1PERLPORT(1) Perl Programmers Reference Guide PERLPORT(1)
2
6 perlport - Writing portable Perl
7
9 Perl runs on numerous operating systems. While most of them share much
10 in common, they also have their own unique features.
11 This document is meant to help you to find out what constitutes
13 portable Perl code. That way once you make a decision to write
14 portably, you know where the lines are drawn, and you can stay within
15 them.
16 There is a tradeoff between taking full advantage of one particular
18 type of computer and taking advantage of a full range of them.
19 Naturally, as you broaden your range and become more diverse, the
20 common factors drop, and you are left with an increasingly smaller area
21 of common ground in which you can operate to accomplish a particular
22 task. Thus, when you begin attacking a problem, it is important to
23 consider under which part of the tradeoff curve you want to operate.
24 Specifically, you must decide whether it is important that the task
25 that you are coding has the full generality of being portable, or
26 whether to just get the job done right now. This is the hardest choice
27 to be made. The rest is easy, because Perl provides many choices,
28 whichever way you want to approach your problem.
29 Looking at it another way, writing portable code is usually about
31 willfully limiting your available choices. Naturally, it takes
32 discipline and sacrifice to do that. The product of portability and
33 convenience may be a constant. You have been warned.
34 Be aware of two important points:
36 Not all Perl programs have to be portable
38 There is no reason you should not use Perl as a language to glue
39 Unix tools together, or to prototype a Macintosh application, or to
40 manage the Windows registry. If it makes no sense to aim for
41 portability for one reason or another in a given program, then
42 don't bother.
43 Nearly all of Perl already is portable
45 Don't be fooled into thinking that it is hard to create portable
46 Perl code. It isn't. Perl tries its level-best to bridge the gaps
47 between what's available on different platforms, and all the means
48 available to use those features. Thus almost all Perl code runs on
49 any machine without modification. But there are some significant
50 issues in writing portable code, and this document is entirely
51 about those issues.
52 Here's the general rule: When you approach a task commonly done using a
54 whole range of platforms, think about writing portable code. That way,
55 you don't sacrifice much by way of the implementation choices you can
56 avail yourself of, and at the same time you can give your users lots of
57 platform choices. On the other hand, when you have to take advantage
58 of some unique feature of a particular platform, as is often the case
59 with systems programming (whether for Unix, Windows, VMS, etc.),
60 consider writing platform-specific code.
61 When the code will run on only two or three operating systems, you may
63 need to consider only the differences of those particular systems. The
64 important thing is to decide where the code will run and to be
65 deliberate in your decision.
66 The material below is separated into three main sections: main issues
68 of portability ("ISSUES"), platform-specific issues ("PLATFORMS"), and
69 built-in Perl functions that behave differently on various ports
70 ("FUNCTION IMPLEMENTATIONS").
71 This information should not be considered complete; it includes
73 possibly transient information about idiosyncrasies of some of the
74 ports, almost all of which are in a state of constant evolution. Thus,
75 this material should be considered a perpetual work in progress ("<IMG
76 SRC="yellow_sign.gif" ALT="Under Construction">").
77
79 Newlines
80 In most operating systems, lines in files are terminated by newlines.
81 Just what is used as a newline may vary from OS to OS. Unix
82 traditionally uses "\012", one type of DOSish I/O uses "\015\012",
83 Mac OS uses "\015", and z/OS uses "\025".
84 Perl uses "\n" to represent the "logical" newline, where what is
86 logical may depend on the platform in use. In MacPerl, "\n" always
87 means "\015". On EBCDIC platforms, "\n" could be "\025" or "\045". In
88 DOSish perls, "\n" usually means "\012", but when accessing a file in
89 "text" mode, perl uses the ":crlf" layer that translates it to (or
90 from) "\015\012", depending on whether you're reading or writing. Unix
91 does the same thing on ttys in canonical mode. "\015\012" is commonly
92 referred to as CRLF.
93 To trim trailing newlines from text lines use "chomp". With default
95 settings that function looks for a trailing "\n" character and thus
96 trims in a portable way.
97 When dealing with binary files (or text files in binary mode) be sure
99 to explicitly set $/ to the appropriate value for your file format
100 before using "chomp".
101 Because of the "text" mode translation, DOSish perls have limitations
103 in using "seek" and "tell" on a file accessed in "text" mode. Stick to
104 "seek"-ing to locations you got from "tell" (and no others), and you
105 are usually free to use "seek" and "tell" even in "text" mode. Using
106 "seek" or "tell" or other file operations may be non-portable. If you
107 use "binmode" on a file, however, you can usually "seek" and "tell"
108 with arbitrary values safely.
109 A common misconception in socket programming is that "\n eq \012"
111 everywhere. When using protocols such as common Internet protocols,
112 "\012" and "\015" are called for specifically, and the values of the
113 logical "\n" and "\r" (carriage return) are not reliable.
114 print $socket "Hi there, client!\r\n"; # WRONG
116 print $socket "Hi there, client!\015\012"; # RIGHT
117 However, using "\015\012" (or "\cM\cJ", or "\x0D\x0A") can be tedious
119 and unsightly, as well as confusing to those maintaining the code. As
120 such, the "Socket" module supplies the Right Thing for those who want
121 it.
122 use Socket qw(:DEFAULT :crlf);
124 print $socket "Hi there, client!$CRLF" # RIGHT
125 When reading from a socket, remember that the default input record
127 separator $/ is "\n", but robust socket code will recognize as either
128 "\012" or "\015\012" as end of line:
129 while (<$socket>) { # NOT ADVISABLE!
131 # ...
132 }
133 Because both CRLF and LF end in LF, the input record separator can be
135 set to LF and any CR stripped later. Better to write:
136 use Socket qw(:DEFAULT :crlf);
138 local($/) = LF; # not needed if $/ is already \012
139 while (<$socket>) {
141 s/$CR?$LF/\n/; # not sure if socket uses LF or CRLF, OK
142 # s/\015?\012/\n/; # same thing
143 }
144 This example is preferred over the previous one--even for Unix
146 platforms--because now any "\015"'s ("\cM"'s) are stripped out (and
147 there was much rejoicing).
148 Similarly, functions that return text data--such as a function that
150 fetches a web page--should sometimes translate newlines before
151 returning the data, if they've not yet been translated to the local
152 newline representation. A single line of code will often suffice:
153 $data =~ s/\015?\012/\n/g;
155 return $data;
156 Some of this may be confusing. Here's a handy reference to the ASCII
158 CR and LF characters. You can print it out and stick it in your
159 wallet.
160 LF eq \012 eq \x0A eq \cJ eq chr(10) eq ASCII 10
162 CR eq \015 eq \x0D eq \cM eq chr(13) eq ASCII 13
163 | Unix | DOS | Mac |
165 ---------------------------
166 \n | LF | LF | CR |
167 \r | CR | CR | LF |
168 \n * | LF | CRLF | CR |
169 \r * | CR | CR | LF |
170 ---------------------------
171 * text-mode STDIO
172 The Unix column assumes that you are not accessing a serial line (like
174 a tty) in canonical mode. If you are, then CR on input becomes "\n",
175 and "\n" on output becomes CRLF.
176 These are just the most common definitions of "\n" and "\r" in Perl.
178 There may well be others. For example, on an EBCDIC implementation
179 such as z/OS (OS/390) or OS/400 (using the ILE, the PASE is ASCII-
180 based) the above material is similar to "Unix" but the code numbers
181 change:
182 LF eq \025 eq \x15 eq \cU eq chr(21) eq CP-1047 21
184 LF eq \045 eq \x25 eq chr(37) eq CP-0037 37
185 CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-1047 13
186 CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-0037 13
187 | z/OS | OS/400 |
189 ----------------------
190 \n | LF | LF |
191 \r | CR | CR |
192 \n * | LF | LF |
193 \r * | CR | CR |
194 ----------------------
195 * text-mode STDIO
196 Numbers endianness and Width
198 Different CPUs store integers and floating point numbers in different
199 orders (called endianness) and widths (32-bit and 64-bit being the most
200 common today). This affects your programs when they attempt to
201 transfer numbers in binary format from one CPU architecture to another,
202 usually either "live" via network connection, or by storing the numbers
203 to secondary storage such as a disk file or tape.
204 Conflicting storage orders make an utter mess out of the numbers. If a
206 little-endian host (Intel, VAX) stores 0x12345678 (305419896 in
207 decimal), a big-endian host (Motorola, Sparc, PA) reads it as
208 0x78563412 (2018915346 in decimal). Alpha and MIPS can be either:
209 Digital/Compaq used/uses them in little-endian mode; SGI/Cray uses them
210 in big-endian mode. To avoid this problem in network (socket)
211 connections use the "pack" and "unpack" formats "n" and "N", the
212 "network" orders. These are guaranteed to be portable.
213 As of Perl 5.10.0, you can also use the ">" and "<" modifiers to force
215 big- or little-endian byte-order. This is useful if you want to store
216 signed integers or 64-bit integers, for example.
217 You can explore the endianness of your platform by unpacking a data
219 structure packed in native format such as:
220 print unpack("h*", pack("s2", 1, 2)), "\n";
222 # '10002000' on e.g. Intel x86 or Alpha 21064 in little-endian mode
223 # '00100020' on e.g. Motorola 68040
224 If you need to distinguish between endian architectures you could use
226 either of the variables set like so:
227 $is_big_endian = unpack("h*", pack("s", 1)) =~ /01/;
229 $is_little_endian = unpack("h*", pack("s", 1)) =~ /^1/;
230 Differing widths can cause truncation even between platforms of equal
232 endianness. The platform of shorter width loses the upper parts of the
233 number. There is no good solution for this problem except to avoid
234 transferring or storing raw binary numbers.
235 One can circumnavigate both these problems in two ways. Either
237 transfer and store numbers always in text format, instead of raw
238 binary, or else consider using modules like "Data::Dumper" and
239 "Storable" (included as of Perl 5.8). Keeping all data as text
240 significantly simplifies matters.
241 Files and Filesystems
243 Most platforms these days structure files in a hierarchical fashion.
244 So, it is reasonably safe to assume that all platforms support the
245 notion of a "path" to uniquely identify a file on the system. How that
246 path is really written, though, differs considerably.
247 Although similar, file path specifications differ between Unix,
249 Windows, Mac OS, OS/2, VMS, VOS, RISC OS, and probably others. Unix,
250 for example, is one of the few OSes that has the elegant idea of a
251 single root directory.
