPHA 2 4 xx
Windows NT MSWin32 MSWin32-ppc 2 4 xx
Windows 2000 MSWin32 MSWin32-x86 2 5 00
Windows XP MSWin32 MSWin32-x86 2 5 01
Windows 2003 MSWin32 MSWin32-x86 2 5 02
Windows CE MSWin32 ? 3
Cygwin cygwin cygwin
The various MSWin32 Perl's can distinguish the OS they are running on
via the value of the fifth element of the list returned from
Win32::GetOSVersion(). For example:
if ($^O eq 'MSWin32') {
my @os_version_info = Win32::GetOSVersion();
print +('3.1','95','NT')[$os_version_info[4]],"\n";
}
There are also Win32::IsWinNT() and Win32::IsWin95(), try "perldoc
Win32", and as of libwin32 0.19 (not part of the core Perl distribu-
tion) Win32::GetOSName(). The very portable POSIX::uname() will work
too:
c:\> perl -MPOSIX -we "print join '|', uname"
Windows NT|moonru|5.0|Build 2195 (Service Pack 2)|x86
Also see:
o The djgpp environment for DOS, http://www.delorie.com/djgpp/ and
perldos.
o The EMX environment for DOS, OS/2, etc. emx@iaehv.nl,
http://www.leo.org/pub/comp/os/os2/leo/gnu/emx+gcc/index.html or
ftp://hobbes.nmsu.edu/pub/os2/dev/emx/ Also perlos2.
o Build instructions for Win32 in perlwin32, or under the Cygnus
environment in perlcygwin.
o The "Win32::*" modules in Win32.
o The ActiveState Pages, http://www.activestate.com/
o The Cygwin environment for Win32; README.cygwin (installed as perl-
cygwin), http://www.cygwin.com/
o The U/WIN environment for Win32,
http://www.research.att.com/sw/tools/uwin/
o Build instructions for OS/2, perlos2
Mac OS
Any module requiring XS compilation is right out for most people,
because MacPerl is built using non-free (and non-cheap!) compilers.
Some XS modules that can work with MacPerl are built and distributed in
binary form on CPAN.
Directories are specified as:
volume:folder:file for absolute pathnames
volume:folder: for absolute pathnames
:folder:file for relative pathnames
:folder: for relative pathnames
:file for relative pathnames
file for relative pathnames
Files are stored in the directory in alphabetical order. Filenames are
limited to 31 characters, and may include any character except for null
and ":", which is reserved as the path separator.
Instead of "flock", see "FSpSetFLock" and "FSpRstFLock" in the
Mac::Files module, or "chmod(0444, ...)" and "chmod(0666, ...)".
In the MacPerl application, you can't run a program from the command
line; programs that expect @ARGV to be populated can be edited with
something like the following, which brings up a dialog box asking for
the command line arguments.
if (!@ARGV) {
@ARGV = split /\s+/, MacPerl::Ask('Arguments?');
}
A MacPerl script saved as a "droplet" will populate @ARGV with the full
pathnames of the files dropped onto the script.
Mac users can run programs under a type of command line interface under
MPW (Macintosh Programmer's Workshop, a free development environment
from Apple). MacPerl was first introduced as an MPW tool, and MPW can
be used like a shell:
perl myscript.plx some arguments
ToolServer is another app from Apple that provides access to MPW tools
from MPW and the MacPerl app, which allows MacPerl programs to use
"system", backticks, and piped "open".
"Mac OS" is the proper name for the operating system, but the value in
$^O is "MacOS". To determine architecture, version, or whether the
application or MPW tool version is running, check:
$is_app = $MacPerl::Version =~ /App/;
$is_tool = $MacPerl::Version =~ /MPW/;
($version) = $MacPerl::Version =~ /^(\S+)/;
$is_ppc = $MacPerl::Architecture eq 'MacPPC';
$is_68k = $MacPerl::Architecture eq 'Mac68K';
Mac OS X, based on NeXT's OpenStep OS, runs MacPerl natively, under the
"Classic" environment. There is no "Carbon" version of MacPerl to run
under the primary Mac OS X environment. Mac OS X and its Open Source
version, Darwin, both run Unix perl natively.
Also see:
o MacPerl Development, http://dev.macperl.org/ .
o The MacPerl Pages, http://www.macperl.com/ .
o The MacPerl mailing lists, http://lists.perl.org/ .
o MPW, ftp://ftp.apple.com/developer/Tool_Chest/Core_Mac_OS_Tools/
VMS
Perl on VMS is discussed in perlvms in the perl distribution. Perl on
VMS can accept either VMS- or Unix-style file specifications as in
either of the following:
$ perl -ne "print if /perl_setup/i" SYS$LOGIN:LOGIN.COM
$ perl -ne "print if /perl_setup/i" /sys$login/login.com
but not a mixture of both as in:
$ perl -ne "print if /perl_setup/i" sys$login:/login.com
Can't open sys$login:/login.com: file specification syntax error
Interacting with Perl from the Digital Command Language (DCL) shell
often requires a different set of quotation marks than Unix shells do.
