Debian Codegraph @

upstream/debian-unstable/man1/perlport.1

raw
1 .\" -*- mode: troff; coding: utf-8 -*-
2 .\" Automatically generated by Pod::Man 5.0102 (Pod::Simple 3.45)
3 .\"
4 .\" Standard preamble:
5 .\" ========================================================================
6 .de Sp \" Vertical space (when we can't use .PP)
7 .if t .sp .5v
8 .if n .sp
9 ..
10 .de Vb \" Begin verbatim text
11 .ft CW
12 .nf
13 .ne \\$1
14 ..
15 .de Ve \" End verbatim text
16 .ft R
17 .fi
18 ..
19 .\" \*(C` and \*(C' are quotes in nroff, nothing in troff, for use with C<>.
20 .ie n \{\
21 . ds C` ""
22 . ds C' ""
23 'br\}
24 .el\{\
25 . ds C`
26 . ds C'
27 'br\}
28 .\"
29 .\" Escape single quotes in literal strings from groff's Unicode transform.
30 .ie \n(.g .ds Aq \(aq
31 .el .ds Aq '
32 .\"
33 .\" If the F register is >0, we'll generate index entries on stderr for
34 .\" titles (.TH), headers (.SH), subsections (.SS), items (.Ip), and index
35 .\" entries marked with X<> in POD. Of course, you'll have to process the
36 .\" output yourself in some meaningful fashion.
37 .\"
38 .\" Avoid warning from groff about undefined register 'F'.
39 .de IX
40 ..
41 .nr rF 0
42 .if \n(.g .if rF .nr rF 1
43 .if (\n(rF:(\n(.g==0)) \{\
44 . if \nF \{\
45 . de IX
46 . tm Index:\\$1\t\\n%\t"\\$2"
47 ..
48 . if !\nF==2 \{\
49 . nr % 0
50 . nr F 2
51 . \}
52 . \}
53 .\}
54 .rr rF
55 .\" ========================================================================
56 .\"
57 .IX Title "PERLPORT 1"
58 .TH PERLPORT 1 2025-11-16 "perl v5.40.1" "Perl Programmers Reference Guide"
59 .\" For nroff, turn off justification. Always turn off hyphenation; it makes
60 .\" way too many mistakes in technical documents.
61 .if n .ad l
62 .nh
63 .SH NAME
64 perlport \- Writing portable Perl
65 .SH DESCRIPTION
66 .IX Header "DESCRIPTION"
67 Perl runs on numerous operating systems. While most of them share
68 much in common, they also have their own unique features.
69 .PP
70 This document is meant to help you to find out what constitutes portable
71 Perl code. That way once you make a decision to write portably,
72 you know where the lines are drawn, and you can stay within them.
73 .PP
74 There is a tradeoff between taking full advantage of one particular
75 type of computer and taking advantage of a full range of them.
76 Naturally, as you broaden your range and become more diverse, the
77 common factors drop, and you are left with an increasingly smaller
78 area of common ground in which you can operate to accomplish a
79 particular task. Thus, when you begin attacking a problem, it is
80 important to consider under which part of the tradeoff curve you
81 want to operate. Specifically, you must decide whether it is
82 important that the task that you are coding has the full generality
83 of being portable, or whether to just get the job done right now.
84 This is the hardest choice to be made. The rest is easy, because
85 Perl provides many choices, whichever way you want to approach your
86 problem.
87 .PP
88 Looking at it another way, writing portable code is usually about
89 willfully limiting your available choices. Naturally, it takes
90 discipline and sacrifice to do that. The product of portability
91 and convenience may be a constant. You have been warned.
92 .PP
93 Be aware of two important points:
94 .IP "Not all Perl programs have to be portable" 4
95 .IX Item "Not all Perl programs have to be portable"
96 There is no reason you should not use Perl as a language to glue Unix
97 tools together, or to prototype a Macintosh application, or to manage the
98 Windows registry. If it makes no sense to aim for portability for one
99 reason or another in a given program, then don't bother.
100 .IP "Nearly all of Perl already \fIis\fR portable" 4
101 .IX Item "Nearly all of Perl already is portable"
102 Don't be fooled into thinking that it is hard to create portable Perl
103 code. It isn't. Perl tries its level-best to bridge the gaps between
104 what's available on different platforms, and all the means available to
105 use those features. Thus almost all Perl code runs on any machine
106 without modification. But there are some significant issues in
107 writing portable code, and this document is entirely about those issues.