252 DOS, OS/2, VMS, VOS, and Windows can work similarly to Unix with "/" as
254 path separator, or in their own idiosyncratic ways (such as having
255 several root directories and various "unrooted" device files such NIL:
256 and LPT:).
257 Mac OS 9 and earlier used ":" as a path separator instead of "/".
259 The filesystem may support neither hard links ("link") nor symbolic
261 links ("symlink", "readlink", "lstat").
262 The filesystem may support neither access timestamp nor change
264 timestamp (meaning that about the only portable timestamp is the
265 modification timestamp), or one second granularity of any timestamps
266 (e.g. the FAT filesystem limits the time granularity to two seconds).
267 The "inode change timestamp" (the "-C" filetest) may really be the
269 "creation timestamp" (which it is not in Unix).
270 VOS perl can emulate Unix filenames with "/" as path separator. The
272 native pathname characters greater-than, less-than, number-sign, and
273 percent-sign are always accepted.
274 RISC OS perl can emulate Unix filenames with "/" as path separator, or
276 go native and use "." for path separator and ":" to signal filesystems
277 and disk names.
278 Don't assume Unix filesystem access semantics: that read, write, and
280 execute are all the permissions there are, and even if they exist, that
281 their semantics (for example what do "r", "w", and "x" mean on a
282 directory) are the Unix ones. The various Unix/POSIX compatibility
283 layers usually try to make interfaces like "chmod" work, but sometimes
284 there simply is no good mapping.
285 The "File::Spec" modules provide methods to manipulate path
287 specifications and return the results in native format for each
288 platform. This is often unnecessary as Unix-style paths are understood
289 by Perl on every supported platform, but if you need to produce native
290 paths for a native utility that does not understand Unix syntax, or if
291 you are operating on paths or path components in unknown (and thus
292 possibly native) syntax, "File::Spec" is your friend. Here are two
293 brief examples:
294 use File::Spec::Functions;
296 chdir(updir()); # go up one directory
297 # Concatenate a path from its components
299 my $file = catfile(updir(), 'temp', 'file.txt');
300 # on Unix: '../temp/file.txt'
301 # on Win32: '..\temp\file.txt'
302 # on VMS: '[-.temp]file.txt'
303 In general, production code should not have file paths hardcoded.
305 Making them user-supplied or read from a configuration file is better,
306 keeping in mind that file path syntax varies on different machines.
307 This is especially noticeable in scripts like Makefiles and test
309 suites, which often assume "/" as a path separator for subdirectories.
310 Also of use is "File::Basename" from the standard distribution, which
312 splits a pathname into pieces (base filename, full path to directory,
313 and file suffix).
314 Even when on a single platform (if you can call Unix a single
316 platform), remember not to count on the existence or the contents of
317 particular system-specific files or directories, like /etc/passwd,
318 /etc/sendmail.conf, /etc/resolv.conf, or even /tmp/. For example,
319 /etc/passwd may exist but not contain the encrypted passwords, because
320 the system is using some form of enhanced security. Or it may not
321 contain all the accounts, because the system is using NIS. If code
322 does need to rely on such a file, include a description of the file and
323 its format in the code's documentation, then make it easy for the user
324 to override the default location of the file.
325 Don't assume a text file will end with a newline. They should, but
327 people forget.
328 Do not have two files or directories of the same name with different
330 case, like test.pl and Test.pl, as many platforms have case-insensitive
331 (or at least case-forgiving) filenames. Also, try not to have non-word
332 characters (except for ".") in the names, and keep them to the 8.3
333 convention, for maximum portability, onerous a burden though this may
334 appear.
335 Likewise, when using the "AutoSplit" module, try to keep your functions
337 to 8.3 naming and case-insensitive conventions; or, at the least, make
338 it so the resulting files have a unique (case-insensitively) first 8
339 characters.
340 Whitespace in filenames is tolerated on most systems, but not all, and
342 even on systems where it might be tolerated, some utilities might
343 become confused by such whitespace.
344 Many systems (DOS, VMS ODS-2) cannot have more than one "." in their
346 filenames.
347 Don't assume ">" won't be the first character of a filename. Always
349 use the three-arg version of "open":
350 open my $fh, '<', $existing_file) or die $!;
352 Two-arg "open" is magic and can translate characters like ">", "<", and
354 "|" in filenames, which is usually the wrong thing to do. "sysopen"
355 and three-arg "open" don't have this problem.
356 Don't use ":" as a part of a filename since many systems use that for
358 their own semantics (Mac OS Classic for separating pathname components,
359 many networking schemes and utilities for separating the nodename and
360 the pathname, and so on). For the same reasons, avoid "@", ";" and
361 "|".
362 Don't assume that in pathnames you can collapse two leading slashes
364 "//" into one: some networking and clustering filesystems have special
365 semantics for that. Let the operating system sort it out.
366 The portable filename characters as defined by ANSI C are
368 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
370 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
371 0 1 2 3 4 5 6 7 8 9
372 . _ -
373 and "-" shouldn't be the first character. If you want to be
375 hypercorrect, stay case-insensitive and within the 8.3 naming
376 convention (all the files and directories have to be unique within one
377 directory if their names are lowercased and truncated to eight
378 characters before the ".", if any, and to three characters after the
379 ".", if any). (And do not use "."s in directory names.)
380 System Interaction
382 Not all platforms provide a command line. These are usually platforms
383 that rely primarily on a Graphical User Interface (GUI) for user
384 interaction. A program requiring a command line interface might not
385 work everywhere. This is probably for the user of the program to deal
386 with, so don't stay up late worrying about it.
387 Some platforms can't delete or rename files held open by the system,
389 this limitation may also apply to changing filesystem metainformation
390 like file permissions or owners. Remember to "close" files when you
391 are done with them. Don't "unlink" or "rename" an open file. Don't
392 "tie" or "open" a file already tied or opened; "untie" or "close" it
393 first.
394 Don't open the same file more than once at a time for writing, as some
396 operating systems put mandatory locks on such files.
397 Don't assume that write/modify permission on a directory gives the
399 right to add or delete files/directories in that directory. That is
400 filesystem specific: in some filesystems you need write/modify
401 permission also (or even just) in the file/directory itself. In some
402 filesystems (AFS, DFS) the permission to add/delete directory entries
403 is a completely separate permission.
404 Don't assume that a single "unlink" completely gets rid of the file:
406 some filesystems (most notably the ones in VMS) have versioned
407 filesystems, and "unlink" removes only the most recent one (it doesn't
408 remove all the versions because by default the native tools on those
409 platforms remove just the most recent version, too). The portable
410 idiom to remove all the versions of a file is
411 1 while unlink "file";
413 This will terminate if the file is undeletable for some reason
415 (protected, not there, and so on).
416 Don't count on a specific environment variable existing in %ENV. Don't
418 count on %ENV entries being case-sensitive, or even case-preserving.
419 Don't try to clear %ENV by saying "%ENV = ();", or, if you really have
420 to, make it conditional on "$^O ne 'VMS'" since in VMS the %ENV table
421 is much more than a per-process key-value string table.
422 On VMS, some entries in the %ENV hash are dynamically created when
424 their key is used on a read if they did not previously exist. The
425 values for $ENV{HOME}, $ENV{TERM}, $ENV{PATH}, and $ENV{USER}, are
426 known to be dynamically generated. The specific names that are
427 dynamically generated may vary with the version of the C library on
428 VMS, and more may exist than are documented.
429 On VMS by default, changes to the %ENV hash persist after perl exits.
431 Subsequent invocations of perl in the same process can inadvertently
432 inherit environment settings that were meant to be temporary.
433 Don't count on signals or %SIG for anything.
435 Don't count on filename globbing. Use "opendir", "readdir", and
437 "closedir" instead.
438 Don't count on per-program environment variables, or per-program
440 current directories.
441 Don't count on specific values of $!, neither numeric nor especially
443 the string values. Users may switch their locales causing error
444 messages to be translated into their languages. If you can trust a
445 POSIXish environment, you can portably use the symbols defined by the
446 "Errno" module, like "ENOENT". And don't trust on the values of $! at
447 all except immediately after a failed system call.
448 Command names versus file pathnames
450 Don't assume that the name used to invoke a command or program with
451 "system" or "exec" can also be used to test for the existence of the
452 file that holds the executable code for that command or program.
453 First, many systems have "internal" commands that are built-in to the
454 shell or OS and while these commands can be invoked, there is no
455 corresponding file. Second, some operating systems (e.g., Cygwin,
456 DJGPP, OS/2, and VOS) have required suffixes for executable files;
457 these suffixes are generally permitted on the command name but are not
458 required. Thus, a command like "perl" might exist in a file named
459 perl, perl.exe, or perl.pm, depending on the operating system. The
460 variable $Config{_exe} in the "Config" module holds the executable
461 suffix, if any. Third, the VMS port carefully sets up $^X and
462 $Config{perlpath} so that no further processing is required. This is
463 just as well, because the matching regular expression used below would
464 then have to deal with a possible trailing version number in the VMS
465 file name.
466 To convert $^X to a file pathname, taking account of the requirements
468 of the various operating system possibilities, say:
469 use Config;
471 my $thisperl = $^X;
472 if ($^O ne 'VMS') {
473 $thisperl .= $Config{_exe}
474 unless $thisperl =~ m/\Q$Config{_exe}\E$/i;
475 }
476 To convert $Config{perlpath} to a file pathname, say:
478 use Config;
480 my $thisperl = $Config{perlpath};
481 if ($^O ne 'VMS') {
482 $thisperl .= $Config{_exe}
483 unless $thisperl =~ m/\Q$Config{_exe}\E$/i;
484 }
485 Networking
487 Don't assume that you can reach the public Internet.
488 Don't assume that there is only one way to get through firewalls to the
490 public Internet.
491 Don't assume that you can reach outside world through any other port
493 than 80, or some web proxy. ftp is blocked by many firewalls.
494 Don't assume that you can send email by connecting to the local SMTP
496 port.
497 Don't assume that you can reach yourself or any node by the name
499 'localhost'. The same goes for '127.0.0.1'. You will have to try
500 both.
501 Don't assume that the host has only one network card, or that it can't
503 bind to many virtual IP addresses.
504 Don't assume a particular network device name.
506 Don't assume a particular set of "ioctl"s will work.
508 Don't assume that you can ping hosts and get replies.
510 Don't assume that any particular port (service) will respond.
512 Don't assume that "Sys::Hostname" (or any other API or command) returns
514 either a fully qualified hostname or a non-qualified hostname: it all
515 depends on how the system had been configured. Also remember that for
516 things such as DHCP and NAT, the hostname you get back might not be
517 very useful.