For example:
$ perl -e "print ""Hello, world.\n"""
Hello, world.
There are several ways to wrap your perl scripts in DCL .COM files, if
you are so inclined. For example:
$ write sys$output "Hello from DCL!"
$ if p1 .eqs. ""
$ then perl -x 'f$environment("PROCEDURE")
$ else perl -x - 'p1 'p2 'p3 'p4 'p5 'p6 'p7 'p8
$ deck/dollars="__END__"
#!/usr/bin/perl
print "Hello from Perl!\n";
__END__
$ endif
Do take care with "$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT" if your
perl-in-DCL script expects to do things like "$read = <STDIN>;".
Filenames are in the format "name.extension;version". The maximum
length for filenames is 39 characters, and the maximum length for
extensions is also 39 characters. Version is a number from 1 to 32767.
Valid characters are "/[A-Z0-9$_-]/".
VMS's RMS filesystem is case-insensitive and does not preserve case.
"readdir" returns lowercased filenames, but specifying a file for open-
ing remains case-insensitive. Files without extensions have a trailing
period on them, so doing a "readdir" with a file named A.;5 will return
a. (though that file could be opened with "open(FH, 'A')").
RMS had an eight level limit on directory depths from any rooted logi-
cal (allowing 16 levels overall) prior to VMS 7.2. Hence
"PERL_ROOT:[LIB.2.3.4.5.6.7.8]" is a valid directory specification but
"PERL_ROOT:[LIB.2.3.4.5.6.7.8.9]" is not. Makefile.PL authors might
have to take this into account, but at least they can refer to the for-
mer as "/PERL_ROOT/lib/2/3/4/5/6/7/8/".
The VMS::Filespec module, which gets installed as part of the build
process on VMS, is a pure Perl module that can easily be installed on
non-VMS platforms and can be helpful for conversions to and from RMS
native formats.
What "\n" represents depends on the type of file opened. It usually
represents "\012" but it could also be "\015", "\012", "\015\012",
"\000", "\040", or nothing depending on the file organization and
record format. The VMS::Stdio module provides access to the special
fopen() requirements of files with unusual attributes on VMS.
TCP/IP stacks are optional on VMS, so socket routines might not be
implemented. UDP sockets may not be supported.
The value of $^O on OpenVMS is "VMS". To determine the architecture
that you are running on without resorting to loading all of %Config you
can examine the content of the @INC array like so:
if (grep(/VMS_AXP/, @INC)) {
print "I'm on Alpha!\n";
} elsif (grep(/VMS_VAX/, @INC)) {
print "I'm on VAX!\n";
} else {
print "I'm not so sure about where $^O is...\n";
}
On VMS, perl determines the UTC offset from the "SYS$TIMEZONE_DIFFEREN-
TIAL" logical name. Although the VMS epoch began at 17-NOV-1858
00:00:00.00, calls to "localtime" are adjusted to count offsets from
01-JAN-1970 00:00:00.00, just like Unix.
Also see:
o README.vms (installed as README_vms), perlvms
o vmsperl list, majordomo@perl.org
(Put the words "subscribe vmsperl" in message body.)
o vmsperl on the web, http://www.sidhe.org/vmsperl/index.html
VOS
Perl on VOS is discussed in README.vos in the perl distribution
(installed as perlvos). Perl on VOS can accept either VOS- or Unix-
style file specifications as in either of the following:
C<< $ perl -ne "print if /perl_setup/i" >system>notices >>
C<< $ perl -ne "print if /perl_setup/i" /system/notices >>
or even a mixture of both as in:
C<< $ perl -ne "print if /perl_setup/i" >system/notices >>
Even though VOS allows the slash character to appear in object names,
because the VOS port of Perl interprets it as a pathname delimiting
character, VOS files, directories, or links whose names contain a slash
character cannot be processed. Such files must be renamed before they
can be processed by Perl. Note that VOS limits file names to 32 or
fewer characters.
Perl on VOS can be built using two different compilers and two differ-
ent versions of the POSIX runtime. The recommended method for building
full Perl is with the GNU C compiler and the generally-available ver-
sion of VOS POSIX support. See README.vos (installed as perlvos) for
restrictions that apply when Perl is built using the VOS Standard C
compiler or the alpha version of VOS POSIX support.
The value of $^O on VOS is "VOS". To determine the architecture that
you are running on without resorting to loading all of %Config you can
examine the content of the @INC array like so:
if ($^O =~ /VOS/) {
print "I'm on a Stratus box!\n";
} else {
print "I'm not on a Stratus box!\n";
die;
}
if (grep(/860/, @INC)) {
print "This box is a Stratus XA/R!\n";
} elsif (grep(/7100/, @INC)) {
print "This box is a Stratus HP 7100 or 8xxx!\n";
} elsif (grep(/8000/, @INC)) {
print "This box is a Stratus HP 8xxx!\n";
} else {
print "This box is a Stratus 68K!\n";
}
Also see:
o README.vos (installed as perlvos)
o The VOS mailing list.
