3 perlport - Writing portable Perl
7 Perl runs on numerous operating systems. While most of them share
8 much in common, they also have their own unique features.
10 This document is meant to help you to find out what constitutes portable
11 Perl code. That way once you make a decision to write portably,
12 you know where the lines are drawn, and you can stay within them.
14 There is a tradeoff between taking full advantage of one particular
15 type of computer and taking advantage of a full range of them.
16 Naturally, as you broaden your range and become more diverse, the
17 common factors drop, and you are left with an increasingly smaller
18 area of common ground in which you can operate to accomplish a
19 particular task. Thus, when you begin attacking a problem, it is
20 important to consider under which part of the tradeoff curve you
21 want to operate. Specifically, you must decide whether it is
22 important that the task that you are coding has the full generality
23 of being portable, or whether to just get the job done right now.
24 This is the hardest choice to be made. The rest is easy, because
25 Perl provides many choices, whichever way you want to approach your
28 Looking at it another way, writing portable code is usually about
29 willfully limiting your available choices. Naturally, it takes
30 discipline and sacrifice to do that. The product of portability
31 and convenience may be a constant. You have been warned.
33 Be aware of two important points:
37 =item Not all Perl programs have to be portable
39 There is no reason you should not use Perl as a language to glue Unix
40 tools together, or to prototype a Macintosh application, or to manage the
41 Windows registry. If it makes no sense to aim for portability for one
42 reason or another in a given program, then don't bother.
44 =item Nearly all of Perl already I<is> portable
46 Don't be fooled into thinking that it is hard to create portable Perl
47 code. It isn't. Perl tries its level-best to bridge the gaps between
48 what's available on different platforms, and all the means available to
49 use those features. Thus almost all Perl code runs on any machine
50 without modification. But there are some significant issues in
51 writing portable code, and this document is entirely about those issues.
55 Here's the general rule: When you approach a task commonly done
56 using a whole range of platforms, think about writing portable
57 code. That way, you don't sacrifice much by way of the implementation
58 choices you can avail yourself of, and at the same time you can give
59 your users lots of platform choices. On the other hand, when you have to
60 take advantage of some unique feature of a particular platform, as is
61 often the case with systems programming (whether for Unix, Windows,
62 VMS, etc.), consider writing platform-specific code.
64 When the code will run on only two or three operating systems, you
65 may need to consider only the differences of those particular systems.
66 The important thing is to decide where the code will run and to be
67 deliberate in your decision.
69 The material below is separated into three main sections: main issues of
70 portability (L</"ISSUES">), platform-specific issues (L</"PLATFORMS">), and
71 built-in Perl functions that behave differently on various ports
72 (L</"FUNCTION IMPLEMENTATIONS">).
74 This information should not be considered complete; it includes possibly
75 transient information about idiosyncrasies of some of the ports, almost
76 all of which are in a state of constant evolution. Thus, this material
77 should be considered a perpetual work in progress
78 (C<< <IMG SRC="yellow_sign.gif" ALT="Under Construction"> >>).
84 In most operating systems, lines in files are terminated by newlines.
85 Just what is used as a newline may vary from OS to OS. Unix
86 traditionally uses C<\012>, one type of DOSish I/O uses C<\015\012>,
87 S<Mac OS> uses C<\015>, and z/OS uses C<\025>.
89 Perl uses C<\n> to represent the "logical" newline, where what is
90 logical may depend on the platform in use. In MacPerl, C<\n> always
91 means C<\015>. On EBCDIC platforms, C<\n> could be C<\025> or C<\045>.
92 In DOSish perls, C<\n> usually means C<\012>, but when
93 accessing a file in "text" mode, perl uses the C<:crlf> layer that
94 translates it to (or from) C<\015\012>, depending on whether you're
95 reading or writing. Unix does the same thing on ttys in canonical
96 mode. C<\015\012> is commonly referred to as CRLF.
98 To trim trailing newlines from text lines use
99 L<C<chomp>|perlfunc/chomp VARIABLE>. With default settings that function
100 looks for a trailing C<\n> character and thus trims in a portable way.
102 When dealing with binary files (or text files in binary mode) be sure
103 to explicitly set L<C<$E<sol>>|perlvar/$E<sol>> to the appropriate value for
104 your file format before using L<C<chomp>|perlfunc/chomp VARIABLE>.
106 Because of the "text" mode translation, DOSish perls have limitations in
107 using L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> and
108 L<C<tell>|perlfunc/tell FILEHANDLE> on a file accessed in "text" mode.
109 Stick to L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE>-ing to
110 locations you got from L<C<tell>|perlfunc/tell FILEHANDLE> (and no
111 others), and you are usually free to use
112 L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> and
113 L<C<tell>|perlfunc/tell FILEHANDLE> even in "text" mode. Using
114 L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> or
115 L<C<tell>|perlfunc/tell FILEHANDLE> or other file operations may be
116 non-portable. If you use L<C<binmode>|perlfunc/binmode FILEHANDLE> on a
117 file, however, you can usually
118 L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> and
119 L<C<tell>|perlfunc/tell FILEHANDLE> with arbitrary values safely.
121 A common misconception in socket programming is that S<C<\n eq \012>>
122 everywhere. When using protocols such as common Internet protocols,
123 C<\012> and C<\015> are called for specifically, and the values of
124 the logical C<\n> and C<\r> (carriage return) are not reliable.
126 print $socket "Hi there, client!\r\n"; # WRONG
127 print $socket "Hi there, client!\015\012"; # RIGHT
129 However, using C<\015\012> (or C<\cM\cJ>, or C<\x0D\x0A>) can be tedious
130 and unsightly, as well as confusing to those maintaining the code. As
131 such, the L<C<Socket>|Socket> module supplies the Right Thing for those
134 use Socket qw(:DEFAULT :crlf);
135 print $socket "Hi there, client!$CRLF" # RIGHT
137 When reading from a socket, remember that the default input record
138 separator L<C<$E<sol>>|perlvar/$E<sol>> is C<\n>, but robust socket code
139 will recognize as either C<\012> or C<\015\012> as end of line:
141 while (<$socket>) { # NOT ADVISABLE!
145 Because both CRLF and LF end in LF, the input record separator can
146 be set to LF and any CR stripped later. Better to write:
148 use Socket qw(:DEFAULT :crlf);
149 local($/) = LF; # not needed if $/ is already \012
152 s/$CR?$LF/\n/; # not sure if socket uses LF or CRLF, OK
153 # s/\015?\012/\n/; # same thing
156 This example is preferred over the previous one--even for Unix
157 platforms--because now any C<\015>'s (C<\cM>'s) are stripped out
158 (and there was much rejoicing).
160 Similarly, functions that return text data--such as a function that
161 fetches a web page--should sometimes translate newlines before
162 returning the data, if they've not yet been translated to the local
163 newline representation. A single line of code will often suffice:
165 $data =~ s/\015?\012/\n/g;
168 Some of this may be confusing. Here's a handy reference to the ASCII CR
169 and LF characters. You can print it out and stick it in your wallet.
171 LF eq \012 eq \x0A eq \cJ eq chr(10) eq ASCII 10
172 CR eq \015 eq \x0D eq \cM eq chr(13) eq ASCII 13
175 ---------------------------
178 \n * | LF | CRLF | CR |
179 \r * | CR | CR | LF |
180 ---------------------------
183 The Unix column assumes that you are not accessing a serial line
184 (like a tty) in canonical mode. If you are, then CR on input becomes
185 "\n", and "\n" on output becomes CRLF.
187 These are just the most common definitions of C<\n> and C<\r> in Perl.
188 There may well be others. For example, on an EBCDIC implementation
189 such as z/OS (OS/390) or OS/400 (using the ILE, the PASE is ASCII-based)
190 the above material is similar to "Unix" but the code numbers change:
192 LF eq \025 eq \x15 eq \cU eq chr(21) eq CP-1047 21
193 LF eq \045 eq \x25 eq chr(37) eq CP-0037 37
194 CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-1047 13
195 CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-0037 13
198 ----------------------
203 ----------------------
206 =head2 Numbers endianness and Width
208 Different CPUs store integers and floating point numbers in different
209 orders (called I<endianness>) and widths (32-bit and 64-bit being the
210 most common today). This affects your programs when they attempt to transfer
211 numbers in binary format from one CPU architecture to another,
212 usually either "live" via network connection, or by storing the
213 numbers to secondary storage such as a disk file or tape.
215 Conflicting storage orders make an utter mess out of the numbers. If a
216 little-endian host (Intel, VAX) stores 0x12345678 (305419896 in
217 decimal), a big-endian host (Motorola, Sparc, PA) reads it as
218 0x78563412 (2018915346 in decimal). Alpha and MIPS can be either:
219 Digital/Compaq used/uses them in little-endian mode; SGI/Cray uses
220 them in big-endian mode. To avoid this problem in network (socket)
221 connections use the L<C<pack>|perlfunc/pack TEMPLATE,LIST> and
222 L<C<unpack>|perlfunc/unpack TEMPLATE,EXPR> formats C<n> and C<N>, the
223 "network" orders. These are guaranteed to be portable.
225 As of Perl 5.10.0, you can also use the C<E<gt>> and C<E<lt>> modifiers
226 to force big- or little-endian byte-order. This is useful if you want
227 to store signed integers or 64-bit integers, for example.
229 You can explore the endianness of your platform by unpacking a
230 data structure packed in native format such as:
232 print unpack("h*", pack("s2", 1, 2)), "\n";
233 # '10002000' on e.g. Intel x86 or Alpha 21064 in little-endian mode
234 # '00100020' on e.g. Motorola 68040
236 If you need to distinguish between endian architectures you could use
237 either of the variables set like so:
239 $is_big_endian = unpack("h*", pack("s", 1)) =~ /01/;
240 $is_little_endian = unpack("h*", pack("s", 1)) =~ /^1/;
242 Differing widths can cause truncation even between platforms of equal
243 endianness. The platform of shorter width loses the upper parts of the
244 number. There is no good solution for this problem except to avoid
245 transferring or storing raw binary numbers.
247 One can circumnavigate both these problems in two ways. Either
248 transfer and store numbers always in text format, instead of raw
249 binary, or else consider using modules like
250 L<C<Data::Dumper>|Data::Dumper> and L<C<Storable>|Storable> (included as
251 of Perl 5.8). Keeping all data as text significantly simplifies matters.
253 =head2 Files and Filesystems
255 Most platforms these days structure files in a hierarchical fashion.
256 So, it is reasonably safe to assume that all platforms support the
257 notion of a "path" to uniquely identify a file on the system. How
258 that path is really written, though, differs considerably.