518 All the above don'ts may look daunting, and they are, but the key is to
520 degrade gracefully if one cannot reach the particular network service
521 one wants. Croaking or hanging do not look very professional.
522 Interprocess Communication (IPC)
524 In general, don't directly access the system in code meant to be
525 portable. That means, no "system", "exec", "fork", "pipe", "``" or
526 "qx//", "open" with a "|", nor any of the other things that makes being
527 a Perl hacker worth being.
528 Commands that launch external processes are generally supported on most
530 platforms (though many of them do not support any type of forking).
531 The problem with using them arises from what you invoke them on.
532 External tools are often named differently on different platforms, may
533 not be available in the same location, might accept different
534 arguments, can behave differently, and often present their results in a
535 platform-dependent way. Thus, you should seldom depend on them to
536 produce consistent results. (Then again, if you're calling "netstat
537 -a", you probably don't expect it to run on both Unix and CP/M.)
538 One especially common bit of Perl code is opening a pipe to sendmail:
540 open(my $mail, '|-', '/usr/lib/sendmail -t')
542 or die "cannot fork sendmail: $!";
543 This is fine for systems programming when sendmail is known to be
545 available. But it is not fine for many non-Unix systems, and even some
546 Unix systems that may not have sendmail installed. If a portable
547 solution is needed, see the various distributions on CPAN that deal
548 with it. "Mail::Mailer" and "Mail::Send" in the "MailTools"
549 distribution are commonly used, and provide several mailing methods,
550 including "mail", "sendmail", and direct SMTP (via "Net::SMTP") if a
551 mail transfer agent is not available. "Mail::Sendmail" is a standalone
552 module that provides simple, platform-independent mailing.
553 The Unix System V IPC ("msg*(), sem*(), shm*()") is not available even
555 on all Unix platforms.
556 Do not use either the bare result of "pack("N", 10, 20, 30, 40)" or
558 bare v-strings (such as "v10.20.30.40") to represent IPv4 addresses:
559 both forms just pack the four bytes into network order. That this
560 would be equal to the C language "in_addr" struct (which is what the
561 socket code internally uses) is not guaranteed. To be portable use the
562 routines of the "Socket" module, such as "inet_aton", "inet_ntoa", and
563 "sockaddr_in".
564 The rule of thumb for portable code is: Do it all in portable Perl, or
566 use a module (that may internally implement it with platform-specific
567 code, but exposes a common interface).
568 External Subroutines (XS)
570 XS code can usually be made to work with any platform, but dependent
571 libraries, header files, etc., might not be readily available or
572 portable, or the XS code itself might be platform-specific, just as
573 Perl code might be. If the libraries and headers are portable, then it
574 is normally reasonable to make sure the XS code is portable, too.
575 A different type of portability issue arises when writing XS code:
577 availability of a C compiler on the end-user's system. C brings with
578 it its own portability issues, and writing XS code will expose you to
579 some of those. Writing purely in Perl is an easier way to achieve
580 portability.
581 Standard Modules
583 In general, the standard modules work across platforms. Notable
584 exceptions are the "CPAN" module (which currently makes connections to
585 external programs that may not be available), platform-specific modules
586 (like "ExtUtils::MM_VMS"), and DBM modules.
587 There is no one DBM module available on all platforms. "SDBM_File" and
589 the others are generally available on all Unix and DOSish ports, but
590 not in MacPerl, where only "NDBM_File" and "DB_File" are available.
591 The good news is that at least some DBM module should be available, and
593 "AnyDBM_File" will use whichever module it can find. Of course, then
594 the code needs to be fairly strict, dropping to the greatest common
595 factor (e.g., not exceeding 1K for each record), so that it will work
596 with any DBM module. See AnyDBM_File for more details.
597 Time and Date
599 The system's notion of time of day and calendar date is controlled in
600 widely different ways. Don't assume the timezone is stored in
601 $ENV{TZ}, and even if it is, don't assume that you can control the
602 timezone through that variable. Don't assume anything about the three-
603 letter timezone abbreviations (for example that MST would be the
604 Mountain Standard Time, it's been known to stand for Moscow Standard
605 Time). If you need to use timezones, express them in some unambiguous
606 format like the exact number of minutes offset from UTC, or the POSIX
607 timezone format.
608 Don't assume that the epoch starts at 00:00:00, January 1, 1970,
610 because that is OS- and implementation-specific. It is better to store
611 a date in an unambiguous representation. The ISO 8601 standard defines
612 YYYY-MM-DD as the date format, or YYYY-MM-DDTHH:MM:SS (that's a literal
613 "T" separating the date from the time). Please do use the ISO 8601
614 instead of making us guess what date 02/03/04 might be. ISO 8601 even
615 sorts nicely as-is. A text representation (like "1987-12-18") can be
616 easily converted into an OS-specific value using a module like
617 "Time::Piece" (see "Date Parsing" in Time::Piece) or "Date::Parse". An
618 array of values, such as those returned by "localtime", can be
619 converted to an OS-specific representation using "Time::Local".
620 When calculating specific times, such as for tests in time or date
622 modules, it may be appropriate to calculate an offset for the epoch.
623 use Time::Local qw(timegm);
625 my $offset = timegm(0, 0, 0, 1, 0, 1970);
626 The value for $offset in Unix will be 0, but in Mac OS Classic will be
628 some large number. $offset can then be added to a Unix time value to
629 get what should be the proper value on any system.
630 Character sets and character encoding
632 Assume very little about character sets.
633 Assume nothing about numerical values ("ord", "chr") of characters. Do
635 not use explicit code point ranges (like "\xHH-\xHH)". However,
636 starting in Perl v5.22, regular expression pattern bracketed character
637 class ranges specified like "qr/[\N{U+HH}-\N{U+HH}]/" are portable, and
638 starting in Perl v5.24, the same ranges are portable in "tr///". You
639 can portably use symbolic character classes like "[:print:]".
640 Do not assume that the alphabetic characters are encoded contiguously
642 (in the numeric sense). There may be gaps. Special coding in Perl,
643 however, guarantees that all subsets of "qr/[A-Z]/", "qr/[a-z]/", and
644 "qr/[0-9]/" behave as expected. "tr///" behaves the same for these
645 ranges. In patterns, any ranges specified with end points using the
646 "\N{...}" notations ensures character set portability, but it is a bug
647 in Perl v5.22 that this isn't true of "tr///", fixed in v5.24.
648 Do not assume anything about the ordering of the characters. The
650 lowercase letters may come before or after the uppercase letters; the
651 lowercase and uppercase may be interlaced so that both "a" and "A" come
652 before "b"; the accented and other international characters may be
653 interlaced so that ae comes before "b". Unicode::Collate can be used
654 to sort this all out.
655 Internationalisation
657 If you may assume POSIX (a rather large assumption), you may read more
658 about the POSIX locale system from perllocale. The locale system at
659 least attempts to make things a little bit more portable, or at least
660 more convenient and native-friendly for non-English users. The system
661 affects character sets and encoding, and date and time
662 formatting--amongst other things.
663 If you really want to be international, you should consider Unicode.
665 See perluniintro and perlunicode for more information.
666 By default Perl assumes your source code is written in an 8-bit ASCII
668 superset. To embed Unicode characters in your strings and regexes, you
669 can use the "\x{HH}" or (more portably) "\N{U+HH}" notations. You can
670 also use the "utf8" pragma and write your code in UTF-8, which lets you
671 use Unicode characters directly (not just in quoted constructs but also
672 in identifiers).
673 System Resources
675 If your code is destined for systems with severely constrained (or
676 missing!) virtual memory systems then you want to be especially mindful
677 of avoiding wasteful constructs such as:
678 my @lines = <$very_large_file>; # bad
680 while (<$fh>) {$file .= $_} # sometimes bad
682 my $file = join('', <$fh>); # better
683 The last two constructs may appear unintuitive to most people. The
685 first repeatedly grows a string, whereas the second allocates a large
686 chunk of memory in one go. On some systems, the second is more
687 efficient than the first.
688 Security
690 Most multi-user platforms provide basic levels of security, usually
691 implemented at the filesystem level. Some, however, unfortunately do
692 not. Thus the notion of user id, or "home" directory, or even the
693 state of being logged-in, may be unrecognizable on many platforms. If
694 you write programs that are security-conscious, it is usually best to
695 know what type of system you will be running under so that you can
696 write code explicitly for that platform (or class of platforms).
697 Don't assume the Unix filesystem access semantics: the operating system
699 or the filesystem may be using some ACL systems, which are richer
700 languages than the usual "rwx". Even if the "rwx" exist, their
701 semantics might be different.
702 (From the security viewpoint, testing for permissions before attempting
704 to do something is silly anyway: if one tries this, there is potential
705 for race conditions. Someone or something might change the permissions
706 between the permissions check and the actual operation. Just try the
707 operation.)
708 Don't assume the Unix user and group semantics: especially, don't
710 expect $< and $> (or $( and $)) to work for switching identities (or
711 memberships).
712 Don't assume set-uid and set-gid semantics. (And even if you do, think
714 twice: set-uid and set-gid are a known can of security worms.)
715 Style
717 For those times when it is necessary to have platform-specific code,
718 consider keeping the platform-specific code in one place, making
719 porting to other platforms easier. Use the "Config" module and the
720 special variable $^O to differentiate platforms, as described in
721 "PLATFORMS".
722 Beware of the "else syndrome":
724 if ($^O eq 'MSWin32') {
726 # code that assumes Windows
727 } else {
728 # code that assumes Linux
729 }
730 The "else" branch should be used for the really ultimate fallback, not
732 for code specific to some platform.
733 Be careful in the tests you supply with your module or programs.
735 Module code may be fully portable, but its tests might not be. This
736 often happens when tests spawn off other processes or call external
737 programs to aid in the testing, or when (as noted above) the tests
738 assume certain things about the filesystem and paths. Be careful not
739 to depend on a specific output style for errors, such as when checking
740 $! after a failed system call. Using $! for anything else than
741 displaying it as output is doubtful (though see the "Errno" module for
742 testing reasonably portably for error value). Some platforms expect a
743 certain output format, and Perl on those platforms may have been
744 adjusted accordingly. Most specifically, don't anchor a regex when
745 testing an error value.
746
748 Modules uploaded to CPAN are tested by a variety of volunteers on
749 different platforms. These CPAN testers are notified by mail of each
750 new upload, and reply to the list with PASS, FAIL, NA (not applicable
751 to this platform), or UNKNOWN (unknown), along with any relevant
752 notations.