There is no specific mailing list for Perl on VOS. You can post
comments to the comp.sys.stratus newsgroup, or subscribe to the
general Stratus mailing list. Send a letter with "subscribe
Info-Stratus" in the message body to majordomo@list.stratagy.com.
o VOS Perl on the web at http://ftp.stra-
tus.com/pub/vos/posix/posix.html
EBCDIC Platforms
Recent versions of Perl have been ported to platforms such as OS/400 on
AS/400 minicomputers as well as OS/390, VM/ESA, and BS2000 for S/390
Mainframes. Such computers use EBCDIC character sets internally (usu-
ally Character Code Set ID 0037 for OS/400 and either 1047 or POSIX-BC
for S/390 systems). On the mainframe perl currently works under the
"Unix system services for OS/390" (formerly known as OpenEdition),
VM/ESA OpenEdition, or the BS200 POSIX-BC system (BS2000 is supported
in perl 5.6 and greater). See perlos390 for details. Note that for
OS/400 there is also a port of Perl 5.8.1/5.9.0 or later to the PASE
which is ASCII-based (as opposed to ILE which is EBCDIC-based), see
perlos400.
As of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix sub-
systems do not support the "#!" shebang trick for script invocation.
Hence, on OS/390 and VM/ESA perl scripts can be executed with a header
similar to the following simple script:
: # use perl
eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}'
if 0;
#!/usr/local/bin/perl # just a comment really
print "Hello from perl!\n";
OS/390 will support the "#!" shebang trick in release 2.8 and beyond.
Calls to "system" and backticks can use POSIX shell syntax on all S/390
systems.
On the AS/400, if PERL5 is in your library list, you may need to wrap
your perl scripts in a CL procedure to invoke them like so:
BEGIN
CALL PGM(PERL5/PERL) PARM('/QOpenSys/hello.pl')
ENDPGM
This will invoke the perl script hello.pl in the root of the QOpenSys
file system. On the AS/400 calls to "system" or backticks must use CL
syntax.
On these platforms, bear in mind that the EBCDIC character set may have
an effect on what happens with some perl functions (such as "chr",
"pack", "print", "printf", "ord", "sort", "sprintf", "unpack"), as well
as bit-fiddling with ASCII constants using operators like "^", "&" and
"|", not to mention dealing with socket interfaces to ASCII computers
(see "Newlines").
Fortunately, most web servers for the mainframe will correctly trans-
late the "\n" in the following statement to its ASCII equivalent ("\r"
is the same under both Unix and OS/390 & VM/ESA):
print "Content-type: text/html\r\n\r\n";
The values of $^O on some of these platforms includes:
uname $^O $Config{'archname'}
--------------------------------------------
OS/390 os390 os390
OS400 os400 os400
POSIX-BC posix-bc BS2000-posix-bc
VM/ESA vmesa vmesa
Some simple tricks for determining if you are running on an EBCDIC
platform could include any of the following (perhaps all):
if ("\t" eq "\05") { print "EBCDIC may be spoken here!\n"; }
if (ord('A') == 193) { print "EBCDIC may be spoken here!\n"; }
if (chr(169) eq 'z') { print "EBCDIC may be spoken here!\n"; }
One thing you may not want to rely on is the EBCDIC encoding of punctu-
ation characters since these may differ from code page to code page
(and once your module or script is rumoured to work with EBCDIC, folks
will want it to work with all EBCDIC character sets).
Also see:
o perlos390, README.os390, perlbs2000, README.vmesa, perlebcdic.
o The perl-mvs@perl.org list is for discussion of porting issues as
well as general usage issues for all EBCDIC Perls. Send a message
body of "subscribe perl-mvs" to majordomo@perl.org.
o AS/400 Perl information at http://as400.rochester.ibm.com/ as well
as on CPAN in the ports/ directory.
Acorn RISC OS
Because Acorns use ASCII with newlines ("\n") in text files as "\012"
like Unix, and because Unix filename emulation is turned on by default,
most simple scripts will probably work "out of the box". The native
filesystem is modular, and individual filesystems are free to be case-
sensitive or insensitive, and are usually case-preserving. Some native
filesystems have name length limits, which file and directory names are
silently truncated to fit. Scripts should be aware that the standard
filesystem currently has a name length limit of 10 characters, with up
to 77 items in a directory, but other filesystems may not impose such
limitations.
Native filenames are of the form
Filesystem#Special_Field::DiskName.$.Directory.Directory.File
where
Special_Field is not usually present, but may contain . and $ .
Filesystem =~ m|[A-Za-z0-9_]|
DsicName =~ m|[A-Za-z0-9_/]|
$ represents the root directory
. is the path separator
@ is the current directory (per filesystem but machine global)
^ is the parent directory
Directory and File =~ m|[^\0- "\.\$\%\&:\@\\^\|\177]+|
The default filename translation is roughly "tr|/.|./|;"
Note that ""ADFS::HardDisk.$.File" ne 'ADFS::HardDisk.$.File'" and that
the second stage of "$" interpolation in regular expressions will fall
foul of the $. if scripts are not careful.