260 Although similar, file path specifications differ between Unix,
261 Windows, S<Mac OS>, OS/2, VMS, VOS, S<RISC OS>, and probably others.
262 Unix, for example, is one of the few OSes that has the elegant idea
263 of a single root directory.
265 DOS, OS/2, VMS, VOS, and Windows can work similarly to Unix with C</>
266 as path separator, or in their own idiosyncratic ways (such as having
267 several root directories and various "unrooted" device files such NIL:
270 S<Mac OS> 9 and earlier used C<:> as a path separator instead of C</>.
272 The filesystem may support neither hard links
273 (L<C<link>|perlfunc/link OLDFILE,NEWFILE>) nor symbolic links
274 (L<C<symlink>|perlfunc/symlink OLDFILE,NEWFILE>,
275 L<C<readlink>|perlfunc/readlink EXPR>,
276 L<C<lstat>|perlfunc/lstat FILEHANDLE>).
278 The filesystem may support neither access timestamp nor change
279 timestamp (meaning that about the only portable timestamp is the
280 modification timestamp), or one second granularity of any timestamps
281 (e.g. the FAT filesystem limits the time granularity to two seconds).
283 The "inode change timestamp" (the L<C<-C>|perlfunc/-X FILEHANDLE>
284 filetest) may really be the "creation timestamp" (which it is not in
287 VOS perl can emulate Unix filenames with C</> as path separator. The
288 native pathname characters greater-than, less-than, number-sign, and
289 percent-sign are always accepted.
291 S<RISC OS> perl can emulate Unix filenames with C</> as path
292 separator, or go native and use C<.> for path separator and C<:> to
293 signal filesystems and disk names.
295 Don't assume Unix filesystem access semantics: that read, write,
296 and execute are all the permissions there are, and even if they exist,
297 that their semantics (for example what do C<r>, C<w>, and C<x> mean on
298 a directory) are the Unix ones. The various Unix/POSIX compatibility
299 layers usually try to make interfaces like L<C<chmod>|perlfunc/chmod LIST>
300 work, but sometimes there simply is no good mapping.
302 The L<C<File::Spec>|File::Spec> modules provide methods to manipulate path
303 specifications and return the results in native format for each
304 platform. This is often unnecessary as Unix-style paths are
305 understood by Perl on every supported platform, but if you need to
306 produce native paths for a native utility that does not understand
307 Unix syntax, or if you are operating on paths or path components
308 in unknown (and thus possibly native) syntax, L<C<File::Spec>|File::Spec>
309 is your friend. Here are two brief examples:
311 use File::Spec::Functions;
312 chdir(updir()); # go up one directory
314 # Concatenate a path from its components
315 my $file = catfile(updir(), 'temp', 'file.txt');
316 # on Unix: '../temp/file.txt'
317 # on Win32: '..\temp\file.txt'
318 # on VMS: '[-.temp]file.txt'
320 In general, production code should not have file paths hardcoded.
321 Making them user-supplied or read from a configuration file is
322 better, keeping in mind that file path syntax varies on different
325 This is especially noticeable in scripts like Makefiles and test suites,
326 which often assume C</> as a path separator for subdirectories.
328 Also of use is L<C<File::Basename>|File::Basename> from the standard
329 distribution, which splits a pathname into pieces (base filename, full
330 path to directory, and file suffix).
332 Even when on a single platform (if you can call Unix a single platform),
333 remember not to count on the existence or the contents of particular
334 system-specific files or directories, like F</etc/passwd>,
335 F</etc/sendmail.conf>, F</etc/resolv.conf>, or even F</tmp/>. For
336 example, F</etc/passwd> may exist but not contain the encrypted
337 passwords, because the system is using some form of enhanced security.
338 Or it may not contain all the accounts, because the system is using NIS.
339 If code does need to rely on such a file, include a description of the
340 file and its format in the code's documentation, then make it easy for
341 the user to override the default location of the file.
343 Don't assume a text file will end with a newline. They should,
346 Do not have two files or directories of the same name with different
347 case, like F<test.pl> and F<Test.pl>, as many platforms have
348 case-insensitive (or at least case-forgiving) filenames. Also, try
349 not to have non-word characters (except for C<.>) in the names, and
350 keep them to the 8.3 convention, for maximum portability, onerous a
351 burden though this may appear.
353 Likewise, when using the L<C<AutoSplit>|AutoSplit> module, try to keep
354 your functions to 8.3 naming and case-insensitive conventions; or, at the
355 least, make it so the resulting files have a unique (case-insensitively)
358 Whitespace in filenames is tolerated on most systems, but not all,
359 and even on systems where it might be tolerated, some utilities
360 might become confused by such whitespace.
362 Many systems (DOS, VMS ODS-2) cannot have more than one C<.> in their
365 Don't assume C<< > >> won't be the first character of a filename.
366 Always use the three-arg version of
367 L<C<open>|perlfunc/open FILEHANDLE,EXPR>:
369 open my $fh, '<', $existing_file) or die $!;
371 Two-arg L<C<open>|perlfunc/open FILEHANDLE,EXPR> is magic and can
372 translate characters like C<< > >>, C<< < >>, and C<|> in filenames,
373 which is usually the wrong thing to do.
374 L<C<sysopen>|perlfunc/sysopen FILEHANDLE,FILENAME,MODE> and three-arg
375 L<C<open>|perlfunc/open FILEHANDLE,EXPR> don't have this problem.
377 Don't use C<:> as a part of a filename since many systems use that for
378 their own semantics (Mac OS Classic for separating pathname components,
379 many networking schemes and utilities for separating the nodename and
380 the pathname, and so on). For the same reasons, avoid C<@>, C<;> and
383 Don't assume that in pathnames you can collapse two leading slashes
384 C<//> into one: some networking and clustering filesystems have special
385 semantics for that. Let the operating system sort it out.
387 The I<portable filename characters> as defined by ANSI C are
389 a b c d e f g h i j k l m n o p q r s t u v w x y z
390 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
394 and C<-> shouldn't be the first character. If you want to be
395 hypercorrect, stay case-insensitive and within the 8.3 naming
396 convention (all the files and directories have to be unique within one
397 directory if their names are lowercased and truncated to eight
398 characters before the C<.>, if any, and to three characters after the
399 C<.>, if any). (And do not use C<.>s in directory names.)
401 =head2 System Interaction
403 Not all platforms provide a command line. These are usually platforms
404 that rely primarily on a Graphical User Interface (GUI) for user
405 interaction. A program requiring a command line interface might
406 not work everywhere. This is probably for the user of the program
407 to deal with, so don't stay up late worrying about it.
409 Some platforms can't delete or rename files held open by the system,
410 this limitation may also apply to changing filesystem metainformation
411 like file permissions or owners. Remember to
412 L<C<close>|perlfunc/close FILEHANDLE> files when you are done with them.
413 Don't L<C<unlink>|perlfunc/unlink LIST> or
414 L<C<rename>|perlfunc/rename OLDNAME,NEWNAME> an open file. Don't
415 L<C<tie>|perlfunc/tie VARIABLE,CLASSNAME,LIST> or
416 L<C<open>|perlfunc/open FILEHANDLE,EXPR> a file already tied or opened;
417 L<C<untie>|perlfunc/untie VARIABLE> or
418 L<C<close>|perlfunc/close FILEHANDLE> it first.
420 Don't open the same file more than once at a time for writing, as some
421 operating systems put mandatory locks on such files.
423 Don't assume that write/modify permission on a directory gives the
424 right to add or delete files/directories in that directory. That is
425 filesystem specific: in some filesystems you need write/modify
426 permission also (or even just) in the file/directory itself. In some
427 filesystems (AFS, DFS) the permission to add/delete directory entries
428 is a completely separate permission.
430 Don't assume that a single L<C<unlink>|perlfunc/unlink LIST> completely
431 gets rid of the file: some filesystems (most notably the ones in VMS) have
432 versioned filesystems, and L<C<unlink>|perlfunc/unlink LIST> removes only
433 the most recent one (it doesn't remove all the versions because by default
434 the native tools on those platforms remove just the most recent version,
435 too). The portable idiom to remove all the versions of a file is
437 1 while unlink "file";
439 This will terminate if the file is undeleteable for some reason
440 (protected, not there, and so on).
442 Don't count on a specific environment variable existing in
443 L<C<%ENV>|perlvar/%ENV>. Don't count on L<C<%ENV>|perlvar/%ENV> entries
444 being case-sensitive, or even case-preserving. Don't try to clear
445 L<C<%ENV>|perlvar/%ENV> by saying C<%ENV = ();>, or, if you really have
446 to, make it conditional on C<$^O ne 'VMS'> since in VMS the
447 L<C<%ENV>|perlvar/%ENV> table is much more than a per-process key-value
450 On VMS, some entries in the L<C<%ENV>|perlvar/%ENV> hash are dynamically
451 created when their key is used on a read if they did not previously
452 exist. The values for C<$ENV{HOME}>, C<$ENV{TERM}>, C<$ENV{PATH}>, and
453 C<$ENV{USER}>, are known to be dynamically generated. The specific names
454 that are dynamically generated may vary with the version of the C library
455 on VMS, and more may exist than are documented.
457 On VMS by default, changes to the L<C<%ENV>|perlvar/%ENV> hash persist
458 after perl exits. Subsequent invocations of perl in the same process can
459 inadvertently inherit environment settings that were meant to be
462 Don't count on signals or L<C<%SIG>|perlvar/%SIG> for anything.
464 Don't count on filename globbing. Use
465 L<C<opendir>|perlfunc/opendir DIRHANDLE,EXPR>,
466 L<C<readdir>|perlfunc/readdir DIRHANDLE>, and
467 L<C<closedir>|perlfunc/closedir DIRHANDLE> instead.
469 Don't count on per-program environment variables, or per-program current
472 Don't count on specific values of L<C<$!>|perlvar/$!>, neither numeric nor
473 especially the string values. Users may switch their locales causing
474 error messages to be translated into their languages. If you can
475 trust a POSIXish environment, you can portably use the symbols defined
476 by the L<C<Errno>|Errno> module, like C<ENOENT>. And don't trust on the
477 values of L<C<$!>|perlvar/$!> at all except immediately after a failed
480 =head2 Command names versus file pathnames
482 Don't assume that the name used to invoke a command or program with
483 L<C<system>|perlfunc/system LIST> or L<C<exec>|perlfunc/exec LIST> can
484 also be used to test for the existence of the file that holds the
485 executable code for that command or program.