753 The purpose of the testing is twofold: one, to help developers fix any
755 problems in their code that crop up because of lack of testing on other
756 platforms; two, to provide users with information about whether a given
757 module works on a given platform.
758 Also see:
760 • Mailing list: cpan-testers-discuss@perl.org
762 • Testing results: <https://www.cpantesters.org/>
764
766 Perl is built with a $^O variable that indicates the operating system
767 it was built on. This was implemented to help speed up code that would
768 otherwise have to "use Config" and use the value of $Config{osname}.
769 Of course, to get more detailed information about the system, looking
770 into %Config is certainly recommended.
771 %Config cannot always be trusted, however, because it was built at
773 compile time. If perl was built in one place, then transferred
774 elsewhere, some values may be wrong. The values may even have been
775 edited after the fact.
776 Unix
778 Perl works on a bewildering variety of Unix and Unix-like platforms
779 (see e.g. most of the files in the hints/ directory in the source code
780 kit). On most of these systems, the value of $^O (hence
781 $Config{osname}, too) is determined either by lowercasing and stripping
782 punctuation from the first field of the string returned by typing
783 "uname -a" (or a similar command) at the shell prompt or by testing the
784 file system for the presence of uniquely named files such as a kernel
785 or header file. Here, for example, are a few of the more popular Unix
786 flavors:
787 uname $^O $Config{archname}
789 --------------------------------------------
790 AIX aix aix
791 BSD/OS bsdos i386-bsdos
792 Darwin darwin darwin
793 DYNIX/ptx dynixptx i386-dynixptx
794 FreeBSD freebsd freebsd-i386
795 Haiku haiku BePC-haiku
796 Linux linux arm-linux
797 Linux linux armv5tel-linux
798 Linux linux i386-linux
799 Linux linux i586-linux
800 Linux linux ppc-linux
801 HP-UX hpux PA-RISC1.1
802 IRIX irix irix
803 Mac OS X darwin darwin
804 NeXT 3 next next-fat
805 NeXT 4 next OPENSTEP-Mach
806 openbsd openbsd i386-openbsd
807 OSF1 dec_osf alpha-dec_osf
808 reliantunix-n svr4 RM400-svr4
809 SCO_SV sco_sv i386-sco_sv
810 SINIX-N svr4 RM400-svr4
811 sn4609 unicos CRAY_C90-unicos
812 sn6521 unicosmk t3e-unicosmk
813 sn9617 unicos CRAY_J90-unicos
814 SunOS solaris sun4-solaris
815 SunOS solaris i86pc-solaris
816 SunOS4 sunos sun4-sunos
817 Because the value of $Config{archname} may depend on the hardware
819 architecture, it can vary more than the value of $^O.
820 DOS and Derivatives
822 Perl has long been ported to Intel-style microcomputers running under
823 systems like PC-DOS, MS-DOS, OS/2, and most Windows platforms you can
824 bring yourself to mention (except for Windows CE, if you count that).
825 Users familiar with COMMAND.COM or CMD.EXE style shells should be aware
826 that each of these file specifications may have subtle differences:
827 my $filespec0 = "c:/foo/bar/file.txt";
829 my $filespec1 = "c:\\foo\\bar\\file.txt";
830 my $filespec2 = 'c:\foo\bar\file.txt';
831 my $filespec3 = 'c:\\foo\\bar\\file.txt';
832 System calls accept either "/" or "\" as the path separator. However,
834 many command-line utilities of DOS vintage treat "/" as the option
835 prefix, so may get confused by filenames containing "/". Aside from
836 calling any external programs, "/" will work just fine, and probably
837 better, as it is more consistent with popular usage, and avoids the
838 problem of remembering what to backwhack and what not to.
839 The DOS FAT filesystem can accommodate only "8.3" style filenames.
841 Under the "case-insensitive, but case-preserving" HPFS (OS/2) and NTFS
842 (NT) filesystems you may have to be careful about case returned with
843 functions like "readdir" or used with functions like "open" or
844 "opendir".
845 DOS also treats several filenames as special, such as AUX, PRN, NUL,
847 CON, COM1, LPT1, LPT2, etc. Unfortunately, sometimes these filenames
848 won't even work if you include an explicit directory prefix. It is
849 best to avoid such filenames, if you want your code to be portable to
850 DOS and its derivatives. It's hard to know what these all are,
851 unfortunately.
852 Users of these operating systems may also wish to make use of scripts
854 such as pl2bat.bat to put wrappers around your scripts.
855 Newline ("\n") is translated as "\015\012" by the I/O system when
857 reading from and writing to files (see "Newlines").
858 "binmode($filehandle)" will keep "\n" translated as "\012" for that
859 filehandle. "binmode" should always be used for code that deals with
860 binary data. That's assuming you realize in advance that your data is
861 in binary. General-purpose programs should often assume nothing about
862 their data.
863 The $^O variable and the $Config{archname} values for various DOSish
865 perls are as follows:
866 OS $^O $Config{archname} ID Version
868 ---------------------------------------------------------
869 MS-DOS dos ?
870 PC-DOS dos ?
871 OS/2 os2 ?
872 Windows 3.1 ? ? 0 3 01
873 Windows 95 MSWin32 MSWin32-x86 1 4 00
874 Windows 98 MSWin32 MSWin32-x86 1 4 10
875 Windows ME MSWin32 MSWin32-x86 1 ?
876 Windows NT MSWin32 MSWin32-x86 2 4 xx
877 Windows NT MSWin32 MSWin32-ALPHA 2 4 xx
878 Windows NT MSWin32 MSWin32-ppc 2 4 xx
879 Windows 2000 MSWin32 MSWin32-x86 2 5 00
880 Windows XP MSWin32 MSWin32-x86 2 5 01
881 Windows 2003 MSWin32 MSWin32-x86 2 5 02
882 Windows Vista MSWin32 MSWin32-x86 2 6 00
883 Windows 7 MSWin32 MSWin32-x86 2 6 01
884 Windows 7 MSWin32 MSWin32-x64 2 6 01
885 Windows 2008 MSWin32 MSWin32-x86 2 6 01
886 Windows 2008 MSWin32 MSWin32-x64 2 6 01
887 Windows CE MSWin32 ? 3
888 Cygwin cygwin cygwin
889 The various MSWin32 Perl's can distinguish the OS they are running on
891 via the value of the fifth element of the list returned from
892 "Win32::GetOSVersion()". For example:
893 if ($^O eq 'MSWin32') {
895 my @os_version_info = Win32::GetOSVersion();
896 print +('3.1','95','NT')[$os_version_info[4]],"\n";
897 }
898 There are also "Win32::IsWinNT()|Win32/Win32::IsWinNT()",
900 "Win32::IsWin95()|Win32/Win32::IsWin95()", and "Win32::GetOSName()";
901 try "perldoc Win32". The very portable "POSIX::uname()" will work too:
902 c:\> perl -MPOSIX -we "print join '|', uname"
904 Windows NT|moonru|5.0|Build 2195 (Service Pack 2)|x86
905 Errors set by Winsock functions are now put directly into $^E, and the
907 relevant "WSAE*" error codes are now exported from the Errno and POSIX
908 modules for testing this against.
909 The previous behavior of putting the errors (converted to POSIX-style
911 "E*" error codes since Perl 5.20.0) into $! was buggy due to the non-
912 equivalence of like-named Winsock and POSIX error constants, a
913 relationship between which has unfortunately been established in one
914 way or another since Perl 5.8.0.
915 The new behavior provides a much more robust solution for checking
917 Winsock errors in portable software without accidentally matching POSIX
918 tests that were intended for other OSes and may have different meanings
919 for Winsock.
920 The old behavior is currently retained, warts and all, for backwards
922 compatibility, but users are encouraged to change any code that tests
923 $! against "E*" constants for Winsock errors to instead test $^E
924 against "WSAE*" constants. After a suitable deprecation period, which
925 started with Perl 5.24, the old behavior may be removed, leaving $!
926 unchanged after Winsock function calls, to avoid any possible confusion
927 over which error variable to check.
928 Also see:
930 • The djgpp environment for DOS, <http://www.delorie.com/djgpp/> and
932 perldos.
933 • The EMX environment for DOS, OS/2, etc. emx@iaehv.nl,
935 <ftp://hobbes.nmsu.edu/pub/os2/dev/emx/> Also perlos2.
936 • Build instructions for Win32 in perlwin32, or under the Cygnus
938 environment in perlcygwin.
939 • The "Win32::*" modules in Win32.
941 • The ActiveState Pages, <https://www.activestate.com/>
943 • The Cygwin environment for Win32; README.cygwin (installed as
945 perlcygwin), <https://www.cygwin.com/>
946 • The U/WIN environment for Win32,
948 <http://www.research.att.com/sw/tools/uwin/>
949 • Build instructions for OS/2, perlos2
951 VMS
953 Perl on VMS is discussed in perlvms in the Perl distribution.
954 The official name of VMS as of this writing is OpenVMS.
956 Interacting with Perl from the Digital Command Language (DCL) shell
958 often requires a different set of quotation marks than Unix shells do.
959 For example:
960 $ perl -e "print ""Hello, world.\n"""
962 Hello, world.
963 There are several ways to wrap your Perl scripts in DCL .COM files, if
965 you are so inclined. For example:
966 $ write sys$output "Hello from DCL!"
968 $ if p1 .eqs. ""
969 $ then perl -x 'f$environment("PROCEDURE")
970 $ else perl -x - 'p1 'p2 'p3 'p4 'p5 'p6 'p7 'p8
971 $ deck/dollars="__END__"
972 #!/usr/bin/perl
973 print "Hello from Perl!\n";
975 __END__
977 $ endif
978 Do take care with "$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT" if your
980 Perl-in-DCL script expects to do things like "$read = <STDIN>;".
981 The VMS operating system has two filesystems, designated by their on-
983 disk structure (ODS) level: ODS-2 and its successor ODS-5. The initial
984 port of Perl to VMS pre-dates ODS-5, but all current testing and
985 development assumes ODS-5 and its capabilities, including case
986 preservation, extended characters in filespecs, and names up to 8192
987 bytes long.