Logical paths specified by system variables containing comma-separated
search lists are also allowed; hence "System:Modules" is a valid file-
name, and the filesystem will prefix "Modules" with each section of
"System$Path" until a name is made that points to an object on disk.
Writing to a new file "System:Modules" would be allowed only if "Sys-
tem$Path" contains a single item list. The filesystem will also expand
system variables in filenames if enclosed in angle brackets, so "<Sys-
tem$Dir>.Modules" would look for the file "$ENV{'System$Dir'} . 'Mod-
ules'". The obvious implication of this is that fully qualified file-
names can start with "<>" and should be protected when "open" is used
for input.
Because "." was in use as a directory separator and filenames could not
be assumed to be unique after 10 characters, Acorn implemented the C
compiler to strip the trailing ".c" ".h" ".s" and ".o" suffix from
filenames specified in source code and store the respective files in
subdirectories named after the suffix. Hence files are translated:
foo.h h.foo
C:foo.h C:h.foo (logical path variable)
sys/os.h sys.h.os (C compiler groks Unix-speak)
10charname.c c.10charname
10charname.o o.10charname
11charname_.c c.11charname (assuming filesystem truncates at 10)
The Unix emulation library's translation of filenames to native assumes
that this sort of translation is required, and it allows a user-defined
list of known suffixes that it will transpose in this fashion. This
may seem transparent, but consider that with these rules
"foo/bar/baz.h" and "foo/bar/h/baz" both map to "foo.bar.h.baz", and
that "readdir" and "glob" cannot and do not attempt to emulate the
reverse mapping. Other "."'s in filenames are translated to "/".
As implied above, the environment accessed through %ENV is global, and
the convention is that program specific environment variables are of
the form "Program$Name". Each filesystem maintains a current direc-
tory, and the current filesystem's current directory is the global cur-
rent directory. Consequently, sociable programs don't change the cur-
rent directory but rely on full pathnames, and programs (and Makefiles)
cannot assume that they can spawn a child process which can change the
current directory without affecting its parent (and everyone else for
that matter).
Because native operating system filehandles are global and are cur-
rently allocated down from 255, with 0 being a reserved value, the Unix
emulation library emulates Unix filehandles. Consequently, you can't
rely on passing "STDIN", "STDOUT", or "STDERR" to your children.
The desire of users to express filenames of the form "<Foo$Dir>.Bar" on
the command line unquoted causes problems, too: `` command output cap-
ture has to perform a guessing game. It assumes that a string
"<[^<>]+\$[^<>]>" is a reference to an environment variable, whereas
anything else involving "<" or ">" is redirection, and generally man-
ages to be 99% right. Of course, the problem remains that scripts can-
not rely on any Unix tools being available, or that any tools found
have Unix-like command line arguments.
Extensions and XS are, in theory, buildable by anyone using free tools.
In practice, many don't, as users of the Acorn platform are used to
binary distributions. MakeMaker does run, but no available make cur-
rently copes with MakeMaker's makefiles; even if and when this should
be fixed, the lack of a Unix-like shell will cause problems with make-
file rules, especially lines of the form "cd sdbm && make all", and
anything using quoting.
"RISC OS" is the proper name for the operating system, but the value in
$^O is "riscos" (because we don't like shouting).
Other perls
Perl has been ported to many platforms that do not fit into any of the
categories listed above. Some, such as AmigaOS, Atari MiNT, BeOS, HP
MPE/iX, QNX, Plan 9, and VOS, have been well-integrated into the stan-
dard Perl source code kit. You may need to see the ports/ directory on
CPAN for information, and possibly binaries, for the likes of: aos,
Atari ST, lynxos, riscos, Novell Netware, Tandem Guardian, etc. (Yes,
we know that some of these OSes may fall under the Unix category, but
we are not a standards body.)
Some approximate operating system names and their $^O values in the
"OTHER" category include:
OS $^O $Config{'archname'}
------------------------------------------
Amiga DOS amigaos m68k-amigos
BeOS beos
MPE/iX mpeix PA-RISC1.1
See also:
o Amiga, README.amiga (installed as perlamiga).
o Atari, README.mint and Guido Flohr's web page
http://stud.uni-sb.de/~gufl0000/
o Be OS, README.beos
o HP 300 MPE/iX, README.mpeix and Mark Bixby's web page
http://www.bixby.org/mark/perlix.html
o A free perl5-based PERL.NLM for Novell Netware is available in pre-
compiled binary and source code form from http://www.novell.com/ as
well as from CPAN.
o Plan 9, README.plan9
FUNCTION IMPLEMENTATIONS
Listed below are functions that are either completely unimplemented or
else have been implemented differently on various platforms. Following
each description will be, in parentheses, a list of platforms that the
description applies to.
The list may well be incomplete, or even wrong in some places. When in
doubt, consult the platform-specific README files in the Perl source
distribution, and any other documentation resources accompanying a
given port.
Be aware, moreover, that even among Unix-ish systems there are varia-
tions.
For many functions, you can also query %Config, exported by default
from the Config module. For example, to check whether the platform has
the "lstat" call, check $Config{d_lstat}. See Config for a full
description of available variables.