486 First, many systems have "internal" commands that are built-in to the
487 shell or OS and while these commands can be invoked, there is no
488 corresponding file. Second, some operating systems (e.g., Cygwin,
489 DJGPP, OS/2, and VOS) have required suffixes for executable files;
490 these suffixes are generally permitted on the command name but are not
491 required. Thus, a command like C<perl> might exist in a file named
492 F<perl>, F<perl.exe>, or F<perl.pm>, depending on the operating system.
493 The variable L<C<$Config{_exe}>|Config/C<_exe>> in the
494 L<C<Config>|Config> module holds the executable suffix, if any. Third,
495 the VMS port carefully sets up L<C<$^X>|perlvar/$^X> and
496 L<C<$Config{perlpath}>|Config/C<perlpath>> so that no further processing
497 is required. This is just as well, because the matching regular
498 expression used below would then have to deal with a possible trailing
499 version number in the VMS file name.
501 To convert L<C<$^X>|perlvar/$^X> to a file pathname, taking account of
502 the requirements of the various operating system possibilities, say:
507 $thisperl .= $Config{_exe}
508 unless $thisperl =~ m/\Q$Config{_exe}\E$/i;
511 To convert L<C<$Config{perlpath}>|Config/C<perlpath>> to a file pathname, say:
514 my $thisperl = $Config{perlpath};
516 $thisperl .= $Config{_exe}
517 unless $thisperl =~ m/\Q$Config{_exe}\E$/i;
522 Don't assume that you can reach the public Internet.
524 Don't assume that there is only one way to get through firewalls
525 to the public Internet.
527 Don't assume that you can reach outside world through any other port
528 than 80, or some web proxy. ftp is blocked by many firewalls.
530 Don't assume that you can send email by connecting to the local SMTP port.
532 Don't assume that you can reach yourself or any node by the name
533 'localhost'. The same goes for '127.0.0.1'. You will have to try both.
535 Don't assume that the host has only one network card, or that it
536 can't bind to many virtual IP addresses.
538 Don't assume a particular network device name.
540 Don't assume a particular set of
541 L<C<ioctl>|perlfunc/ioctl FILEHANDLE,FUNCTION,SCALAR>s will work.
543 Don't assume that you can ping hosts and get replies.
545 Don't assume that any particular port (service) will respond.
547 Don't assume that L<C<Sys::Hostname>|Sys::Hostname> (or any other API or
548 command) returns either a fully qualified hostname or a non-qualified
549 hostname: it all depends on how the system had been configured. Also
550 remember that for things such as DHCP and NAT, the hostname you get back
551 might not be very useful.
553 All the above I<don't>s may look daunting, and they are, but the key
554 is to degrade gracefully if one cannot reach the particular network
555 service one wants. Croaking or hanging do not look very professional.
557 =head2 Interprocess Communication (IPC)
559 In general, don't directly access the system in code meant to be
560 portable. That means, no L<C<system>|perlfunc/system LIST>,
561 L<C<exec>|perlfunc/exec LIST>, L<C<fork>|perlfunc/fork>,
562 L<C<pipe>|perlfunc/pipe READHANDLE,WRITEHANDLE>,
563 L<C<``> or C<qxE<sol>E<sol>>|perlop/C<qxE<sol>I<STRING>E<sol>>>,
564 L<C<open>|perlfunc/open FILEHANDLE,EXPR> with a C<|>, nor any of the other
565 things that makes being a Perl hacker worth being.
567 Commands that launch external processes are generally supported on
568 most platforms (though many of them do not support any type of
569 forking). The problem with using them arises from what you invoke
570 them on. External tools are often named differently on different
571 platforms, may not be available in the same location, might accept
572 different arguments, can behave differently, and often present their
573 results in a platform-dependent way. Thus, you should seldom depend
574 on them to produce consistent results. (Then again, if you're calling
575 C<netstat -a>, you probably don't expect it to run on both Unix and CP/M.)
577 One especially common bit of Perl code is opening a pipe to B<sendmail>:
579 open(my $mail, '|-', '/usr/lib/sendmail -t')
580 or die "cannot fork sendmail: $!";
582 This is fine for systems programming when sendmail is known to be
583 available. But it is not fine for many non-Unix systems, and even
584 some Unix systems that may not have sendmail installed. If a portable
585 solution is needed, see the various distributions on CPAN that deal
586 with it. L<C<Mail::Mailer>|Mail::Mailer> and L<C<Mail::Send>|Mail::Send>
587 in the C<MailTools> distribution are commonly used, and provide several
588 mailing methods, including C<mail>, C<sendmail>, and direct SMTP (via
589 L<C<Net::SMTP>|Net::SMTP>) if a mail transfer agent is not available.
590 L<C<Mail::Sendmail>|Mail::Sendmail> is a standalone module that provides
591 simple, platform-independent mailing.
593 The Unix System V IPC (C<msg*(), sem*(), shm*()>) is not available
594 even on all Unix platforms.
596 Do not use either the bare result of C<pack("N", 10, 20, 30, 40)> or
597 bare v-strings (such as C<v10.20.30.40>) to represent IPv4 addresses:
598 both forms just pack the four bytes into network order. That this
599 would be equal to the C language C<in_addr> struct (which is what the
600 socket code internally uses) is not guaranteed. To be portable use
601 the routines of the L<C<Socket>|Socket> module, such as
602 L<C<inet_aton>|Socket/$ip_address = inet_aton $string>,
603 L<C<inet_ntoa>|Socket/$string = inet_ntoa $ip_address>, and
604 L<C<sockaddr_in>|Socket/$sockaddr = sockaddr_in $port, $ip_address>.
606 The rule of thumb for portable code is: Do it all in portable Perl, or
607 use a module (that may internally implement it with platform-specific
608 code, but exposes a common interface).
610 =head2 External Subroutines (XS)
612 XS code can usually be made to work with any platform, but dependent
613 libraries, header files, etc., might not be readily available or
614 portable, or the XS code itself might be platform-specific, just as Perl
615 code might be. If the libraries and headers are portable, then it is
616 normally reasonable to make sure the XS code is portable, too.
618 A different type of portability issue arises when writing XS code:
619 availability of a C compiler on the end-user's system. C brings
620 with it its own portability issues, and writing XS code will expose
621 you to some of those. Writing purely in Perl is an easier way to
624 =head2 Standard Modules
626 In general, the standard modules work across platforms. Notable
627 exceptions are the L<C<CPAN>|CPAN> module (which currently makes
628 connections to external programs that may not be available),
629 platform-specific modules (like L<C<ExtUtils::MM_VMS>|ExtUtils::MM_VMS>),
632 There is no one DBM module available on all platforms.
633 L<C<SDBM_File>|SDBM_File> and the others are generally available on all
634 Unix and DOSish ports, but not in MacPerl, where only
635 L<C<NDBM_File>|NDBM_File> and L<C<DB_File>|DB_File> are available.
637 The good news is that at least some DBM module should be available, and
638 L<C<AnyDBM_File>|AnyDBM_File> will use whichever module it can find. Of
639 course, then the code needs to be fairly strict, dropping to the greatest
640 common factor (e.g., not exceeding 1K for each record), so that it will
641 work with any DBM module. See L<AnyDBM_File> for more details.
645 The system's notion of time of day and calendar date is controlled in
646 widely different ways. Don't assume the timezone is stored in C<$ENV{TZ}>,
647 and even if it is, don't assume that you can control the timezone through
648 that variable. Don't assume anything about the three-letter timezone
649 abbreviations (for example that MST would be the Mountain Standard Time,
650 it's been known to stand for Moscow Standard Time). If you need to
651 use timezones, express them in some unambiguous format like the
652 exact number of minutes offset from UTC, or the POSIX timezone
655 Don't assume that the epoch starts at 00:00:00, January 1, 1970,
656 because that is OS- and implementation-specific. It is better to
657 store a date in an unambiguous representation. The ISO 8601 standard
658 defines YYYY-MM-DD as the date format, or YYYY-MM-DDTHH:MM:SS
659 (that's a literal "T" separating the date from the time).
660 Please do use the ISO 8601 instead of making us guess what
661 date 02/03/04 might be. ISO 8601 even sorts nicely as-is.
662 A text representation (like "1987-12-18") can be easily converted
663 into an OS-specific value using a module like
664 L<C<Time::Piece>|Time::Piece> (see L<Time::Piece/Date Parsing>) or
665 L<C<Date::Parse>|Date::Parse>. An array of values, such as those
666 returned by L<C<localtime>|perlfunc/localtime EXPR>, can be converted to an OS-specific
667 representation using L<C<Time::Local>|Time::Local>.
669 When calculating specific times, such as for tests in time or date modules,
670 it may be appropriate to calculate an offset for the epoch.
672 use Time::Local qw(timegm);
673 my $offset = timegm(0, 0, 0, 1, 0, 70);
675 The value for C<$offset> in Unix will be C<0>, but in Mac OS Classic
676 will be some large number. C<$offset> can then be added to a Unix time
677 value to get what should be the proper value on any system.
679 =head2 Character sets and character encoding
681 Assume very little about character sets.
683 Assume nothing about numerical values (L<C<ord>|perlfunc/ord EXPR>,
684 L<C<chr>|perlfunc/chr NUMBER>) of characters.
685 Do not use explicit code point ranges (like C<\xHH-\xHH)>. However,
686 starting in Perl v5.22, regular expression pattern bracketed character
687 class ranges specified like C<qr/[\N{U+HH}-\N{U+HH}]/> are portable,
688 and starting in Perl v5.24, the same ranges are portable in
689 L<C<trE<sol>E<sol>E<sol>>|perlop/C<trE<sol>I<SEARCHLIST>E<sol>I<REPLACEMENTLIST>E<sol>cdsr>>.
690 You can portably use symbolic character classes like C<[:print:]>.
692 Do not assume that the alphabetic characters are encoded contiguously
693 (in the numeric sense). There may be gaps. Special coding in Perl,
694 however, guarantees that all subsets of C<qr/[A-Z]/>, C<qr/[a-z]/>, and
695 C<qr/[0-9]/> behave as expected.
696 L<C<trE<sol>E<sol>E<sol>>|perlop/C<trE<sol>I<SEARCHLIST>E<sol>I<REPLACEMENTLIST>E<sol>cdsr>>
697 behaves the same for these ranges. In patterns, any ranges specified with
698 end points using the C<\N{...}> notations ensures character set
699 portability, but it is a bug in Perl v5.22 that this isn't true of
700 L<C<trE<sol>E<sol>E<sol>>|perlop/C<trE<sol>I<SEARCHLIST>E<sol>I<REPLACEMENTLIST>E<sol>cdsr>>,
703 Do not assume anything about the ordering of the characters.
704 The lowercase letters may come before or after the uppercase letters;
705 the lowercase and uppercase may be interlaced so that both "a" and "A"
706 come before "b"; the accented and other international characters may
707 be interlaced so that E<auml> comes before "b".