108 .PP
109 Here's the general rule: When you approach a task commonly done
110 using a whole range of platforms, think about writing portable
111 code. That way, you don't sacrifice much by way of the implementation
112 choices you can avail yourself of, and at the same time you can give
113 your users lots of platform choices. On the other hand, when you have to
114 take advantage of some unique feature of a particular platform, as is
115 often the case with systems programming (whether for Unix, Windows,
116 VMS, etc.), consider writing platform-specific code.
117 .PP
118 When the code will run on only two or three operating systems, you
119 may need to consider only the differences of those particular systems.
120 The important thing is to decide where the code will run and to be
121 deliberate in your decision.
122 .PP
123 The material below is separated into three main sections: main issues of
124 portability ("ISSUES"), platform-specific issues ("PLATFORMS"), and
125 built-in Perl functions that behave differently on various ports
126 ("FUNCTION IMPLEMENTATIONS").
127 .PP
128 This information should not be considered complete; it includes possibly
129 transient information about idiosyncrasies of some of the ports, almost
130 all of which are in a state of constant evolution. Thus, this material
131 should be considered a perpetual work in progress
132 (\f(CW\*(C`<IMG SRC="yellow_sign.gif" ALT="Under Construction">\*(C'\fR).
133 .SH ISSUES
134 .IX Header "ISSUES"
135 .SS Newlines
136 .IX Subsection "Newlines"
137 In most operating systems, lines in files are terminated by newlines.
138 Just what is used as a newline may vary from OS to OS. Unix
139 traditionally uses \f(CW\*(C`\e012\*(C'\fR, one type of DOSish I/O uses \f(CW\*(C`\e015\e012\*(C'\fR,
140 Mac\ OS uses \f(CW\*(C`\e015\*(C'\fR, and z/OS uses \f(CW\*(C`\e025\*(C'\fR.
141 .PP
142 Perl uses \f(CW\*(C`\en\*(C'\fR to represent the "logical" newline, where what is
143 logical may depend on the platform in use. In MacPerl, \f(CW\*(C`\en\*(C'\fR always
144 means \f(CW\*(C`\e015\*(C'\fR. On EBCDIC platforms, \f(CW\*(C`\en\*(C'\fR could be \f(CW\*(C`\e025\*(C'\fR or \f(CW\*(C`\e045\*(C'\fR.
145 In DOSish perls, \f(CW\*(C`\en\*(C'\fR usually means \f(CW\*(C`\e012\*(C'\fR, but when
146 accessing a file in "text" mode, perl uses the \f(CW\*(C`:crlf\*(C'\fR layer that
147 translates it to (or from) \f(CW\*(C`\e015\e012\*(C'\fR, depending on whether you're
148 reading or writing. Unix does the same thing on ttys in canonical
149 mode. \f(CW\*(C`\e015\e012\*(C'\fR is commonly referred to as CRLF.
150 .PP
151 To trim trailing newlines from text lines use
152 \&\f(CW\*(C`chomp\*(C'\fR. With default settings that function
153 looks for a trailing \f(CW\*(C`\en\*(C'\fR character and thus trims in a portable way.
154 .PP
155 When dealing with binary files (or text files in binary mode) be sure
156 to explicitly set \f(CW$/\fR to the appropriate value for
157 your file format before using \f(CW\*(C`chomp\*(C'\fR.
158 .PP
159 Because of the "text" mode translation, DOSish perls have limitations in
160 using \f(CW\*(C`seek\*(C'\fR and
161 \&\f(CW\*(C`tell\*(C'\fR on a file accessed in "text" mode.
162 Stick to \f(CW\*(C`seek\*(C'\fR\-ing to
163 locations you got from \f(CW\*(C`tell\*(C'\fR (and no
164 others), and you are usually free to use
165 \&\f(CW\*(C`seek\*(C'\fR and
166 \&\f(CW\*(C`tell\*(C'\fR even in "text" mode. Using
167 \&\f(CW\*(C`seek\*(C'\fR or
168 \&\f(CW\*(C`tell\*(C'\fR or other file operations may be
169 non-portable. If you use \f(CW\*(C`binmode\*(C'\fR on a
170 file, however, you can usually
171 \&\f(CW\*(C`seek\*(C'\fR and
172 \&\f(CW\*(C`tell\*(C'\fR with arbitrary values safely.
173 .PP
174 A common misconception in socket programming is that \f(CW\*(C`\en\ eq\ \e012\*(C'\fR
175 everywhere. When using protocols such as common Internet protocols,
176 \&\f(CW\*(C`\e012\*(C'\fR and \f(CW\*(C`\e015\*(C'\fR are called for specifically, and the values of
177 the logical \f(CW\*(C`\en\*(C'\fR and \f(CW\*(C`\er\*(C'\fR (carriage return) are not reliable.