Alphabetical Listing of Perl Functions
-X "-r", "-w", and "-x" have a limited meaning only; directories
and applications are executable, and there are no uid/gid con-
siderations. "-o" is not supported. (Mac OS)
"-r", "-w", "-x", and "-o" tell whether the file is accessible,
which may not reflect UIC-based file protections. (VMS)
"-s" returns the size of the data fork, not the total size of
data fork plus resource fork. (Mac OS).
"-s" by name on an open file will return the space reserved on
disk, rather than the current extent. "-s" on an open filehan-
dle returns the current size. (RISC OS)
"-R", "-W", "-X", "-O" are indistinguishable from "-r", "-w",
"-x", "-o". (Mac OS, Win32, VMS, RISC OS)
"-b", "-c", "-k", "-g", "-p", "-u", "-A" are not implemented.
(Mac OS)
"-g", "-k", "-l", "-p", "-u", "-A" are not particularly mean-
ingful. (Win32, VMS, RISC OS)
"-d" is true if passed a device spec without an explicit direc-
tory. (VMS)
"-T" and "-B" are implemented, but might misclassify Mac text
files with foreign characters; this is the case will all plat-
forms, but may affect Mac OS often. (Mac OS)
"-x" (or "-X") determine if a file ends in one of the exe-
cutable suffixes. "-S" is meaningless. (Win32)
"-x" (or "-X") determine if a file has an executable file type.
(RISC OS)
atan2 Y,X
Due to issues with various CPUs, math libraries, compilers, and
standards, results for "atan2()" may vary depending on any com-
bination of the above. Perl attempts to conform to the Open
Group/IEEE standards for the results returned from "atan2()",
but cannot force the issue if the system Perl is run on does
not allow it. (Tru64, HP-UX 10.20)
The current version of the standards for "atan2()" is available
at <http://www.opengroup.org/onlinepubs/009695399/func-
tions/atan2.html>.
atan2 Due to issues with various CPUs, math libraries, compilers, and
standards, results for "atan2()" may vary depending on any com-
bination of the above. Perl attempts to conform to the Open
Group/IEEE standards for the results returned from "atan2()",
but cannot force the issue if the system Perl is run on does
not allow it. (Tru64, HP-UX 10.20)
The current version of the standards for "atan2()" is available
at <http://www.opengroup.org/onlinepubs/009695399/func-
tions/atan2.html>.
binmode Meaningless. (Mac OS, RISC OS)
Reopens file and restores pointer; if function fails, underly-
ing filehandle may be closed, or pointer may be in a different
position. (VMS)
The value returned by "tell" may be affected after the call,
and the filehandle may be flushed. (Win32)
chmod Only limited meaning. Disabling/enabling write permission is
mapped to locking/unlocking the file. (Mac OS)
Only good for changing "owner" read-write access, "group", and
"other" bits are meaningless. (Win32)
Only good for changing "owner" and "other" read-write access.
(RISC OS)
Access permissions are mapped onto VOS access-control list
changes. (VOS)
The actual permissions set depend on the value of the "CYGWIN"
in the SYSTEM environment settings. (Cygwin)
chown Not implemented. (Mac OS, Win32, Plan 9, RISC OS, VOS)
Does nothing, but won't fail. (Win32)
chroot Not implemented. (Mac OS, Win32, VMS, Plan 9, RISC OS, VOS,
VM/ESA)
crypt May not be available if library or source was not provided when
building perl. (Win32)
Not implemented. (VOS)
dbmclose
Not implemented. (VMS, Plan 9, VOS)
dbmopen Not implemented. (VMS, Plan 9, VOS)
dump Not useful. (Mac OS, RISC OS)
Not implemented. (Win32)
Invokes VMS debugger. (VMS)
exec Not implemented. (Mac OS)
Implemented via Spawn. (VM/ESA)
Does not automatically flush output handles on some platforms.
(SunOS, Solaris, HP-UX)
exit Emulates UNIX exit() (which considers "exit 1" to indicate an
error) by mapping the 1 to SS$_ABORT (44). This behavior may
be overridden with the pragma "use vmsish 'exit'". As with the
CRTL's exit() function, "exit 0" is also mapped to an exit sta-
tus of SS$_NORMAL (1); this mapping cannot be overridden. Any
other argument to exit() is used directly as Perl's exit sta-
tus. (VMS)
fcntl Not implemented. (Win32, VMS)
flock Not implemented (Mac OS, VMS, RISC OS, VOS).
Available only on Windows NT (not on Windows 95). (Win32)
fork Not implemented. (Mac OS, AmigaOS, RISC OS, VOS, VM/ESA, VMS)
Emulated using multiple interpreters. See perlfork. (Win32)
Does not automatically flush output handles on some platforms.