708 L<Unicode::Collate> can be used to sort this all out.
710 =head2 Internationalisation
712 If you may assume POSIX (a rather large assumption), you may read
713 more about the POSIX locale system from L<perllocale>. The locale
714 system at least attempts to make things a little bit more portable,
715 or at least more convenient and native-friendly for non-English
716 users. The system affects character sets and encoding, and date
717 and time formatting--amongst other things.
719 If you really want to be international, you should consider Unicode.
720 See L<perluniintro> and L<perlunicode> for more information.
722 By default Perl assumes your source code is written in an 8-bit ASCII
723 superset. To embed Unicode characters in your strings and regexes, you can
724 use the L<C<\x{HH}> or (more portably) C<\N{U+HH}>
725 notations|perlop/Quote and Quote-like Operators>. You can also use the
726 L<C<utf8>|utf8> pragma and write your code in UTF-8, which lets you use
727 Unicode characters directly (not just in quoted constructs but also in
730 =head2 System Resources
732 If your code is destined for systems with severely constrained (or
733 missing!) virtual memory systems then you want to be I<especially> mindful
734 of avoiding wasteful constructs such as:
736 my @lines = <$very_large_file>; # bad
738 while (<$fh>) {$file .= $_} # sometimes bad
739 my $file = join('', <$fh>); # better
741 The last two constructs may appear unintuitive to most people. The
742 first repeatedly grows a string, whereas the second allocates a
743 large chunk of memory in one go. On some systems, the second is
744 more efficient than the first.
748 Most multi-user platforms provide basic levels of security, usually
749 implemented at the filesystem level. Some, however, unfortunately do
750 not. Thus the notion of user id, or "home" directory,
751 or even the state of being logged-in, may be unrecognizable on many
752 platforms. If you write programs that are security-conscious, it
753 is usually best to know what type of system you will be running
754 under so that you can write code explicitly for that platform (or
757 Don't assume the Unix filesystem access semantics: the operating
758 system or the filesystem may be using some ACL systems, which are
759 richer languages than the usual C<rwx>. Even if the C<rwx> exist,
760 their semantics might be different.
762 (From the security viewpoint, testing for permissions before attempting to
763 do something is silly anyway: if one tries this, there is potential
764 for race conditions. Someone or something might change the
765 permissions between the permissions check and the actual operation.
766 Just try the operation.)
768 Don't assume the Unix user and group semantics: especially, don't
769 expect L<C<< $< >>|perlvar/$E<lt>> and L<C<< $> >>|perlvar/$E<gt>> (or
770 L<C<$(>|perlvar/$(> and L<C<$)>|perlvar/$)>) to work for switching
771 identities (or memberships).
773 Don't assume set-uid and set-gid semantics. (And even if you do,
774 think twice: set-uid and set-gid are a known can of security worms.)
778 For those times when it is necessary to have platform-specific code,
779 consider keeping the platform-specific code in one place, making porting
780 to other platforms easier. Use the L<C<Config>|Config> module and the
781 special variable L<C<$^O>|perlvar/$^O> to differentiate platforms, as
782 described in L</"PLATFORMS">.
784 Beware of the "else syndrome":
786 if ($^O eq 'MSWin32') {
787 # code that assumes Windows
789 # code that assumes Linux
792 The C<else> branch should be used for the really ultimate fallback,
793 not for code specific to some platform.
795 Be careful in the tests you supply with your module or programs.
796 Module code may be fully portable, but its tests might not be. This
797 often happens when tests spawn off other processes or call external
798 programs to aid in the testing, or when (as noted above) the tests
799 assume certain things about the filesystem and paths. Be careful not
800 to depend on a specific output style for errors, such as when checking
801 L<C<$!>|perlvar/$!> after a failed system call. Using
802 L<C<$!>|perlvar/$!> for anything else than displaying it as output is
803 doubtful (though see the L<C<Errno>|Errno> module for testing reasonably
804 portably for error value). Some platforms expect a certain output format,
805 and Perl on those platforms may have been adjusted accordingly. Most
806 specifically, don't anchor a regex when testing an error value.
810 Modules uploaded to CPAN are tested by a variety of volunteers on
811 different platforms. These CPAN testers are notified by mail of each
812 new upload, and reply to the list with PASS, FAIL, NA (not applicable to
813 this platform), or UNKNOWN (unknown), along with any relevant notations.
815 The purpose of the testing is twofold: one, to help developers fix any
816 problems in their code that crop up because of lack of testing on other
817 platforms; two, to provide users with information about whether
818 a given module works on a given platform.
826 Mailing list: cpan-testers-discuss@perl.org
830 Testing results: L<http://www.cpantesters.org/>
836 Perl is built with a L<C<$^O>|perlvar/$^O> variable that indicates the
837 operating system it was built on. This was implemented
838 to help speed up code that would otherwise have to C<use Config>
839 and use the value of L<C<$Config{osname}>|Config/C<osname>>. Of course,
840 to get more detailed information about the system, looking into
841 L<C<%Config>|Config/DESCRIPTION> is certainly recommended.
843 L<C<%Config>|Config/DESCRIPTION> cannot always be trusted, however,
844 because it was built at compile time. If perl was built in one place,
845 then transferred elsewhere, some values may be wrong. The values may
846 even have been edited after the fact.
850 Perl works on a bewildering variety of Unix and Unix-like platforms (see
851 e.g. most of the files in the F<hints/> directory in the source code kit).
852 On most of these systems, the value of L<C<$^O>|perlvar/$^O> (hence
853 L<C<$Config{osname}>|Config/C<osname>>, too) is determined either by
854 lowercasing and stripping punctuation from the first field of the string
855 returned by typing C<uname -a> (or a similar command) at the shell prompt
856 or by testing the file system for the presence of uniquely named files
857 such as a kernel or header file. Here, for example, are a few of the
858 more popular Unix flavors:
860 uname $^O $Config{archname}
861 --------------------------------------------
863 BSD/OS bsdos i386-bsdos
865 DYNIX/ptx dynixptx i386-dynixptx
866 FreeBSD freebsd freebsd-i386
867 Haiku haiku BePC-haiku
868 Linux linux arm-linux
869 Linux linux armv5tel-linux
870 Linux linux i386-linux
871 Linux linux i586-linux
872 Linux linux ppc-linux
873 HP-UX hpux PA-RISC1.1
875 Mac OS X darwin darwin
877 NeXT 4 next OPENSTEP-Mach
878 openbsd openbsd i386-openbsd
879 OSF1 dec_osf alpha-dec_osf
880 reliantunix-n svr4 RM400-svr4
881 SCO_SV sco_sv i386-sco_sv
882 SINIX-N svr4 RM400-svr4
883 sn4609 unicos CRAY_C90-unicos
884 sn6521 unicosmk t3e-unicosmk
885 sn9617 unicos CRAY_J90-unicos
886 SunOS solaris sun4-solaris
887 SunOS solaris i86pc-solaris
888 SunOS4 sunos sun4-sunos
890 Because the value of L<C<$Config{archname}>|Config/C<archname>> may
891 depend on the hardware architecture, it can vary more than the value of
892 L<C<$^O>|perlvar/$^O>.
894 =head2 DOS and Derivatives
896 Perl has long been ported to Intel-style microcomputers running under
897 systems like PC-DOS, MS-DOS, OS/2, and most Windows platforms you can
898 bring yourself to mention (except for Windows CE, if you count that).
899 Users familiar with I<COMMAND.COM> or I<CMD.EXE> style shells should
900 be aware that each of these file specifications may have subtle
903 my $filespec0 = "c:/foo/bar/file.txt";
904 my $filespec1 = "c:\\foo\\bar\\file.txt";
905 my $filespec2 = 'c:\foo\bar\file.txt';
906 my $filespec3 = 'c:\\foo\\bar\\file.txt';
908 System calls accept either C</> or C<\> as the path separator.
909 However, many command-line utilities of DOS vintage treat C</> as
910 the option prefix, so may get confused by filenames containing C</>.
911 Aside from calling any external programs, C</> will work just fine,
912 and probably better, as it is more consistent with popular usage,
913 and avoids the problem of remembering what to backwhack and what
916 The DOS FAT filesystem can accommodate only "8.3" style filenames. Under
917 the "case-insensitive, but case-preserving" HPFS (OS/2) and NTFS (NT)
918 filesystems you may have to be careful about case returned with functions
919 like L<C<readdir>|perlfunc/readdir DIRHANDLE> or used with functions like
920 L<C<open>|perlfunc/open FILEHANDLE,EXPR> or
921 L<C<opendir>|perlfunc/opendir DIRHANDLE,EXPR>.
923 DOS also treats several filenames as special, such as F<AUX>, F<PRN>,
924 F<NUL>, F<CON>, F<COM1>, F<LPT1>, F<LPT2>, etc. Unfortunately, sometimes
925 these filenames won't even work if you include an explicit directory
926 prefix. It is best to avoid such filenames, if you want your code to be
927 portable to DOS and its derivatives. It's hard to know what these all
930 Users of these operating systems may also wish to make use of
931 scripts such as F<pl2bat.bat> to put wrappers around your scripts.
933 Newline (C<\n>) is translated as C<\015\012> by the I/O system when
934 reading from and writing to files (see L</"Newlines">).
935 C<binmode($filehandle)> will keep C<\n> translated as C<\012> for that
937 L<C<binmode>|perlfunc/binmode FILEHANDLE> should always be used for code
938 that deals with binary data. That's assuming you realize in advance that
939 your data is in binary. General-purpose programs should often assume
940 nothing about their data.