988 Perl on VMS can accept either VMS- or Unix-style file specifications as
990 in either of the following:
991 $ perl -ne "print if /perl_setup/i" SYS$LOGIN:LOGIN.COM
993 $ perl -ne "print if /perl_setup/i" /sys$login/login.com
994 but not a mixture of both as in:
996 $ perl -ne "print if /perl_setup/i" sys$login:/login.com
998 Can't open sys$login:/login.com: file specification syntax error
999 In general, the easiest path to portability is always to specify
1001 filenames in Unix format unless they will need to be processed by
1002 native commands or utilities. Because of this latter consideration,
1003 the File::Spec module by default returns native format specifications
1004 regardless of input format. This default may be reversed so that
1005 filenames are always reported in Unix format by specifying the
1006 "DECC$FILENAME_UNIX_REPORT" feature logical in the environment.
1007 The file type, or extension, is always present in a VMS-format file
1009 specification even if it's zero-length. This means that, by default,
1010 "readdir" will return a trailing dot on a file with no extension, so
1011 where you would see "a" on Unix you'll see "a." on VMS. However, the
1012 trailing dot may be suppressed by enabling the
1013 "DECC$READDIR_DROPDOTNOTYPE" feature in the environment (see the CRTL
1014 documentation on feature logical names).
1015 What "\n" represents depends on the type of file opened. It usually
1017 represents "\012" but it could also be "\015", "\012", "\015\012",
1018 "\000", "\040", or nothing depending on the file organization and
1019 record format. The "VMS::Stdio" module provides access to the special
1020 "fopen()" requirements of files with unusual attributes on VMS.
1021 The value of $^O on OpenVMS is "VMS". To determine the architecture
1023 that you are running on refer to $Config{archname}.
1024 On VMS, perl determines the UTC offset from the
1026 "SYS$TIMEZONE_DIFFERENTIAL" logical name. Although the VMS epoch began
1027 at 17-NOV-1858 00:00:00.00, calls to "localtime" are adjusted to count
1028 offsets from 01-JAN-1970 00:00:00.00, just like Unix.
1029 Also see:
1031 • README.vms (installed as README_vms), perlvms
1033 • vmsperl list, vmsperl-subscribe@perl.org
1035 • vmsperl on the web, <http://www.sidhe.org/vmsperl/index.html>
1037 • VMS Software Inc. web site, <http://www.vmssoftware.com>
1039 VOS
1041 Perl on VOS (also known as OpenVOS) is discussed in README.vos in the
1042 Perl distribution (installed as perlvos). Perl on VOS can accept
1043 either VOS- or Unix-style file specifications as in either of the
1044 following:
1045 $ perl -ne "print if /perl_setup/i" >system>notices
1047 $ perl -ne "print if /perl_setup/i" /system/notices
1048 or even a mixture of both as in:
1050 $ perl -ne "print if /perl_setup/i" >system/notices
1052 Even though VOS allows the slash character to appear in object names,
1054 because the VOS port of Perl interprets it as a pathname delimiting
1055 character, VOS files, directories, or links whose names contain a slash
1056 character cannot be processed. Such files must be renamed before they
1057 can be processed by Perl.
1058 Older releases of VOS (prior to OpenVOS Release 17.0) limit file names
1060 to 32 or fewer characters, prohibit file names from starting with a "-"
1061 character, and prohibit file names from containing " " (space) or any
1062 character from the set "!#%&'()*;<=>?".
1063 Newer releases of VOS (OpenVOS Release 17.0 or later) support a feature
1065 known as extended names. On these releases, file names can contain up
1066 to 255 characters, are prohibited from starting with a "-" character,
1067 and the set of prohibited characters is reduced to "#%*<>?". There are
1068 restrictions involving spaces and apostrophes: these characters must
1069 not begin or end a name, nor can they immediately precede or follow a
1070 period. Additionally, a space must not immediately precede another
1071 space or hyphen. Specifically, the following character combinations
1072 are prohibited: space-space, space-hyphen, period-space, space-period,
1073 period-apostrophe, apostrophe-period, leading or trailing space, and
1074 leading or trailing apostrophe. Although an extended file name is
1075 limited to 255 characters, a path name is still limited to 256
1076 characters.
1077 The value of $^O on VOS is "vos". To determine the architecture that
1079 you are running on refer to $Config{archname}.
1080 Also see:
1082 • README.vos (installed as perlvos)
1084 • The VOS mailing list.
1086 There is no specific mailing list for Perl on VOS. You can contact
1088 the Stratus Technologies Customer Assistance Center (CAC) for your
1089 region, or you can use the contact information located in the
1090 distribution files on the Stratus Anonymous FTP site.
1091 • Stratus Technologies on the web at <http://www.stratus.com>
1093 • VOS Open-Source Software on the web at
1095 <http://ftp.stratus.com/pub/vos/vos.html>
1096 EBCDIC Platforms
1098 v5.22 core Perl runs on z/OS (formerly OS/390). Theoretically it could
1099 run on the successors of OS/400 on AS/400 minicomputers as well as
1100 VM/ESA, and BS2000 for S/390 Mainframes. Such computers use EBCDIC
1101 character sets internally (usually Character Code Set ID 0037 for
1102 OS/400 and either 1047 or POSIX-BC for S/390 systems).
1103 The rest of this section may need updating, but we don't know what it
1105 should say. Please submit comments to
1106 <https://github.com/Perl/perl5/issues>.
1107 On the mainframe Perl currently works under the "Unix system services
1109 for OS/390" (formerly known as OpenEdition), VM/ESA OpenEdition, or the
1110 BS200 POSIX-BC system (BS2000 is supported in Perl 5.6 and greater).
1111 See perlos390 for details. Note that for OS/400 there is also a port
1112 of Perl 5.8.1/5.10.0 or later to the PASE which is ASCII-based (as
1113 opposed to ILE which is EBCDIC-based), see perlos400.
1114 As of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix sub-
1116 systems do not support the "#!" shebang trick for script invocation.
1117 Hence, on OS/390 and VM/ESA Perl scripts can be executed with a header
1118 similar to the following simple script:
1119 : # use perl
1121 eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}'
1122 if 0;
1123 #!/usr/local/bin/perl # just a comment really
1124 print "Hello from perl!\n";
1126 OS/390 will support the "#!" shebang trick in release 2.8 and beyond.
1128 Calls to "system" and backticks can use POSIX shell syntax on all S/390
1129 systems.
1130 On the AS/400, if PERL5 is in your library list, you may need to wrap
1132 your Perl scripts in a CL procedure to invoke them like so:
1133 BEGIN
1135 CALL PGM(PERL5/PERL) PARM('/QOpenSys/hello.pl')
1136 ENDPGM
1137 This will invoke the Perl script hello.pl in the root of the QOpenSys
1139 file system. On the AS/400 calls to "system" or backticks must use CL
1140 syntax.
1141 On these platforms, bear in mind that the EBCDIC character set may have
1143 an effect on what happens with some Perl functions (such as "chr",
1144 "pack", "print", "printf", "ord", "sort", "sprintf", "unpack"), as well
1145 as bit-fiddling with ASCII constants using operators like "^", "&" and
1146 "|", not to mention dealing with socket interfaces to ASCII computers
1147 (see "Newlines").
1148 Fortunately, most web servers for the mainframe will correctly
1150 translate the "\n" in the following statement to its ASCII equivalent
1151 ("\r" is the same under both Unix and z/OS):
1152 print "Content-type: text/html\r\n\r\n";
1154 The values of $^O on some of these platforms include:
1156 uname $^O $Config{archname}
1158 --------------------------------------------
1159 OS/390 os390 os390
1160 OS400 os400 os400
1161 POSIX-BC posix-bc BS2000-posix-bc
1162 Some simple tricks for determining if you are running on an EBCDIC
1164 platform could include any of the following (perhaps all):
1165 if ("\t" eq "\005") { print "EBCDIC may be spoken here!\n"; }
1167 if (ord('A') == 193) { print "EBCDIC may be spoken here!\n"; }
1169 if (chr(169) eq 'z') { print "EBCDIC may be spoken here!\n"; }
1171 One thing you may not want to rely on is the EBCDIC encoding of
1173 punctuation characters since these may differ from code page to code
1174 page (and once your module or script is rumoured to work with EBCDIC,
1175 folks will want it to work with all EBCDIC character sets).
1176 Also see:
1178 • perlos390, perlos400, perlbs2000, perlebcdic.
1180 • The perl-mvs@perl.org list is for discussion of porting issues as
1182 well as general usage issues for all EBCDIC Perls. Send a message
1183 body of "subscribe perl-mvs" to majordomo@perl.org.
1184 • AS/400 Perl information at <http://as400.rochester.ibm.com/> as
1186 well as on CPAN in the ports/ directory.
1187 Acorn RISC OS
1189 Because Acorns use ASCII with newlines ("\n") in text files as "\012"
1190 like Unix, and because Unix filename emulation is turned on by default,
1191 most simple scripts will probably work "out of the box". The native
1192 filesystem is modular, and individual filesystems are free to be case-
1193 sensitive or insensitive, and are usually case-preserving. Some native
1194 filesystems have name length limits, which file and directory names are
1195 silently truncated to fit. Scripts should be aware that the standard
1196 filesystem currently has a name length limit of 10 characters, with up
1197 to 77 items in a directory, but other filesystems may not impose such
1198 limitations.
1199 Native filenames are of the form
1201 Filesystem#Special_Field::DiskName.$.Directory.Directory.File
1203 where
1205 Special_Field is not usually present, but may contain . and $ .
1207 Filesystem =~ m|[A-Za-z0-9_]|
1208 DsicName =~ m|[A-Za-z0-9_/]|
1209 $ represents the root directory
1210 . is the path separator
1211 @ is the current directory (per filesystem but machine global)
1212 ^ is the parent directory
1213 Directory and File =~ m|[^\0- "\.\$\%\&:\@\\^\|\177]+|
1214 The default filename translation is roughly "tr|/.|./|", swapping dots
1216 and slashes.
1217 Note that ""ADFS::HardDisk.$.File" ne 'ADFS::HardDisk.$.File'" and that
1219 the second stage of "$" interpolation in regular expressions will fall
1220 foul of the $. variable if scripts are not careful.
1221 Logical paths specified by system variables containing comma-separated
1223 search lists are also allowed; hence "System:Modules" is a valid
1224 filename, and the filesystem will prefix "Modules" with each section of
1225 "System$Path" until a name is made that points to an object on disk.
1226 Writing to a new file "System:Modules" would be allowed only if
1227 "System$Path" contains a single item list. The filesystem will also
1228 expand system variables in filenames if enclosed in angle brackets, so
1229 "<System$Dir>.Modules" would look for the file
1230 "$ENV{'System$Dir'} . 'Modules'". The obvious implication of this is
1231 that fully qualified filenames can start with "<>" and the three-
1232 argument form of "open" should always be used.