(SunOS, Solaris, HP-UX)
getlogin
Not implemented. (Mac OS, RISC OS)
getpgrp Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS)
getppid Not implemented. (Mac OS, Win32, RISC OS)
getpriority
Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS, VM/ESA)
getpwnam
Not implemented. (Mac OS, Win32)
Not useful. (RISC OS)
getgrnam
Not implemented. (Mac OS, Win32, VMS, RISC OS)
getnetbyname
Not implemented. (Mac OS, Win32, Plan 9)
getpwuid
Not implemented. (Mac OS, Win32)
Not useful. (RISC OS)
getgrgid
Not implemented. (Mac OS, Win32, VMS, RISC OS)
getnetbyaddr
Not implemented. (Mac OS, Win32, Plan 9)
getprotobynumber
Not implemented. (Mac OS)
getservbyport
Not implemented. (Mac OS)
getpwent
Not implemented. (Mac OS, Win32, VM/ESA)
getgrent
Not implemented. (Mac OS, Win32, VMS, VM/ESA)
gethostbyname
"gethostbyname('localhost')" does not work everywhere: you may
have to use "gethostbyname('127.0.0.1')". (Mac OS, Irix 5)
gethostent
Not implemented. (Mac OS, Win32)
getnetent
Not implemented. (Mac OS, Win32, Plan 9)
getprotoent
Not implemented. (Mac OS, Win32, Plan 9)
getservent
Not implemented. (Win32, Plan 9)
sethostent
Not implemented. (Mac OS, Win32, Plan 9, RISC OS)
setnetent
Not implemented. (Mac OS, Win32, Plan 9, RISC OS)
setprotoent
Not implemented. (Mac OS, Win32, Plan 9, RISC OS)
setservent
Not implemented. (Plan 9, Win32, RISC OS)
endpwent
Not implemented. (Mac OS, MPE/iX, VM/ESA, Win32)
endgrent
Not implemented. (Mac OS, MPE/iX, RISC OS, VM/ESA, VMS, Win32)
endhostent
Not implemented. (Mac OS, Win32)
endnetent
Not implemented. (Mac OS, Win32, Plan 9)
endprotoent
Not implemented. (Mac OS, Win32, Plan 9)
endservent
Not implemented. (Plan 9, Win32)
getsockopt SOCKET,LEVEL,OPTNAME
Not implemented. (Plan 9)
glob This operator is implemented via the File::Glob extension on
most platforms. See File::Glob for portability information.
gmtime Same portability caveats as localtime.
ioctl FILEHANDLE,FUNCTION,SCALAR
Not implemented. (VMS)
Available only for socket handles, and it does what the ioctl-
socket() call in the Winsock API does. (Win32)
Available only for socket handles. (RISC OS)
kill "kill(0, LIST)" is implemented for the sake of taint checking;
use with other signals is unimplemented. (Mac OS)
Not implemented, hence not useful for taint checking. (RISC OS)
"kill()" doesn't have the semantics of "raise()", i.e. it
doesn't send a signal to the identified process like it does on
Unix platforms. Instead "kill($sig, $pid)" terminates the
process identified by $pid, and makes it exit immediately with
exit status $sig. As in Unix, if $sig is 0 and the specified
process exists, it returns true without actually terminating
it. (Win32)
link Not implemented. (Mac OS, MPE/iX, VMS, RISC OS)
Link count not updated because hard links are not quite that
hard (They are sort of half-way between hard and soft links).
(AmigaOS)
Hard links are implemented on Win32 (Windows NT and Windows
2000) under NTFS only.
localtime
Because Perl currently relies on the native standard C local-
time() function, it is only safe to use times between 0 and
(2**31)-1. Times outside this range may result in unexpected
behavior depending on your operating system's implementation of
localtime().
lstat Not implemented. (VMS, RISC OS)
Return values (especially for device and inode) may be bogus.
(Win32)
msgctl
msgget
msgsnd
msgrcv Not implemented. (Mac OS, Win32, VMS, Plan 9, RISC OS, VOS)
open The "|" variants are supported only if ToolServer is installed.
(Mac OS)
open to "|-" and "-|" are unsupported. (Mac OS, Win32, RISC OS)
Opening a process does not automatically flush output handles
on some platforms. (SunOS, Solaris, HP-UX)
pipe Very limited functionality. (MiNT)
readlink
Not implemented. (Win32, VMS, RISC OS)
rename Can't move directories between directories on different logical
volumes. (Win32)
select Only implemented on sockets. (Win32, VMS)
Only reliable on sockets. (RISC OS)
Note that the "select FILEHANDLE" form is generally portable.
semctl
semget
semop Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS)
setgrent
Not implemented. (Mac OS, MPE/iX, VMS, Win32, RISC OS)
setpgrp Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS)
setpriority
Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS)
setpwent
Not implemented. (Mac OS, MPE/iX, Win32, RISC OS)
setsockopt
Not implemented. (Plan 9)
shmctl
shmget
shmread
shmwrite
Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS)
sockatmark
A relatively recent addition to socket functions, may not be
implemented even in UNIX platforms.
socketpair
Not implemented. (Win32, VMS, RISC OS, VOS, VM/ESA)
stat Platforms that do not have rdev, blksize, or blocks will return
these as '', so numeric comparison or manipulation of these
fields may cause 'not numeric' warnings.
mtime and atime are the same thing, and ctime is creation time
instead of inode change time. (Mac OS).
ctime not supported on UFS (Mac OS X).
ctime is creation time instead of inode change time (Win32).
device and inode are not meaningful. (Win32)
device and inode are not necessarily reliable. (VMS)
mtime, atime and ctime all return the last modification time.