942 The L<C<$^O>|perlvar/$^O> variable and the
943 L<C<$Config{archname}>|Config/C<archname>> values for various DOSish
944 perls are as follows:
946 OS $^O $Config{archname} ID Version
947 ---------------------------------------------------------
951 Windows 3.1 ? ? 0 3 01
952 Windows 95 MSWin32 MSWin32-x86 1 4 00
953 Windows 98 MSWin32 MSWin32-x86 1 4 10
954 Windows ME MSWin32 MSWin32-x86 1 ?
955 Windows NT MSWin32 MSWin32-x86 2 4 xx
956 Windows NT MSWin32 MSWin32-ALPHA 2 4 xx
957 Windows NT MSWin32 MSWin32-ppc 2 4 xx
958 Windows 2000 MSWin32 MSWin32-x86 2 5 00
959 Windows XP MSWin32 MSWin32-x86 2 5 01
960 Windows 2003 MSWin32 MSWin32-x86 2 5 02
961 Windows Vista MSWin32 MSWin32-x86 2 6 00
962 Windows 7 MSWin32 MSWin32-x86 2 6 01
963 Windows 7 MSWin32 MSWin32-x64 2 6 01
964 Windows 2008 MSWin32 MSWin32-x86 2 6 01
965 Windows 2008 MSWin32 MSWin32-x64 2 6 01
966 Windows CE MSWin32 ? 3
969 The various MSWin32 Perl's can distinguish the OS they are running on
970 via the value of the fifth element of the list returned from
971 L<C<Win32::GetOSVersion()>|Win32/Win32::GetOSVersion()>. For example:
973 if ($^O eq 'MSWin32') {
974 my @os_version_info = Win32::GetOSVersion();
975 print +('3.1','95','NT')[$os_version_info[4]],"\n";
978 There are also C<Win32::IsWinNT()|Win32/Win32::IsWinNT()>,
979 C<Win32::IsWin95()|Win32/Win32::IsWin95()>, and
980 L<C<Win32::GetOSName()>|Win32/Win32::GetOSName()>; try
981 L<C<perldoc Win32>|Win32>.
982 The very portable L<C<POSIX::uname()>|POSIX/C<uname>> will work too:
984 c:\> perl -MPOSIX -we "print join '|', uname"
985 Windows NT|moonru|5.0|Build 2195 (Service Pack 2)|x86
993 The djgpp environment for DOS, L<http://www.delorie.com/djgpp/>
998 The EMX environment for DOS, OS/2, etc. emx@iaehv.nl,
999 L<ftp://hobbes.nmsu.edu/pub/os2/dev/emx/> Also L<perlos2>.
1003 Build instructions for Win32 in L<perlwin32>, or under the Cygnus environment
1008 The C<Win32::*> modules in L<Win32>.
1012 The ActiveState Pages, L<http://www.activestate.com/>
1016 The Cygwin environment for Win32; F<README.cygwin> (installed
1017 as L<perlcygwin>), L<http://www.cygwin.com/>
1021 The U/WIN environment for Win32,
1022 L<http://www.research.att.com/sw/tools/uwin/>
1026 Build instructions for OS/2, L<perlos2>
1032 Perl on VMS is discussed in L<perlvms> in the Perl distribution.
1034 The official name of VMS as of this writing is OpenVMS.
1036 Interacting with Perl from the Digital Command Language (DCL) shell
1037 often requires a different set of quotation marks than Unix shells do.
1040 $ perl -e "print ""Hello, world.\n"""
1043 There are several ways to wrap your Perl scripts in DCL F<.COM> files, if
1044 you are so inclined. For example:
1046 $ write sys$output "Hello from DCL!"
1048 $ then perl -x 'f$environment("PROCEDURE")
1049 $ else perl -x - 'p1 'p2 'p3 'p4 'p5 'p6 'p7 'p8
1050 $ deck/dollars="__END__"
1053 print "Hello from Perl!\n";
1058 Do take care with C<$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT> if your
1059 Perl-in-DCL script expects to do things like C<< $read = <STDIN>; >>.
1061 The VMS operating system has two filesystems, designated by their
1062 on-disk structure (ODS) level: ODS-2 and its successor ODS-5. The
1063 initial port of Perl to VMS pre-dates ODS-5, but all current testing and
1064 development assumes ODS-5 and its capabilities, including case
1065 preservation, extended characters in filespecs, and names up to 8192
1068 Perl on VMS can accept either VMS- or Unix-style file
1069 specifications as in either of the following:
1071 $ perl -ne "print if /perl_setup/i" SYS$LOGIN:LOGIN.COM
1072 $ perl -ne "print if /perl_setup/i" /sys$login/login.com
1074 but not a mixture of both as in:
1076 $ perl -ne "print if /perl_setup/i" sys$login:/login.com
1077 Can't open sys$login:/login.com: file specification syntax error
1079 In general, the easiest path to portability is always to specify
1080 filenames in Unix format unless they will need to be processed by native
1081 commands or utilities. Because of this latter consideration, the
1082 L<File::Spec> module by default returns native format specifications
1083 regardless of input format. This default may be reversed so that
1084 filenames are always reported in Unix format by specifying the
1085 C<DECC$FILENAME_UNIX_REPORT> feature logical in the environment.
1087 The file type, or extension, is always present in a VMS-format file
1088 specification even if it's zero-length. This means that, by default,
1089 L<C<readdir>|perlfunc/readdir DIRHANDLE> will return a trailing dot on a
1090 file with no extension, so where you would see C<"a"> on Unix you'll see
1091 C<"a."> on VMS. However, the trailing dot may be suppressed by enabling
1092 the C<DECC$READDIR_DROPDOTNOTYPE> feature in the environment (see the CRTL
1093 documentation on feature logical names).
1095 What C<\n> represents depends on the type of file opened. It usually
1096 represents C<\012> but it could also be C<\015>, C<\012>, C<\015\012>,
1097 C<\000>, C<\040>, or nothing depending on the file organization and
1098 record format. The L<C<VMS::Stdio>|VMS::Stdio> module provides access to
1099 the special C<fopen()> requirements of files with unusual attributes on
1102 The value of L<C<$^O>|perlvar/$^O> on OpenVMS is "VMS". To determine the
1103 architecture that you are running on refer to
1104 L<C<$Config{archname}>|Config/C<archname>>.
1106 On VMS, perl determines the UTC offset from the C<SYS$TIMEZONE_DIFFERENTIAL>
1107 logical name. Although the VMS epoch began at 17-NOV-1858 00:00:00.00,
1108 calls to L<C<localtime>|perlfunc/localtime EXPR> are adjusted to count
1109 offsets from 01-JAN-1970 00:00:00.00, just like Unix.
1117 F<README.vms> (installed as F<README_vms>), L<perlvms>
1121 vmsperl list, vmsperl-subscribe@perl.org
1125 vmsperl on the web, L<http://www.sidhe.org/vmsperl/index.html>
1129 VMS Software Inc. web site, L<http://www.vmssoftware.com>
1135 Perl on VOS (also known as OpenVOS) is discussed in F<README.vos>
1136 in the Perl distribution (installed as L<perlvos>). Perl on VOS
1137 can accept either VOS- or Unix-style file specifications as in
1138 either of the following:
1140 $ perl -ne "print if /perl_setup/i" >system>notices
1141 $ perl -ne "print if /perl_setup/i" /system/notices
1143 or even a mixture of both as in:
1145 $ perl -ne "print if /perl_setup/i" >system/notices
1147 Even though VOS allows the slash character to appear in object
1148 names, because the VOS port of Perl interprets it as a pathname
1149 delimiting character, VOS files, directories, or links whose
1150 names contain a slash character cannot be processed. Such files
1151 must be renamed before they can be processed by Perl.
1153 Older releases of VOS (prior to OpenVOS Release 17.0) limit file
1154 names to 32 or fewer characters, prohibit file names from
1155 starting with a C<-> character, and prohibit file names from
1156 containing C< > (space) or any character from the set C<< !#%&'()*;<=>? >>.
1158 Newer releases of VOS (OpenVOS Release 17.0 or later) support a
1159 feature known as extended names. On these releases, file names
1160 can contain up to 255 characters, are prohibited from starting
1161 with a C<-> character, and the set of prohibited characters is
1162 reduced to C<< #%*<>? >>. There are
1163 restrictions involving spaces and apostrophes: these characters
1164 must not begin or end a name, nor can they immediately precede or
1165 follow a period. Additionally, a space must not immediately
1166 precede another space or hyphen. Specifically, the following
1167 character combinations are prohibited: space-space,
1168 space-hyphen, period-space, space-period, period-apostrophe,
1169 apostrophe-period, leading or trailing space, and leading or
1170 trailing apostrophe. Although an extended file name is limited
1171 to 255 characters, a path name is still limited to 256
1174 The value of L<C<$^O>|perlvar/$^O> on VOS is "vos". To determine the
1175 architecture that you are running on refer to
1176 L<C<$Config{archname}>|Config/C<archname>>.
1184 F<README.vos> (installed as L<perlvos>)
1188 The VOS mailing list.
1190 There is no specific mailing list for Perl on VOS. You can contact
1191 the Stratus Technologies Customer Assistance Center (CAC) for your
1192 region, or you can use the contact information located in the
1193 distribution files on the Stratus Anonymous FTP site.
1197 Stratus Technologies on the web at L<http://www.stratus.com>
1201 VOS Open-Source Software on the web at L<http://ftp.stratus.com/pub/vos/vos.html>
1205 =head2 EBCDIC Platforms
1207 v5.22 core Perl runs on z/OS (formerly OS/390). Theoretically it could
1208 run on the successors of OS/400 on AS/400 minicomputers as well as
1209 VM/ESA, and BS2000 for S/390 Mainframes. Such computers use EBCDIC
1210 character sets internally (usually Character Code Set ID 0037 for OS/400
1211 and either 1047 or POSIX-BC for S/390 systems).
1213 The rest of this section may need updating, but we don't know what it
1214 should say. Please email comments to
1215 L<perlbug@perl.org|mailto:perlbug@perl.org>.
1217 On the mainframe Perl currently works under the "Unix system
1218 services for OS/390" (formerly known as OpenEdition), VM/ESA OpenEdition, or
1219 the BS200 POSIX-BC system (BS2000 is supported in Perl 5.6 and greater).
1220 See L<perlos390> for details. Note that for OS/400 there is also a port of
1221 Perl 5.8.1/5.10.0 or later to the PASE which is ASCII-based (as opposed to
1222 ILE which is EBCDIC-based), see L<perlos400>.
1224 As of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix
1225 sub-systems do not support the C<#!> shebang trick for script invocation.
1226 Hence, on OS/390 and VM/ESA Perl scripts can be executed with a header
1227 similar to the following simple script:
1230 eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}'
1232 #!/usr/local/bin/perl # just a comment really
1234 print "Hello from perl!\n";
1236 OS/390 will support the C<#!> shebang trick in release 2.8 and beyond.
1237 Calls to L<C<system>|perlfunc/system LIST> and backticks can use POSIX
1238 shell syntax on all S/390 systems.