1233 Because "." was in use as a directory separator and filenames could not
1235 be assumed to be unique after 10 characters, Acorn implemented the C
1236 compiler to strip the trailing ".c" ".h" ".s" and ".o" suffix from
1237 filenames specified in source code and store the respective files in
1238 subdirectories named after the suffix. Hence files are translated:
1239 foo.h h.foo
1241 C:foo.h C:h.foo (logical path variable)
1242 sys/os.h sys.h.os (C compiler groks Unix-speak)
1243 10charname.c c.10charname
1244 10charname.o o.10charname
1245 11charname_.c c.11charname (assuming filesystem truncates at 10)
1246 The Unix emulation library's translation of filenames to native assumes
1248 that this sort of translation is required, and it allows a user-defined
1249 list of known suffixes that it will transpose in this fashion. This
1250 may seem transparent, but consider that with these rules foo/bar/baz.h
1251 and foo/bar/h/baz both map to foo.bar.h.baz, and that "readdir" and
1252 "glob" cannot and do not attempt to emulate the reverse mapping. Other
1253 "."'s in filenames are translated to "/".
1254 As implied above, the environment accessed through %ENV is global, and
1256 the convention is that program specific environment variables are of
1257 the form "Program$Name". Each filesystem maintains a current
1258 directory, and the current filesystem's current directory is the global
1259 current directory. Consequently, sociable programs don't change the
1260 current directory but rely on full pathnames, and programs (and
1261 Makefiles) cannot assume that they can spawn a child process which can
1262 change the current directory without affecting its parent (and everyone
1263 else for that matter).
1264 Because native operating system filehandles are global and are
1266 currently allocated down from 255, with 0 being a reserved value, the
1267 Unix emulation library emulates Unix filehandles. Consequently, you
1268 can't rely on passing "STDIN", "STDOUT", or "STDERR" to your children.
1269 The desire of users to express filenames of the form "<Foo$Dir>.Bar" on
1271 the command line unquoted causes problems, too: "``" command output
1272 capture has to perform a guessing game. It assumes that a string
1273 "<[^<>]+\$[^<>]>" is a reference to an environment variable, whereas
1274 anything else involving "<" or ">" is redirection, and generally
1275 manages to be 99% right. Of course, the problem remains that scripts
1276 cannot rely on any Unix tools being available, or that any tools found
1277 have Unix-like command line arguments.
1278 Extensions and XS are, in theory, buildable by anyone using free tools.
1280 In practice, many don't, as users of the Acorn platform are used to
1281 binary distributions. MakeMaker does run, but no available make
1282 currently copes with MakeMaker's makefiles; even if and when this
1283 should be fixed, the lack of a Unix-like shell will cause problems with
1284 makefile rules, especially lines of the form "cd sdbm && make all", and
1285 anything using quoting.
1286 "RISC OS" is the proper name for the operating system, but the value in
1288 $^O is "riscos" (because we don't like shouting).
1289 Other perls
1291 Perl has been ported to many platforms that do not fit into any of the
1292 categories listed above. Some, such as AmigaOS, QNX, Plan 9, and VOS,
1293 have been well-integrated into the standard Perl source code kit. You
1294 may need to see the ports/ directory on CPAN for information, and
1295 possibly binaries, for the likes of: aos, Atari ST, lynxos, riscos,
1296 Novell Netware, Tandem Guardian, etc. (Yes, we know that some of these
1297 OSes may fall under the Unix category, but we are not a standards
1298 body.)
1299 Some approximate operating system names and their $^O values in the
1301 "OTHER" category include:
1302 OS $^O $Config{archname}
1304 ------------------------------------------
1305 Amiga DOS amigaos m68k-amigos
1306 See also:
1308 • Amiga, README.amiga (installed as perlamiga).
1310 • A free perl5-based PERL.NLM for Novell Netware is available in
1312 precompiled binary and source code form from
1313 <http://www.novell.com/> as well as from CPAN.
1314 • Plan 9, README.plan9
1316
1318 Listed below are functions that are either completely unimplemented or
1319 else have been implemented differently on various platforms. Preceding
1320 each description will be, in parentheses, a list of platforms that the
1321 description applies to.
1322 The list may well be incomplete, or even wrong in some places. When in
1324 doubt, consult the platform-specific README files in the Perl source
1325 distribution, and any other documentation resources accompanying a
1326 given port.
1327 Be aware, moreover, that even among Unix-ish systems there are
1329 variations.
1330 For many functions, you can also query %Config, exported by default
1332 from the "Config" module. For example, to check whether the platform
1333 has the "lstat" call, check $Config{d_lstat}. See Config for a full
1334 description of available variables.
1335 Alphabetical Listing of Perl Functions
1337 -X (Win32) "-w" only inspects the read-only file attribute
1338 (FILE_ATTRIBUTE_READONLY), which determines whether the
1339 directory can be deleted, not whether it can be written to.
1340 Directories always have read and write access unless denied by
1341 discretionary access control lists (DACLs).
1342 (VMS) "-r", "-w", "-x", and "-o" tell whether the file is
1344 accessible, which may not reflect UIC-based file protections.
1345 (RISC OS) "-s" by name on an open file will return the space
1347 reserved on disk, rather than the current extent. "-s" on an
1348 open filehandle returns the current size.
1349 (Win32, VMS, RISC OS) "-R", "-W", "-X", "-O" are
1351 indistinguishable from "-r", "-w", "-x", "-o".
1352 (Win32, VMS, RISC OS) "-g", "-k", "-l", "-u", "-A" are not
1354 particularly meaningful.
1355 (Win32) "-l" returns true for both symlinks and directory
1357 junctions.
1358 (VMS, RISC OS) "-p" is not particularly meaningful.
1360 (VMS) "-d" is true if passed a device spec without an explicit
1362 directory.
1363 (Win32) "-x" (or "-X") determine if a file ends in one of the
1365 executable suffixes. "-S" is meaningless.
1366 (RISC OS) "-x" (or "-X") determine if a file has an executable
1368 file type.
1369 alarm (Win32) Emulated using timers that must be explicitly polled
1371 whenever Perl wants to dispatch "safe signals" and therefore
1372 cannot interrupt blocking system calls.
1373 atan2 (Tru64, HP-UX 10.20) Due to issues with various CPUs, math
1375 libraries, compilers, and standards, results for "atan2" may
1376 vary depending on any combination of the above. Perl attempts
1377 to conform to the Open Group/IEEE standards for the results
1378 returned from "atan2", but cannot force the issue if the system
1379 Perl is run on does not allow it.
1380 The current version of the standards for "atan2" is available
1382 at
1383 <http://www.opengroup.org/onlinepubs/009695399/functions/atan2.html>.
1384 binmode (RISC OS) Meaningless.
1386 (VMS) Reopens file and restores pointer; if function fails,
1388 underlying filehandle may be closed, or pointer may be in a
1389 different position.
1390 (Win32) The value returned by "tell" may be affected after the
1392 call, and the filehandle may be flushed.
1393 chdir (Win32) The current directory reported by the system may
1395 include any symbolic links specified to chdir().
1396 chmod (Win32) Only good for changing "owner" read-write access;
1398 "group" and "other" bits are meaningless.
1399 (RISC OS) Only good for changing "owner" and "other" read-write
1401 access.
1402 (VOS) Access permissions are mapped onto VOS access-control
1404 list changes.
1405 (Cygwin) The actual permissions set depend on the value of the
1407 "CYGWIN" variable in the SYSTEM environment settings.
1408 (Android) Setting the exec bit on some locations (generally
1410 /sdcard) will return true but not actually set the bit.
1411 (VMS) A mode argument of zero sets permissions to the user's
1413 default permission mask rather than disabling all permissions.
1414 chown (Plan 9, RISC OS) Not implemented.
1416 (Win32) Does nothing, but won't fail.
1418 (VOS) A little funky, because VOS's notion of ownership is a
1420 little funky.
1421 chroot (Win32, VMS, Plan 9, RISC OS, VOS) Not implemented.
1423 crypt (Win32) May not be available if library or source was not
1425 provided when building perl.
1426 (Android) Not implemented.
1428 dbmclose
1430 (VMS, Plan 9, VOS) Not implemented.
1431 dbmopen (VMS, Plan 9, VOS) Not implemented.
1433 dump (RISC OS) Not useful.
1435 (Cygwin, Win32) Not supported.
1437 (VMS) Invokes VMS debugger.
1439 exec (Win32) "exec LIST" without the use of indirect object syntax
1441 ("exec PROGRAM LIST") may fall back to trying the shell if the
1442 first "spawn()" fails.
1443 Note that the list form of exec() is emulated since the Win32
1445 API CreateProcess() accepts a simple string rather than an
1446 array of command-line arguments. This may have security
1447 implications for your code.
1448 (SunOS, Solaris, HP-UX) Does not automatically flush output
1450 handles on some platforms.
1451 exit (VMS) Emulates Unix "exit" (which considers "exit 1" to
1453 indicate an error) by mapping the 1 to "SS$_ABORT" (44). This
1454 behavior may be overridden with the pragma "use vmsish 'exit'".
1455 As with the CRTL's "exit()" function, "exit 0" is also mapped
1456 to an exit status of "SS$_NORMAL" (1); this mapping cannot be
1457 overridden. Any other argument to "exit" is used directly as
1458 Perl's exit status. On VMS, unless the future POSIX_EXIT mode
1459 is enabled, the exit code should always be a valid VMS exit
1460 code and not a generic number. When the POSIX_EXIT mode is
1461 enabled, a generic number will be encoded in a method
1462 compatible with the C library _POSIX_EXIT macro so that it can
1463 be decoded by other programs, particularly ones written in C,
1464 like the GNV package.
1465 (Solaris) "exit" resets file pointers, which is a problem when
1467 called from a child process (created by "fork") in "BEGIN". A
1468 workaround is to use "POSIX::_exit".
1469 exit unless $Config{archname} =~ /\bsolaris\b/;
1471 require POSIX;
1472 POSIX::_exit(0);
1473 fcntl (Win32) Not implemented.
1475 (VMS) Some functions available based on the version of VMS.
1477 flock (VMS, RISC OS, VOS) Not implemented.
1479 fork (AmigaOS, RISC OS, VMS) Not implemented.
1481 (Win32) Emulated using multiple interpreters. See perlfork.
1483 (SunOS, Solaris, HP-UX) Does not automatically flush output
1485 handles on some platforms.