Device and inode are not necessarily reliable. (RISC OS)
dev, rdev, blksize, and blocks are not available. inode is not
meaningful and will differ between stat calls on the same file.
(os2)
some versions of cygwin when doing a stat("foo") and if not
finding it may then attempt to stat("foo.exe") (Cygwin)
symlink Not implemented. (Win32, VMS, RISC OS)
syscall Not implemented. (Mac OS, Win32, VMS, RISC OS, VOS, VM/ESA)
sysopen The traditional "0", "1", and "2" MODEs are implemented with
different numeric values on some systems. The flags exported
by "Fcntl" (O_RDONLY, O_WRONLY, O_RDWR) should work everywhere
though. (Mac OS, OS/390, VM/ESA)
system In general, do not assume the UNIX/POSIX semantics that you can
shift $? right by eight to get the exit value, or that "$? &
127" would give you the number of the signal that terminated
the program, or that "$? & 128" would test true if the program
was terminated by a coredump. Instead, use the POSIX W*()
interfaces: for example, use WIFEXITED($?) and WEXITVALUE($?)
to test for a normal exit and the exit value, WIFSIGNALED($?)
and WTERMSIG($?) for a signal exit and the signal. Core dump-
ing is not a portable concept, so there's no portable way to
test for that.
Only implemented if ToolServer is installed. (Mac OS)
As an optimization, may not call the command shell specified in
$ENV{PERL5SHELL}. "system(1, @args)" spawns an external
process and immediately returns its process designator, without
waiting for it to terminate. Return value may be used subse-
quently in "wait" or "waitpid". Failure to spawn() a subpro-
cess is indicated by setting $? to "255 << 8". $? is set in a
way compatible with Unix (i.e. the exitstatus of the subprocess
is obtained by "$? >> 8", as described in the documentation).
(Win32)
There is no shell to process metacharacters, and the native
standard is to pass a command line terminated by "\n" "\r" or
"\0" to the spawned program. Redirection such as "> foo" is
performed (if at all) by the run time library of the spawned
program. "system" list will call the Unix emulation library's
"exec" emulation, which attempts to provide emulation of the
stdin, stdout, stderr in force in the parent, providing the
child program uses a compatible version of the emulation
library. scalar will call the native command line direct and
no such emulation of a child Unix program will exists. Mileage
will vary. (RISC OS)
Far from being POSIX compliant. Because there may be no under-
lying /bin/sh tries to work around the problem by forking and
execing the first token in its argument string. Handles basic
redirection ("<" or ">") on its own behalf. (MiNT)
Does not automatically flush output handles on some platforms.
(SunOS, Solaris, HP-UX)
The return value is POSIX-like (shifted up by 8 bits), which
only allows room for a made-up value derived from the severity
bits of the native 32-bit condition code (unless overridden by
"use vmsish 'status'"). For more details see "$?" in perlvms.
(VMS)
times Only the first entry returned is nonzero. (Mac OS)
"cumulative" times will be bogus. On anything other than Win-
dows NT or Windows 2000, "system" time will be bogus, and
"user" time is actually the time returned by the clock() func-
tion in the C runtime library. (Win32)
Not useful. (RISC OS)
truncate
Not implemented. (Older versions of VMS)
Truncation to zero-length only. (VOS)
If a FILEHANDLE is supplied, it must be writable and opened in
append mode (i.e., use "open(FH, '>>filename')" or
"sysopen(FH,...,O_APPEND|O_RDWR)". If a filename is supplied,
it should not be held open elsewhere. (Win32)
umask Returns undef where unavailable, as of version 5.005.
"umask" works but the correct permissions are set only when the
file is finally closed. (AmigaOS)
utime Only the modification time is updated. (BeOS, Mac OS, VMS,
RISC OS)
May not behave as expected. Behavior depends on the C runtime
library's implementation of utime(), and the filesystem being
used. The FAT filesystem typically does not support an "access
time" field, and it may limit timestamps to a granularity of
two seconds. (Win32)
wait
waitpid Not implemented. (Mac OS, VOS)
Can only be applied to process handles returned for processes
spawned using "system(1, ...)" or pseudo processes created with
"fork()". (Win32)
Not useful. (RISC OS)
Supported Platforms
As of September 2003 (the Perl release 5.8.1), the following platforms
are able to build Perl from the standard source code distribution
available at http://www.cpan.org/src/index.html
AIX
BeOS
BSD/OS (BSDi)
Cygwin
DG/UX
DOS DJGPP 1)
DYNIX/ptx
EPOC R5
FreeBSD
HI-UXMPP (Hitachi) (5.8.0 worked but we didn't know it)
HP-UX
IRIX
Linux
LynxOS
Mac OS Classic
Mac OS X (Darwin)
MPE/iX
NetBSD
NetWare
NonStop-UX
ReliantUNIX (formerly SINIX)
OpenBSD
OpenVMS (formerly VMS)
Open UNIX (Unixware) (since Perl 5.8.1/5.9.0)
OS/2
OS/400 (using the PASE) (since Perl 5.8.1/5.9.0)
PowerUX
POSIX-BC (formerly BS2000)
QNX
Solaris
SunOS 4
SUPER-UX (NEC)
SVR4
Tru64 UNIX (formerly DEC OSF/1, Digital UNIX)
UNICOS
UNICOS/mk
UTS
VOS
Win95/98/ME/2K/XP 2)
WinCE
z/OS (formerly OS/390)
VM/ESA
1) in DOS mode either the DOS or OS/2 ports can be used
2) compilers: Borland, MinGW (GCC), VC6
The following platforms worked with the previous releases (5.6 and
5.7), but we did not manage either to fix or to test these in time for
the 5.8.1 release. There is a very good chance that many of these will
work fine with the 5.8.1.