1240 On the AS/400, if PERL5 is in your library list, you may need
1241 to wrap your Perl scripts in a CL procedure to invoke them like so:
1244 CALL PGM(PERL5/PERL) PARM('/QOpenSys/hello.pl')
1247 This will invoke the Perl script F<hello.pl> in the root of the
1248 QOpenSys file system. On the AS/400 calls to
1249 L<C<system>|perlfunc/system LIST> or backticks must use CL syntax.
1251 On these platforms, bear in mind that the EBCDIC character set may have
1252 an effect on what happens with some Perl functions (such as
1253 L<C<chr>|perlfunc/chr NUMBER>, L<C<pack>|perlfunc/pack TEMPLATE,LIST>,
1254 L<C<print>|perlfunc/print FILEHANDLE LIST>,
1255 L<C<printf>|perlfunc/printf FILEHANDLE FORMAT, LIST>,
1256 L<C<ord>|perlfunc/ord EXPR>, L<C<sort>|perlfunc/sort SUBNAME LIST>,
1257 L<C<sprintf>|perlfunc/sprintf FORMAT, LIST>,
1258 L<C<unpack>|perlfunc/unpack TEMPLATE,EXPR>), as
1259 well as bit-fiddling with ASCII constants using operators like
1260 L<C<^>, C<&> and C<|>|perlop/Bitwise String Operators>, not to mention
1261 dealing with socket interfaces to ASCII computers (see L</"Newlines">).
1263 Fortunately, most web servers for the mainframe will correctly
1264 translate the C<\n> in the following statement to its ASCII equivalent
1265 (C<\r> is the same under both Unix and z/OS):
1267 print "Content-type: text/html\r\n\r\n";
1269 The values of L<C<$^O>|perlvar/$^O> on some of these platforms include:
1271 uname $^O $Config{archname}
1272 --------------------------------------------
1275 POSIX-BC posix-bc BS2000-posix-bc
1277 Some simple tricks for determining if you are running on an EBCDIC
1278 platform could include any of the following (perhaps all):
1280 if ("\t" eq "\005") { print "EBCDIC may be spoken here!\n"; }
1282 if (ord('A') == 193) { print "EBCDIC may be spoken here!\n"; }
1284 if (chr(169) eq 'z') { print "EBCDIC may be spoken here!\n"; }
1286 One thing you may not want to rely on is the EBCDIC encoding
1287 of punctuation characters since these may differ from code page to code
1288 page (and once your module or script is rumoured to work with EBCDIC,
1289 folks will want it to work with all EBCDIC character sets).
1297 L<perlos390>, L<perlos400>, L<perlbs2000>, L<perlebcdic>.
1301 The perl-mvs@perl.org list is for discussion of porting issues as well as
1302 general usage issues for all EBCDIC Perls. Send a message body of
1303 "subscribe perl-mvs" to majordomo@perl.org.
1307 AS/400 Perl information at
1308 L<http://as400.rochester.ibm.com/>
1309 as well as on CPAN in the F<ports/> directory.
1313 =head2 Acorn RISC OS
1315 Because Acorns use ASCII with newlines (C<\n>) in text files as C<\012> like
1316 Unix, and because Unix filename emulation is turned on by default,
1317 most simple scripts will probably work "out of the box". The native
1318 filesystem is modular, and individual filesystems are free to be
1319 case-sensitive or insensitive, and are usually case-preserving. Some
1320 native filesystems have name length limits, which file and directory
1321 names are silently truncated to fit. Scripts should be aware that the
1322 standard filesystem currently has a name length limit of B<10>
1323 characters, with up to 77 items in a directory, but other filesystems
1324 may not impose such limitations.
1326 Native filenames are of the form
1328 Filesystem#Special_Field::DiskName.$.Directory.Directory.File
1332 Special_Field is not usually present, but may contain . and $ .
1333 Filesystem =~ m|[A-Za-z0-9_]|
1334 DsicName =~ m|[A-Za-z0-9_/]|
1335 $ represents the root directory
1336 . is the path separator
1337 @ is the current directory (per filesystem but machine global)
1338 ^ is the parent directory
1339 Directory and File =~ m|[^\0- "\.\$\%\&:\@\\^\|\177]+|
1341 The default filename translation is roughly C<tr|/.|./|>, swapping dots
1344 Note that C<"ADFS::HardDisk.$.File" ne 'ADFS::HardDisk.$.File'> and that
1345 the second stage of C<$> interpolation in regular expressions will fall
1346 foul of the L<C<$.>|perlvar/$.> variable if scripts are not careful.
1348 Logical paths specified by system variables containing comma-separated
1349 search lists are also allowed; hence C<System:Modules> is a valid
1350 filename, and the filesystem will prefix C<Modules> with each section of
1351 C<System$Path> until a name is made that points to an object on disk.
1352 Writing to a new file C<System:Modules> would be allowed only if
1353 C<System$Path> contains a single item list. The filesystem will also
1354 expand system variables in filenames if enclosed in angle brackets, so
1355 C<< <System$Dir>.Modules >> would look for the file
1356 S<C<$ENV{'System$Dir'} . 'Modules'>>. The obvious implication of this is
1357 that B<fully qualified filenames can start with C<< <> >>> and the
1358 three-argument form of L<C<open>|perlfunc/open FILEHANDLE,EXPR> should
1361 Because C<.> was in use as a directory separator and filenames could not
1362 be assumed to be unique after 10 characters, Acorn implemented the C
1363 compiler to strip the trailing C<.c> C<.h> C<.s> and C<.o> suffix from
1364 filenames specified in source code and store the respective files in
1365 subdirectories named after the suffix. Hence files are translated:
1368 C:foo.h C:h.foo (logical path variable)
1369 sys/os.h sys.h.os (C compiler groks Unix-speak)
1370 10charname.c c.10charname
1371 10charname.o o.10charname
1372 11charname_.c c.11charname (assuming filesystem truncates at 10)
1374 The Unix emulation library's translation of filenames to native assumes
1375 that this sort of translation is required, and it allows a user-defined list
1376 of known suffixes that it will transpose in this fashion. This may
1377 seem transparent, but consider that with these rules F<foo/bar/baz.h>
1378 and F<foo/bar/h/baz> both map to F<foo.bar.h.baz>, and that
1379 L<C<readdir>|perlfunc/readdir DIRHANDLE> and L<C<glob>|perlfunc/glob EXPR>
1380 cannot and do not attempt to emulate the reverse mapping. Other
1381 C<.>'s in filenames are translated to C</>.
1383 As implied above, the environment accessed through
1384 L<C<%ENV>|perlvar/%ENV> is global, and the convention is that program
1385 specific environment variables are of the form C<Program$Name>.
1386 Each filesystem maintains a current directory,
1387 and the current filesystem's current directory is the B<global> current
1388 directory. Consequently, sociable programs don't change the current
1389 directory but rely on full pathnames, and programs (and Makefiles) cannot
1390 assume that they can spawn a child process which can change the current
1391 directory without affecting its parent (and everyone else for that
1394 Because native operating system filehandles are global and are currently
1395 allocated down from 255, with 0 being a reserved value, the Unix emulation
1396 library emulates Unix filehandles. Consequently, you can't rely on
1397 passing C<STDIN>, C<STDOUT>, or C<STDERR> to your children.
1399 The desire of users to express filenames of the form
1400 C<< <Foo$Dir>.Bar >> on the command line unquoted causes problems,
1401 too: L<C<``>|perlop/C<qxE<sol>I<STRING>E<sol>>> command output capture has
1402 to perform a guessing game. It assumes that a string C<< <[^<>]+\$[^<>]> >>
1403 is a reference to an environment variable, whereas anything else involving
1404 C<< < >> or C<< > >> is redirection, and generally manages to be 99%
1405 right. Of course, the problem remains that scripts cannot rely on any
1406 Unix tools being available, or that any tools found have Unix-like command
1409 Extensions and XS are, in theory, buildable by anyone using free
1410 tools. In practice, many don't, as users of the Acorn platform are
1411 used to binary distributions. MakeMaker does run, but no available
1412 make currently copes with MakeMaker's makefiles; even if and when
1413 this should be fixed, the lack of a Unix-like shell will cause
1414 problems with makefile rules, especially lines of the form
1415 C<cd sdbm && make all>, and anything using quoting.
1417 S<"RISC OS"> is the proper name for the operating system, but the value
1418 in L<C<$^O>|perlvar/$^O> is "riscos" (because we don't like shouting).
1422 Perl has been ported to many platforms that do not fit into any of
1423 the categories listed above. Some, such as AmigaOS,
1424 QNX, Plan 9, and VOS, have been well-integrated into the standard
1425 Perl source code kit. You may need to see the F<ports/> directory
1426 on CPAN for information, and possibly binaries, for the likes of:
1427 aos, Atari ST, lynxos, riscos, Novell Netware, Tandem Guardian,
1428 I<etc.> (Yes, we know that some of these OSes may fall under the
1429 Unix category, but we are not a standards body.)
1431 Some approximate operating system names and their L<C<$^O>|perlvar/$^O>
1432 values in the "OTHER" category include:
1434 OS $^O $Config{archname}
1435 ------------------------------------------
1436 Amiga DOS amigaos m68k-amigos
1444 Amiga, F<README.amiga> (installed as L<perlamiga>).
1448 A free perl5-based PERL.NLM for Novell Netware is available in
1449 precompiled binary and source code form from L<http://www.novell.com/>
1450 as well as from CPAN.
1454 S<Plan 9>, F<README.plan9>
1458 =head1 FUNCTION IMPLEMENTATIONS
1460 Listed below are functions that are either completely unimplemented
1461 or else have been implemented differently on various platforms.
1462 Preceding each description will be, in parentheses, a list of
1463 platforms that the description applies to.
1465 The list may well be incomplete, or even wrong in some places. When
1466 in doubt, consult the platform-specific README files in the Perl
1467 source distribution, and any other documentation resources accompanying
1470 Be aware, moreover, that even among Unix-ish systems there are variations.
1472 For many functions, you can also query L<C<%Config>|Config/DESCRIPTION>,
1473 exported by default from the L<C<Config>|Config> module. For example, to
1474 check whether the platform has the L<C<lstat>|perlfunc/lstat FILEHANDLE>
1475 call, check L<C<$Config{d_lstat}>|Config/C<d_lstat>>. See L<Config> for a
1476 full description of available variables.