1486 getlogin
1488 (RISC OS) Not implemented.
1489 getpgrp (Win32, VMS, RISC OS) Not implemented.
1491 getppid (Win32, RISC OS) Not implemented.
1493 getpriority
1495 (Win32, VMS, RISC OS, VOS) Not implemented.
1496 getpwnam
1498 (Win32) Not implemented.
1499 (RISC OS) Not useful.
1501 getgrnam
1503 (Win32, VMS, RISC OS) Not implemented.
1504 getnetbyname
1506 (Android, Win32, Plan 9) Not implemented.
1507 getpwuid
1509 (Win32) Not implemented.
1510 (RISC OS) Not useful.
1512 getgrgid
1514 (Win32, VMS, RISC OS) Not implemented.
1515 getnetbyaddr
1517 (Android, Win32, Plan 9) Not implemented.
1518 getprotobynumber
1520 (Android) Not implemented.
1521 getpwent
1523 (Android, Win32) Not implemented.
1524 getgrent
1526 (Android, Win32, VMS) Not implemented.
1527 gethostbyname
1529 (Irix 5) "gethostbyname('localhost')" does not work everywhere:
1530 you may have to use "gethostbyname('127.0.0.1')".
1531 gethostent
1533 (Win32) Not implemented.
1534 getnetent
1536 (Android, Win32, Plan 9) Not implemented.
1537 getprotoent
1539 (Android, Win32, Plan 9) Not implemented.
1540 getservent
1542 (Win32, Plan 9) Not implemented.
1543 seekdir (Android) Not implemented.
1545 sethostent
1547 (Android, Win32, Plan 9, RISC OS) Not implemented.
1548 setnetent
1550 (Win32, Plan 9, RISC OS) Not implemented.
1551 setprotoent
1553 (Android, Win32, Plan 9, RISC OS) Not implemented.
1554 setservent
1556 (Plan 9, Win32, RISC OS) Not implemented.
1557 endpwent
1559 (Win32) Not implemented.
1560 (Android) Either not implemented or a no-op.
1562 endgrent
1564 (Android, RISC OS, VMS, Win32) Not implemented.
1565 endhostent
1567 (Android, Win32) Not implemented.
1568 endnetent
1570 (Android, Win32, Plan 9) Not implemented.
1571 endprotoent
1573 (Android, Win32, Plan 9) Not implemented.
1574 endservent
1576 (Plan 9, Win32) Not implemented.
1577 getsockopt
1579 (Plan 9) Not implemented.
1580 glob This operator is implemented via the "File::Glob" extension on
1582 most platforms. See File::Glob for portability information.
1583 gmtime In theory, "gmtime" is reliable from -2**63 to 2**63-1.
1585 However, because work-arounds in the implementation use
1586 floating point numbers, it will become inaccurate as the time
1587 gets larger. This is a bug and will be fixed in the future.
1588 (VOS) Time values are 32-bit quantities.
1590 ioctl (VMS) Not implemented.
1592 (Win32) Available only for socket handles, and it does what the
1594 "ioctlsocket()" call in the Winsock API does.
1595 (RISC OS) Available only for socket handles.
1597 kill (RISC OS) Not implemented, hence not useful for taint checking.
1599 (Win32) "kill" doesn't send a signal to the identified process
1601 like it does on Unix platforms. Instead "kill($sig, $pid)"
1602 terminates the process identified by $pid, and makes it exit
1603 immediately with exit status $sig. As in Unix, if $sig is 0
1604 and the specified process exists, it returns true without
1605 actually terminating it.
1606 (Win32) "kill(-9, $pid)" will terminate the process specified
1608 by $pid and recursively all child processes owned by it. This
1609 is different from the Unix semantics, where the signal will be
1610 delivered to all processes in the same process group as the
1611 process specified by $pid.
1612 (VMS) A pid of -1 indicating all processes on the system is not
1614 currently supported.
1615 link (RISC OS, VOS) Not implemented.
1617 (AmigaOS) Link count not updated because hard links are not
1619 quite that hard (They are sort of half-way between hard and
1620 soft links).
1621 (Win32) Hard links are implemented on Win32 under NTFS only.
1623 They are natively supported on Windows 2000 and later. On
1624 Windows NT they are implemented using the Windows POSIX
1625 subsystem support and the Perl process will need Administrator
1626 or Backup Operator privileges to create hard links.
1627 (VMS) Available on 64 bit OpenVMS 8.2 and later.
1629 localtime
1631 "localtime" has the same range as "gmtime", but because time
1632 zone rules change, its accuracy for historical and future times
1633 may degrade but usually by no more than an hour.
1634 lstat (RISC OS) Not implemented.
1636 (Win32) Treats directory junctions as symlinks.
1638 msgctl
1640 msgget
1641 msgsnd
1642 msgrcv (Android, Win32, VMS, Plan 9, RISC OS, VOS) Not implemented.
1643 open (RISC OS) Open modes "|-" and "-|" are unsupported.
1645 (SunOS, Solaris, HP-UX) Opening a process does not
1647 automatically flush output handles on some platforms.
1648 (Win32) Both of modes "|-" and "-|" are supported, but the list
1650 form is emulated since the Win32 API CreateProcess() accepts a
1651 simple string rather than an array of arguments. This may have
1652 security implications for your code.
1653 readlink
1655 (VMS, RISC OS) Not implemented.
1656 (Win32) readlink() on a directory junction returns the object
1658 name, not a simple path.
1659 rename (Win32) Can't move directories between directories on different
1661 logical volumes.
1662 rewinddir
1664 (Win32) Will not cause "readdir" to re-read the directory
1665 stream. The entries already read before the "rewinddir" call
1666 will just be returned again from a cache buffer.
1667 select (Win32, VMS) Only implemented on sockets.
1669 (RISC OS) Only reliable on sockets.
1671 Note that the "select FILEHANDLE" form is generally portable.
1673 semctl
1675 semget
1676 semop (Android, Win32, VMS, RISC OS) Not implemented.
1677 setgrent
1679 (Android, VMS, Win32, RISC OS) Not implemented.
1680 setpgrp (Win32, VMS, RISC OS, VOS) Not implemented.
1682 setpriority
1684 (Win32, VMS, RISC OS, VOS) Not implemented.
1685 setpwent
1687 (Android, Win32, RISC OS) Not implemented.
1688 setsockopt
1690 (Plan 9) Not implemented.
1691 shmctl
1693 shmget
1694 shmread
1695 shmwrite
1696 (Android, Win32, VMS, RISC OS) Not implemented.
1697 sleep (Win32) Emulated using synchronization functions such that it
1699 can be interrupted by "alarm", and limited to a maximum of
1700 4294967 seconds, approximately 49 days.
1701 socketpair
1703 (RISC OS) Not implemented.
1704 (VMS) Available on 64 bit OpenVMS 8.2 and later.
1706 stat Platforms that do not have "rdev", "blksize", or "blocks" will
1708 return these as '', so numeric comparison or manipulation of
1709 these fields may cause 'not numeric' warnings.
1710 (Mac OS X) "ctime" not supported on UFS.
1712 (Win32) "ctime" is creation time instead of inode change time.
1714 (VMS) "dev" and "ino" are not necessarily reliable.
1716 (RISC OS) "mtime", "atime" and "ctime" all return the last
1718 modification time. "dev" and "ino" are not necessarily
1719 reliable.
1720 (OS/2) "dev", "rdev", "blksize", and "blocks" are not
1722 available. "ino" is not meaningful and will differ between
1723 stat calls on the same file.
1724 (Cygwin) Some versions of cygwin when doing a "stat("foo")" and
1726 not finding it may then attempt to "stat("foo.exe")".
1727 symlink (RISC OS) Not implemented.
1729 (Win32) Requires either elevated permissions or developer mode
1731 and a sufficiently recent version of Windows 10. You can check
1732 whether the current process has the required privileges using
1733 the Win32::IsSymlinkCreationAllowed() function.
1734 Since Windows needs to know whether the target is a directory
1736 or not when creating the link the target Perl will only create
1737 the link as a directory link when the target exists and is a
1738 directory.
1739 (VMS) Implemented on 64 bit VMS 8.3. VMS requires the symbolic
1741 link to be in Unix syntax if it is intended to resolve to a
1742 valid path.
1743 syscall (Win32, VMS, RISC OS, VOS) Not implemented.
1745 sysopen (Mac OS, OS/390) The traditional 0, 1, and 2 MODEs are
1747 implemented with different numeric values on some systems. The
1748 flags exported by "Fcntl" ("O_RDONLY", "O_WRONLY", "O_RDWR")
1749 should work everywhere though.
1750 system (Win32) As an optimization, may not call the command shell
1752 specified in $ENV{PERL5SHELL}. "system(1, @args)" spawns an
1753 external process and immediately returns its process
1754 designator, without waiting for it to terminate. Return value
1755 may be used subsequently in "wait" or "waitpid". Failure to
1756 "spawn()" a subprocess is indicated by setting $? to "255 <<
1757 8". $? is set in a way compatible with Unix (i.e. the exit
1758 status of the subprocess is obtained by "$? >> 8", as described
1759 in the documentation).
1760 Note that the list form of system() is emulated since the Win32
1762 API CreateProcess() accepts a simple string rather than an
1763 array of command-line arguments. This may have security
1764 implications for your code.
1765 (RISC OS) There is no shell to process metacharacters, and the
1767 native standard is to pass a command line terminated by "\n"
1768 "\r" or "\0" to the spawned program. Redirection such as ">
1769 foo" is performed (if at all) by the run time library of the
1770 spawned program. "system LIST" will call the Unix emulation
1771 library's "exec" emulation, which attempts to provide emulation
1772 of the stdin, stdout, stderr in force in the parent, provided
1773 the child program uses a compatible version of the emulation
1774 library. "system SCALAR" will call the native command line
1775 directly and no such emulation of a child Unix program will
1776 occur. Mileage will vary.
1777 (Win32) "system LIST" without the use of indirect object syntax
1779 ("system PROGRAM LIST") may fall back to trying the shell if
1780 the first "spawn()" fails.
1781 (SunOS, Solaris, HP-UX) Does not automatically flush output
1783 handles on some platforms.
1784 (VMS) As with Win32, "system(1, @args)" spawns an external
1786 process and immediately returns its process designator without
1787 waiting for the process to terminate. In this case the return
1788 value may be used subsequently in "wait" or "waitpid".