DomainOS
Hurd
MachTen
PowerMAX
SCO SV
Unixware
Windows 3.1
Known to be broken for 5.8.0 and 5.8.1 (but 5.6.1 and 5.7.2 can be
used):
AmigaOS
The following platforms have been known to build Perl from source in
the past (5.005_03 and earlier), but we haven't been able to verify
their status for the current release, either because the hardware/soft-
ware platforms are rare or because we don't have an active champion on
these platforms--or both. They used to work, though, so go ahead and
try compiling them, and let perlbug@perl.org of any trouble.
3b1
A/UX
ConvexOS
CX/UX
DC/OSx
DDE SMES
DOS EMX
Dynix
EP/IX
ESIX
FPS
GENIX
Greenhills
ISC
MachTen 68k
MiNT
MPC
NEWS-OS
NextSTEP
OpenSTEP
Opus
Plan 9
RISC/os
SCO ODT/OSR
Stellar
SVR2
TI1500
TitanOS
Ultrix
Unisys Dynix
The following platforms have their own source code distributions and
binaries available via http://www.cpan.org/ports/
Perl release
OS/400 (ILE) 5.005_02
Tandem Guardian 5.004
The following platforms have only binaries available via
http://www.cpan.org/ports/index.html :
Perl release
Acorn RISCOS 5.005_02
AOS 5.002
LynxOS 5.004_02
Although we do suggest that you always build your own Perl from the
source code, both for maximal configurability and for security, in case
you are in a hurry you can check http://www.cpan.org/ports/index.html
for binary distributions.
SEE ALSO
perlaix, perlamiga, perlapollo, perlbeos, perlbs2000, perlce, perlcyg-
win, perldgux, perldos, perlepoc, perlebcdic, perlfreebsd, perlhurd,
perlhpux, perlirix, perlmachten, perlmacos, perlmacosx, perlmint,
perlmpeix, perlnetware, perlos2, perlos390, perlos400, perlplan9, per-
lqnx, perlsolaris, perltru64, perlunicode, perlvmesa, perlvms, perlvos,
perlwin32, and Win32.
AUTHORS / CONTRIBUTORS
Abigail <abigail@foad.org>, Charles Bailey <bailey@newman.upenn.edu>,
Graham Barr <gbarr@pobox.com>, Tom Christiansen <tchrist@perl.com>,
Nicholas Clark <nick@ccl4.org>, Thomas Dorner <Thomas.Dorner@start.de>,
Andy Dougherty <doughera@lafayette.edu>, Dominic Dunlop <domo@com-
puter.org>, Neale Ferguson <neale@vma.tabnsw.com.au>, David J. Fiander
<davidf@mks.com>, Paul Green <Paul_Green@stratus.com>, M.J.T. Guy
<mjtg@cam.ac.uk>, Jarkko Hietaniemi <jhi@iki.fi>, Luther Huffman
<lutherh@stratcom.com>, Nick Ing-Simmons <nick@ing-simmons.net>,
Andreas J. Koenig <a.koenig@mind.de>, Markus Laker <mlaker@con-
tax.co.uk>, Andrew M. Langmead <aml@world.std.com>, Larry Moore
<ljmoore@freespace.net>, Paul Moore <Paul.Moore@uk.origin-it.com>,
Chris Nandor <pudge@pobox.com>, Matthias Neeracher <neeracher@mac.com>,
Philip Newton <pne@cpan.org>, Gary Ng <71564.1743@CompuServe.COM>, Tom
Phoenix <rootbeer@teleport.com>, Andre Pirard <A.Pirard@ulg.ac.be>,
Peter Prymmer <pvhp@forte.com>, Hugo van der Sanden
<hv@crypt0.demon.co.uk>, Gurusamy Sarathy <gsar@activestate.com>, Paul
J. Schinder <schinder@pobox.com>, Michael G Schwern <schw-
ern@pobox.com>, Dan Sugalski <dan@sidhe.org>, Nathan Torkington
<gnat@frii.com>.
perl v5.8.8 2006-06-14 PERLPORT(1)
Man(1) output converted with
man2html