1478 =head2 Alphabetical Listing of Perl Functions
1485 C<-w> only inspects the read-only file attribute (FILE_ATTRIBUTE_READONLY),
1486 which determines whether the directory can be deleted, not whether it can
1487 be written to. Directories always have read and write access unless denied
1488 by discretionary access control lists (DACLs).
1491 C<-r>, C<-w>, C<-x>, and C<-o> tell whether the file is accessible,
1492 which may not reflect UIC-based file protections.
1495 C<-s> by name on an open file will return the space reserved on disk,
1496 rather than the current extent. C<-s> on an open filehandle returns the
1499 (Win32, VMS, S<RISC OS>)
1500 C<-R>, C<-W>, C<-X>, C<-O> are indistinguishable from C<-r>, C<-w>,
1503 (Win32, VMS, S<RISC OS>)
1504 C<-g>, C<-k>, C<-l>, C<-u>, C<-A> are not particularly meaningful.
1507 C<-p> is not particularly meaningful.
1510 C<-d> is true if passed a device spec without an explicit directory.
1513 C<-x> (or C<-X>) determine if a file ends in one of the executable
1514 suffixes. C<-S> is meaningless.
1517 C<-x> (or C<-X>) determine if a file has an executable file type.
1522 Emulated using timers that must be explicitly polled whenever Perl
1523 wants to dispatch "safe signals" and therefore cannot interrupt
1524 blocking system calls.
1528 (Tru64, HP-UX 10.20)
1529 Due to issues with various CPUs, math libraries, compilers, and standards,
1530 results for C<atan2> may vary depending on any combination of the above.
1531 Perl attempts to conform to the Open Group/IEEE standards for the results
1532 returned from C<atan2>, but cannot force the issue if the system Perl is
1533 run on does not allow it.
1535 The current version of the standards for C<atan2> is available at
1536 L<http://www.opengroup.org/onlinepubs/009695399/functions/atan2.html>.
1544 Reopens file and restores pointer; if function fails, underlying
1545 filehandle may be closed, or pointer may be in a different position.
1548 The value returned by L<C<tell>|perlfunc/tell FILEHANDLE> may be affected
1549 after the call, and the filehandle may be flushed.
1554 Only good for changing "owner" read-write access; "group" and "other"
1555 bits are meaningless.
1558 Only good for changing "owner" and "other" read-write access.
1561 Access permissions are mapped onto VOS access-control list changes.
1564 The actual permissions set depend on the value of the C<CYGWIN> variable
1565 in the SYSTEM environment settings.
1568 Setting the exec bit on some locations (generally F</sdcard>) will return true
1569 but not actually set the bit.
1573 (S<Plan 9>, S<RISC OS>)
1577 Does nothing, but won't fail.
1580 A little funky, because VOS's notion of ownership is a little funky.
1584 (Win32, VMS, S<Plan 9>, S<RISC OS>, VOS)
1590 May not be available if library or source was not provided when building
1598 (VMS, S<Plan 9>, VOS)
1603 (VMS, S<Plan 9>, VOS)
1615 Invokes VMS debugger.
1620 C<exec LIST> without the use of indirect object syntax (C<exec PROGRAM LIST>)
1621 may fall back to trying the shell if the first C<spawn()> fails.
1623 (SunOS, Solaris, HP-UX)
1624 Does not automatically flush output handles on some platforms.
1632 Emulates Unix C<exit> (which considers C<exit 1> to indicate an error) by
1633 mapping the C<1> to C<SS$_ABORT> (C<44>). This behavior may be overridden
1634 with the pragma L<C<use vmsish 'exit'>|vmsish/C<vmsish exit>>. As with
1635 the CRTL's C<exit()> function, C<exit 0> is also mapped to an exit status
1636 of C<SS$_NORMAL> (C<1>); this mapping cannot be overridden. Any other
1638 is used directly as Perl's exit status. On VMS, unless the future
1639 POSIX_EXIT mode is enabled, the exit code should always be a valid
1640 VMS exit code and not a generic number. When the POSIX_EXIT mode is
1641 enabled, a generic number will be encoded in a method compatible with
1642 the C library _POSIX_EXIT macro so that it can be decoded by other
1643 programs, particularly ones written in C, like the GNV package.
1646 C<exit> resets file pointers, which is a problem when called
1647 from a child process (created by L<C<fork>|perlfunc/fork>) in
1648 L<C<BEGIN>|perlmod/BEGIN, UNITCHECK, CHECK, INIT and END>.
1649 A workaround is to use L<C<POSIX::_exit>|POSIX/C<_exit>>.
1651 exit unless $Config{archname} =~ /\bsolaris\b/;
1661 Some functions available based on the version of VMS.
1665 (VMS, S<RISC OS>, VOS)
1670 (AmigaOS, S<RISC OS>, VMS)
1674 Emulated using multiple interpreters. See L<perlfork>.
1676 (SunOS, Solaris, HP-UX)
1677 Does not automatically flush output handles on some platforms.
1686 (Win32, VMS, S<RISC OS>)
1696 (Win32, VMS, S<RISC OS>, VOS)
1709 (Win32, VMS, S<RISC OS>)
1714 (Android, Win32, S<Plan 9>)
1727 (Win32, VMS, S<RISC OS>)
1732 (Android, Win32, S<Plan 9>)
1735 =item getprotobynumber
1747 (Android, Win32, VMS)
1753 C<gethostbyname('localhost')> does not work everywhere: you may have
1754 to use C<gethostbyname('127.0.0.1')>.
1763 (Android, Win32, S<Plan 9>)
1768 (Android, Win32, S<Plan 9>)
1783 (Android, Win32, S<Plan 9>, S<RISC OS>)
1788 (Win32, S<Plan 9>, S<RISC OS>)
1793 (Android, Win32, S<Plan 9>, S<RISC OS>)
1798 (S<Plan 9>, Win32, S<RISC OS>)
1807 Either not implemented or a no-op.
1811 (Android, S<RISC OS>, VMS, Win32)
1821 (Android, Win32, S<Plan 9>)
1826 (Android, Win32, S<Plan 9>)
1841 This operator is implemented via the L<C<File::Glob>|File::Glob> extension
1842 on most platforms. See L<File::Glob> for portability information.
1846 In theory, C<gmtime> is reliable from -2**63 to 2**63-1. However,
1847 because work-arounds in the implementation use floating point numbers,
1848 it will become inaccurate as the time gets larger. This is a bug and
1849 will be fixed in the future.
1852 Time values are 32-bit quantities.
1860 Available only for socket handles, and it does what the C<ioctlsocket()> call
1861 in the Winsock API does.
1864 Available only for socket handles.
1869 Not implemented, hence not useful for taint checking.
1872 C<kill> doesn't send a signal to the identified process like it does on
1873 Unix platforms. Instead C<kill($sig, $pid)> terminates the process
1874 identified by C<$pid>, and makes it exit immediately with exit status
1875 C<$sig>. As in Unix, if C<$sig> is 0 and the specified process exists, it
1876 returns true without actually terminating it.
1879 C<kill(-9, $pid)> will terminate the process specified by C<$pid> and
1880 recursively all child processes owned by it. This is different from
1881 the Unix semantics, where the signal will be delivered to all
1882 processes in the same process group as the process specified by
1886 A pid of -1 indicating all processes on the system is not currently
1895 Link count not updated because hard links are not quite that hard
1896 (They are sort of half-way between hard and soft links).
1899 Hard links are implemented on Win32 under NTFS only. They are
1900 natively supported on Windows 2000 and later. On Windows NT they
1901 are implemented using the Windows POSIX subsystem support and the
1902 Perl process will need Administrator or Backup Operator privileges
1903 to create hard links.
1906 Available on 64 bit OpenVMS 8.2 and later.
1910 C<localtime> has the same range as L</gmtime>, but because time zone
1911 rules change, its accuracy for historical and future times may degrade
1912 but usually by no more than an hour.
1920 Return values (especially for device and inode) may be bogus.
1930 (Android, Win32, VMS, S<Plan 9>, S<RISC OS>, VOS)
1936 Open modes C<|-> and C<-|> are unsupported.
1938 (SunOS, Solaris, HP-UX)
1939 Opening a process does not automatically flush output handles on some
1944 (Win32, VMS, S<RISC OS>)
1950 Can't move directories between directories on different logical volumes.
1955 Will not cause L<C<readdir>|perlfunc/readdir DIRHANDLE> to re-read the
1956 directory stream. The entries already read before the C<rewinddir> call
1957 will just be returned again from a cache buffer.
1962 Only implemented on sockets.
1965 Only reliable on sockets.
1967 Note that the L<C<select FILEHANDLE>|perlfunc/select FILEHANDLE> form is
1976 (Android, Win32, VMS, S<RISC OS>)
1981 (Android, VMS, Win32, S<RISC OS>)
1986 (Win32, VMS, S<RISC OS>, VOS)
1991 (Win32, VMS, S<RISC OS>, VOS)
1996 (Android, Win32, S<RISC OS>)
2012 (Android, Win32, VMS, S<RISC OS>)
2018 Emulated using synchronization functions such that it can be
2019 interrupted by L<C<alarm>|perlfunc/alarm SECONDS>, and limited to a
2020 maximum of 4294967 seconds, approximately 49 days.
2028 Available on 64 bit OpenVMS 8.2 and later.
2032 Platforms that do not have C<rdev>, C<blksize>, or C<blocks> will return
2033 these as C<''>, so numeric comparison or manipulation of these fields may
2034 cause 'not numeric' warnings.
2037 C<ctime> not supported on UFS.
2040 C<ctime> is creation time instead of inode change time.
2043 C<dev> and C<ino> are not meaningful.
2046 C<dev> and C<ino> are not necessarily reliable.
2049 C<mtime>, C<atime> and C<ctime> all return the last modification time.
2050 C<dev> and C<ino> are not necessarily reliable.
2053 C<dev>, C<rdev>, C<blksize>, and C<blocks> are not available. C<ino> is not
2054 meaningful and will differ between stat calls on the same file.
2057 Some versions of cygwin when doing a C<stat("foo")> and not finding it
2058 may then attempt to C<stat("foo.exe")>.
2061 C<stat> needs to open the file to determine the link count
2062 and update attributes that may have been changed through hard links.
2063 Setting L<C<${^WIN32_SLOPPY_STAT}>|perlvar/${^WIN32_SLOPPY_STAT}> to a
2064 true value speeds up C<stat> by not performing this operation.
2072 Implemented on 64 bit VMS 8.3. VMS requires the symbolic link to be in Unix
2073 syntax if it is intended to resolve to a valid path.