1789 Otherwise the return value is POSIX-like (shifted up by 8
1790 bits), which only allows room for a made-up value derived from
1791 the severity bits of the native 32-bit condition code (unless
1792 overridden by "use vmsish 'status'"). If the native condition
1793 code is one that has a POSIX value encoded, the POSIX value
1794 will be decoded to extract the expected exit value. For more
1795 details see "$?" in perlvms.
1796 telldir (Android) Not implemented.
1798 times (Win32) "Cumulative" times will be bogus. On anything other
1800 than Windows NT or Windows 2000, "system" time will be bogus,
1801 and "user" time is actually the time returned by the "clock()"
1802 function in the C runtime library.
1803 (RISC OS) Not useful.
1805 truncate
1807 (Older versions of VMS) Not implemented.
1808 (VOS) Truncation to same-or-shorter lengths only.
1810 (Win32) If a FILEHANDLE is supplied, it must be writable and
1812 opened in append mode (i.e., use "open(my $fh, '>>',
1813 'filename')" or "sysopen(my $fh, ..., O_APPEND|O_RDWR)". If a
1814 filename is supplied, it should not be held open elsewhere.
1815 umask Returns "undef" where unavailable.
1817 (AmigaOS) "umask" works but the correct permissions are set
1819 only when the file is finally closed.
1820 utime (VMS, RISC OS) Only the modification time is updated.
1822 (Win32) May not behave as expected. Behavior depends on the C
1824 runtime library's implementation of "utime()", and the
1825 filesystem being used. The FAT filesystem typically does not
1826 support an "access time" field, and it may limit timestamps to
1827 a granularity of two seconds.
1828 wait
1830 waitpid (Win32) Can only be applied to process handles returned for
1831 processes spawned using "system(1, ...)" or pseudo processes
1832 created with "fork".
1833 (RISC OS) Not useful.
1835
1837 The following platforms are known to build Perl 5.12 (as of April 2010,
1838 its release date) from the standard source code distribution available
1839 at <http://www.cpan.org/src>
1840 Linux (x86, ARM, IA64)
1842 HP-UX
1843 AIX
1844 Win32
1845 Windows 2000
1846 Windows XP
1847 Windows Server 2003
1848 Windows Vista
1849 Windows Server 2008
1850 Windows 7
1851 Cygwin
1852 Some tests are known to fail:
1853 • ext/XS-APItest/t/call_checker.t - see
1855 <https://github.com/Perl/perl5/issues/10750>
1856 • dist/I18N-Collate/t/I18N-Collate.t
1858 • ext/Win32CORE/t/win32core.t - may fail on recent cygwin
1860 installs.
1861 Solaris (x86, SPARC)
1863 OpenVMS
1864 Alpha (7.2 and later)
1865 I64 (8.2 and later)
1866 NetBSD
1867 FreeBSD
1868 Debian GNU/kFreeBSD
1869 Haiku
1870 Irix (6.5. What else?)
1871 OpenBSD
1872 Dragonfly BSD
1873 Midnight BSD
1874 QNX Neutrino RTOS (6.5.0)
1875 MirOS BSD
1876 Stratus OpenVOS (17.0 or later)
1877 Caveats:
1878 time_t issues that may or may not be fixed
1880 Stratus VOS / OpenVOS
1881 AIX
1882 Android
1883 FreeMINT
1884 Perl now builds with FreeMiNT/Atari. It fails a few tests, that
1885 needs some investigation.
1886 The FreeMiNT port uses GNU dld for loadable module capabilities. So
1888 ensure you have that library installed when building perl.
1889
1891 (Perl 5.20)
1892 The following platforms were supported by a previous version of Perl
1893 but have been officially removed from Perl's source code as of 5.20:
1894 AT&T 3b1
1896 (Perl 5.14)
1898 The following platforms were supported up to 5.10. They may still have
1899 worked in 5.12, but supporting code has been removed for 5.14:
1900 Windows 95
1902 Windows 98
1903 Windows ME
1904 Windows NT4
1905 (Perl 5.12)
1907 The following platforms were supported by a previous version of Perl
1908 but have been officially removed from Perl's source code as of 5.12:
1909 Atari MiNT
1911 Apollo Domain/OS
1912 Apple Mac OS 8/9
1913 Tenon Machten
1914
1916 As of July 2002 (the Perl release 5.8.0), the following platforms were
1917 able to build Perl from the standard source code distribution available
1918 at <http://www.cpan.org/src/>
1919 AIX
1921 BeOS
1922 BSD/OS (BSDi)
1923 Cygwin
1924 DG/UX
1925 DOS DJGPP 1)
1926 DYNIX/ptx
1927 EPOC R5
1928 FreeBSD
1929 HI-UXMPP (Hitachi) (5.8.0 worked but we didn't know it)
1930 HP-UX
1931 IRIX
1932 Linux
1933 Mac OS Classic
1934 Mac OS X (Darwin)
1935 MPE/iX
1936 NetBSD
1937 NetWare
1938 NonStop-UX
1939 ReliantUNIX (formerly SINIX)
1940 OpenBSD
1941 OpenVMS (formerly VMS)
1942 Open UNIX (Unixware) (since Perl 5.8.1/5.9.0)
1943 OS/2
1944 OS/400 (using the PASE) (since Perl 5.8.1/5.9.0)
1945 POSIX-BC (formerly BS2000)
1946 QNX
1947 Solaris
1948 SunOS 4
1949 SUPER-UX (NEC)
1950 Tru64 UNIX (formerly DEC OSF/1, Digital UNIX)
1951 UNICOS
1952 UNICOS/mk
1953 UTS
1954 VOS / OpenVOS
1955 Win95/98/ME/2K/XP 2)
1956 WinCE
1957 z/OS (formerly OS/390)
1958 VM/ESA
1959 1) in DOS mode either the DOS or OS/2 ports can be used
1961 2) compilers: Borland, MinGW (GCC), VC6
1962 The following platforms worked with the previous releases (5.6 and
1964 5.7), but we did not manage either to fix or to test these in time for
1965 the 5.8.0 release. There is a very good chance that many of these will
1966 work fine with the 5.8.0.
1967 BSD/OS
1969 DomainOS
1970 Hurd
1971 LynxOS
1972 MachTen
1973 PowerMAX
1974 SCO SV
1975 SVR4
1976 Unixware
1977 Windows 3.1
1978 Known to be broken for 5.8.0 (but 5.6.1 and 5.7.2 can be used):
1980 AmigaOS 3
1982 The following platforms have been known to build Perl from source in
1984 the past (5.005_03 and earlier), but we haven't been able to verify
1985 their status for the current release, either because the
1986 hardware/software platforms are rare or because we don't have an active
1987 champion on these platforms--or both. They used to work, though, so go
1988 ahead and try compiling them, and let
1989 <https://github.com/Perl/perl5/issues> know of any trouble.
1990 3b1
1992 A/UX
1993 ConvexOS
1994 CX/UX
1995 DC/OSx
1996 DDE SMES
1997 DOS EMX
1998 Dynix
1999 EP/IX
2000 ESIX
2001 FPS
2002 GENIX
2003 Greenhills
2004 ISC
2005 MachTen 68k
2006 MPC
2007 NEWS-OS
2008 NextSTEP
2009 OpenSTEP
2010 Opus
2011 Plan 9
2012 RISC/os
2013 SCO ODT/OSR
2014 Stellar
2015 SVR2
2016 TI1500
2017 TitanOS
2018 Ultrix
2019 Unisys Dynix
2020 The following platforms have their own source code distributions and
2022 binaries available via <http://www.cpan.org/ports/>
2023 Perl release
2025 OS/400 (ILE) 5.005_02
2027 Tandem Guardian 5.004
2028 The following platforms have only binaries available via
2030 <http://www.cpan.org/ports/index.html> :
2031 Perl release
2033 Acorn RISCOS 5.005_02
2035 AOS 5.002
2036 LynxOS 5.004_02
2037 Although we do suggest that you always build your own Perl from the
2039 source code, both for maximal configurability and for security, in case
2040 you are in a hurry you can check <http://www.cpan.org/ports/index.html>
2041 for binary distributions.
2042
2044 perlaix, perlamiga, perlbs2000, perlcygwin, perldos, perlebcdic,
2045 perlfreebsd, perlhurd, perlhpux, perlirix, perlmacos, perlmacosx,
2046 perlnetware, perlos2, perlos390, perlos400, perlplan9, perlqnx,
2047 perlsolaris, perltru64, perlunicode, perlvms, perlvos, perlwin32, and
2048 Win32.
2049
2051 Abigail <abigail@abigail.be>, Charles Bailey <bailey@newman.upenn.edu>,
2052 Graham Barr <gbarr@pobox.com>, Tom Christiansen <tchrist@perl.com>,
2053 Nicholas Clark <nick@ccl4.org>, Thomas Dorner <Thomas.Dorner@start.de>,
2054 Andy Dougherty <doughera@lafayette.edu>, Dominic Dunlop
2055 <domo@computer.org>, Neale Ferguson <neale@vma.tabnsw.com.au>, David J.
2056 Fiander <davidf@mks.com>, Paul Green <Paul.Green@stratus.com>, M.J.T.
2057 Guy <mjtg@cam.ac.uk>, Jarkko Hietaniemi <jhi@iki.fi>, Luther Huffman
2058 <lutherh@stratcom.com>, Nick Ing-Simmons <nick@ing-simmons.net>,
2059 Andreas J. Koenig <a.koenig@mind.de>, Markus Laker
2060 <mlaker@contax.co.uk>, Andrew M. Langmead <aml@world.std.com>, Lukas
2061 Mai <l.mai@web.de>, Larry Moore <ljmoore@freespace.net>, Paul Moore
2062 <Paul.Moore@uk.origin-it.com>, Chris Nandor <pudge@pobox.com>, Matthias
2063 Neeracher <neeracher@mac.com>, Philip Newton <pne@cpan.org>, Gary Ng
2064 <71564.1743@CompuServe.COM>, Tom Phoenix <rootbeer@teleport.com>, Andre
2065 Pirard <A.Pirard@ulg.ac.be>, Peter Prymmer <pvhp@forte.com>, Hugo van
2066 der Sanden <hv@crypt0.demon.co.uk>, Gurusamy Sarathy
2067 <gsar@activestate.com>, Paul J. Schinder <schinder@pobox.com>, Michael
2068 G Schwern <schwern@pobox.com>, Dan Sugalski <dan@sidhe.org>, Nathan
2069 Torkington <gnat@frii.com>, John Malmberg <wb8tyw@qsl.net>
2070perl v5.34.0 2021-10-18 PERLPORT(1)