2077 (Win32, VMS, S<RISC OS>, VOS)
2083 The traditional C<0>, C<1>, and C<2> MODEs are implemented with different
2084 numeric values on some systems. The flags exported by L<C<Fcntl>|Fcntl>
2085 (C<O_RDONLY>, C<O_WRONLY>, C<O_RDWR>) should work everywhere though.
2090 As an optimization, may not call the command shell specified in
2091 C<$ENV{PERL5SHELL}>. C<system(1, @args)> spawns an external
2092 process and immediately returns its process designator, without
2093 waiting for it to terminate. Return value may be used subsequently
2094 in L<C<wait>|perlfunc/wait> or L<C<waitpid>|perlfunc/waitpid PID,FLAGS>.
2095 Failure to C<spawn()> a subprocess is indicated by setting
2096 L<C<$?>|perlvar/$?> to C<<< 255 << 8 >>>. L<C<$?>|perlvar/$?> is set in a
2097 way compatible with Unix (i.e. the exit status of the subprocess is
2098 obtained by C<<< $? >> 8 >>>, as described in the documentation).
2101 There is no shell to process metacharacters, and the native standard is
2102 to pass a command line terminated by "\n" "\r" or "\0" to the spawned
2103 program. Redirection such as C<< > foo >> is performed (if at all) by
2104 the run time library of the spawned program. C<system LIST> will call
2105 the Unix emulation library's L<C<exec>|perlfunc/exec LIST> emulation,
2106 which attempts to provide emulation of the stdin, stdout, stderr in force
2107 in the parent, provided the child program uses a compatible version of the
2108 emulation library. C<system SCALAR> will call the native command line
2109 directly and no such emulation of a child Unix program will occur.
2110 Mileage B<will> vary.
2113 C<system LIST> without the use of indirect object syntax (C<system PROGRAM LIST>)
2114 may fall back to trying the shell if the first C<spawn()> fails.
2116 (SunOS, Solaris, HP-UX)
2117 Does not automatically flush output handles on some platforms.
2120 The return value is POSIX-like (shifted up by 8 bits), which only allows
2121 room for a made-up value derived from the severity bits of the native
2122 32-bit condition code (unless overridden by
2123 L<C<use vmsish 'status'>|vmsish/C<vmsish status>>). If the native
2124 condition code is one that has a POSIX value encoded, the POSIX value will
2125 be decoded to extract the expected exit value. For more details see
2136 "Cumulative" times will be bogus. On anything other than Windows NT
2137 or Windows 2000, "system" time will be bogus, and "user" time is
2138 actually the time returned by the L<C<clock()>|clock(3)> function in the C
2146 (Older versions of VMS)
2150 Truncation to same-or-shorter lengths only.
2153 If a FILEHANDLE is supplied, it must be writable and opened in append
2154 mode (i.e., use C<<< open(my $fh, '>>', 'filename') >>>
2155 or C<sysopen(my $fh, ..., O_APPEND|O_RDWR)>. If a filename is supplied, it
2156 should not be held open elsewhere.
2160 Returns C<undef> where unavailable.
2163 C<umask> works but the correct permissions are set only when the file
2169 Only the modification time is updated.
2172 May not behave as expected. Behavior depends on the C runtime
2173 library's implementation of L<C<utime()>|utime(2)>, and the filesystem
2174 being used. The FAT filesystem typically does not support an "access
2175 time" field, and it may limit timestamps to a granularity of two seconds.
2182 Can only be applied to process handles returned for processes spawned
2183 using C<system(1, ...)> or pseudo processes created with
2184 L<C<fork>|perlfunc/fork>.
2192 =head1 Supported Platforms
2194 The following platforms are known to build Perl 5.12 (as of April 2010,
2195 its release date) from the standard source code distribution available
2196 at L<http://www.cpan.org/src>
2200 =item Linux (x86, ARM, IA64)
2214 =item Windows Server 2003
2218 =item Windows Server 2008
2226 Some tests are known to fail:
2232 F<ext/XS-APItest/t/call_checker.t> - see
2233 L<https://rt.perl.org/Ticket/Display.html?id=78502>
2237 F<dist/I18N-Collate/t/I18N-Collate.t>
2241 F<ext/Win32CORE/t/win32core.t> - may fail on recent cygwin installs.
2245 =item Solaris (x86, SPARC)
2251 =item Alpha (7.2 and later)
2253 =item I64 (8.2 and later)
2263 =item Debian GNU/kFreeBSD
2267 =item Irix (6.5. What else?)
2275 =item QNX Neutrino RTOS (6.5.0)
2279 =item Stratus OpenVOS (17.0 or later)
2285 =item time_t issues that may or may not be fixed
2289 =item Symbian (Series 60 v3, 3.2 and 5 - what else?)
2291 =item Stratus VOS / OpenVOS
2299 Perl now builds with FreeMiNT/Atari. It fails a few tests, that needs
2302 The FreeMiNT port uses GNU dld for loadable module capabilities. So
2303 ensure you have that library installed when building perl.
2307 =head1 EOL Platforms
2311 The following platforms were supported by a previous version of
2312 Perl but have been officially removed from Perl's source code
2323 The following platforms were supported up to 5.10. They may still
2324 have worked in 5.12, but supporting code has been removed for 5.14:
2340 The following platforms were supported by a previous version of
2341 Perl but have been officially removed from Perl's source code
2348 =item Apollo Domain/OS
2350 =item Apple Mac OS 8/9
2357 =head1 Supported Platforms (Perl 5.8)
2359 As of July 2002 (the Perl release 5.8.0), the following platforms were
2360 able to build Perl from the standard source code distribution
2361 available at L<http://www.cpan.org/src/>
2372 HI-UXMPP (Hitachi) (5.8.0 worked but we didn't know it)
2382 ReliantUNIX (formerly SINIX)
2384 OpenVMS (formerly VMS)
2385 Open UNIX (Unixware) (since Perl 5.8.1/5.9.0)
2387 OS/400 (using the PASE) (since Perl 5.8.1/5.9.0)
2389 POSIX-BC (formerly BS2000)
2394 Tru64 UNIX (formerly DEC OSF/1, Digital UNIX)
2399 Win95/98/ME/2K/XP 2)
2401 z/OS (formerly OS/390)
2404 1) in DOS mode either the DOS or OS/2 ports can be used
2405 2) compilers: Borland, MinGW (GCC), VC6
2407 The following platforms worked with the previous releases (5.6 and
2408 5.7), but we did not manage either to fix or to test these in time
2409 for the 5.8.0 release. There is a very good chance that many of these
2410 will work fine with the 5.8.0.
2423 Known to be broken for 5.8.0 (but 5.6.1 and 5.7.2 can be used):
2427 The following platforms have been known to build Perl from source in
2428 the past (5.005_03 and earlier), but we haven't been able to verify
2429 their status for the current release, either because the
2430 hardware/software platforms are rare or because we don't have an
2431 active champion on these platforms--or both. They used to work,
2432 though, so go ahead and try compiling them, and let perlbug@perl.org
2465 The following platforms have their own source code distributions and
2466 binaries available via L<http://www.cpan.org/ports/>
2470 OS/400 (ILE) 5.005_02
2471 Tandem Guardian 5.004
2473 The following platforms have only binaries available via
2474 L<http://www.cpan.org/ports/index.html> :
2478 Acorn RISCOS 5.005_02
2482 Although we do suggest that you always build your own Perl from
2483 the source code, both for maximal configurability and for security,
2484 in case you are in a hurry you can check
2485 L<http://www.cpan.org/ports/index.html> for binary distributions.
2489 L<perlaix>, L<perlamiga>, L<perlbs2000>,
2490 L<perlce>, L<perlcygwin>, L<perldos>,
2491 L<perlebcdic>, L<perlfreebsd>, L<perlhurd>, L<perlhpux>, L<perlirix>,
2492 L<perlmacos>, L<perlmacosx>,
2493 L<perlnetware>, L<perlos2>, L<perlos390>, L<perlos400>,
2494 L<perlplan9>, L<perlqnx>, L<perlsolaris>, L<perltru64>,
2495 L<perlunicode>, L<perlvms>, L<perlvos>, L<perlwin32>, and L<Win32>.
2497 =head1 AUTHORS / CONTRIBUTORS
2499 Abigail <abigail@foad.org>,
2500 Charles Bailey <bailey@newman.upenn.edu>,
2501 Graham Barr <gbarr@pobox.com>,
2502 Tom Christiansen <tchrist@perl.com>,
2503 Nicholas Clark <nick@ccl4.org>,
2504 Thomas Dorner <Thomas.Dorner@start.de>,
2505 Andy Dougherty <doughera@lafayette.edu>,
2506 Dominic Dunlop <domo@computer.org>,
2507 Neale Ferguson <neale@vma.tabnsw.com.au>,
2508 David J. Fiander <davidf@mks.com>,
2509 Paul Green <Paul.Green@stratus.com>,
2510 M.J.T. Guy <mjtg@cam.ac.uk>,
2511 Jarkko Hietaniemi <jhi@iki.fi>,
2512 Luther Huffman <lutherh@stratcom.com>,
2513 Nick Ing-Simmons <nick@ing-simmons.net>,
2514 Andreas J. KE<ouml>nig <a.koenig@mind.de>,
2515 Markus Laker <mlaker@contax.co.uk>,
2516 Andrew M. Langmead <aml@world.std.com>,
2517 Lukas Mai <l.mai@web.de>,
2518 Larry Moore <ljmoore@freespace.net>,
2519 Paul Moore <Paul.Moore@uk.origin-it.com>,
2520 Chris Nandor <pudge@pobox.com>,
2521 Matthias Neeracher <neeracher@mac.com>,
2522 Philip Newton <pne@cpan.org>,
2523 Gary Ng <71564.1743@CompuServe.COM>,
2524 Tom Phoenix <rootbeer@teleport.com>,
2525 AndrE<eacute> Pirard <A.Pirard@ulg.ac.be>,
2526 Peter Prymmer <pvhp@forte.com>,
2527 Hugo van der Sanden <hv@crypt0.demon.co.uk>,
2528 Gurusamy Sarathy <gsar@activestate.com>,
2529 Paul J. Schinder <schinder@pobox.com>,
2530 Michael G Schwern <schwern@pobox.com>,
2531 Dan Sugalski <dan@sidhe.org>,
2532 Nathan Torkington <gnat@frii.com>,
2533 John Malmberg <wb8tyw@qsl.net>