178 .PP
179 .Vb 2
180 \& print $socket "Hi there, client!\er\en"; # WRONG
181 \& print $socket "Hi there, client!\e015\e012"; # RIGHT
182 .Ve
183 .PP
184 However, using \f(CW\*(C`\e015\e012\*(C'\fR (or \f(CW\*(C`\ecM\ecJ\*(C'\fR, or \f(CW\*(C`\ex0D\ex0A\*(C'\fR) can be tedious
185 and unsightly, as well as confusing to those maintaining the code. As
186 such, the \f(CW\*(C`Socket\*(C'\fR module supplies the Right Thing for those
187 who want it.
188 .PP
189 .Vb 2
190 \& use Socket qw(:DEFAULT :crlf);
191 \& print $socket "Hi there, client!$CRLF" # RIGHT
192 .Ve
193 .PP
194 When reading from a socket, remember that the default input record
195 separator \f(CW$/\fR is \f(CW\*(C`\en\*(C'\fR, but robust socket code
196 will recognize as either \f(CW\*(C`\e012\*(C'\fR or \f(CW\*(C`\e015\e012\*(C'\fR as end of line:
197 .PP
198 .Vb 3
199 \& while (<$socket>) { # NOT ADVISABLE!
200 \& # ...
201 \& }
202 .Ve
203 .PP
204 Because both CRLF and LF end in LF, the input record separator can
205 be set to LF and any CR stripped later. Better to write:
206 .PP
207 .Vb 2
208 \& use Socket qw(:DEFAULT :crlf);
209 \& local($/) = LF; # not needed if $/ is already \e012
210 \&
211 \& while (<$socket>) {
212 \& s/$CR?$LF/\en/; # not sure if socket uses LF or CRLF, OK
213 \& # s/\e015?\e012/\en/; # same thing
214 \& }
215 .Ve
216 .PP
217 This example is preferred over the previous one\-\-even for Unix
218 platforms\-\-because now any \f(CW\*(C`\e015\*(C'\fR's (\f(CW\*(C`\ecM\*(C'\fR's) are stripped out
219 (and there was much rejoicing).
220 .PP
221 Similarly, functions that return text data\-\-such as a function that
222 fetches a web page\-\-should sometimes translate newlines before
223 returning the data, if they've not yet been translated to the local
224 newline representation. A single line of code will often suffice:
225 .PP
226 .Vb 2
227 \& $data =~ s/\e015?\e012/\en/g;
228 \& return $data;
229 .Ve
230 .PP
231 Some of this may be confusing. Here's a handy reference to the ASCII CR
232 and LF characters. You can print it out and stick it in your wallet.
233 .PP
234 .Vb 2
235 \& LF eq \e012 eq \ex0A eq \ecJ eq chr(10) eq ASCII 10
236 \& CR eq \e015 eq \ex0D eq \ecM eq chr(13) eq ASCII 13
237 \&
238 \& | Unix | DOS | Mac |
239 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
240 \& \en | LF | LF | CR |
241 \& \er | CR | CR | LF |
242 \& \en * | LF | CRLF | CR |
243 \& \er * | CR | CR | LF |
244 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
245 \& * text\-mode STDIO
246 .Ve
247 .PP
248 The Unix column assumes that you are not accessing a serial line
249 (like a tty) in canonical mode. If you are, then CR on input becomes
250 "\en", and "\en" on output becomes CRLF.
251 .PP
252 These are just the most common definitions of \f(CW\*(C`\en\*(C'\fR and \f(CW\*(C`\er\*(C'\fR in Perl.
253 There may well be others. For example, on an EBCDIC implementation
254 such as z/OS (OS/390) or OS/400 (using the ILE, the PASE is ASCII-based)
255 the above material is similar to "Unix" but the code numbers change:
256 .PP
257 .Vb 4
258 \& LF eq \e025 eq \ex15 eq \ecU eq chr(21) eq CP\-1047 21
259 \& LF eq \e045 eq \ex25 eq chr(37) eq CP\-0037 37
260 \& CR eq \e015 eq \ex0D eq \ecM eq chr(13) eq CP\-1047 13
261 \& CR eq \e015 eq \ex0D eq \ecM eq chr(13) eq CP\-0037 13
262 \&
263 \& | z/OS | OS/400 |
264 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
265 \& \en | LF | LF |
266 \& \er | CR | CR |
267 \& \en * | LF | LF |
268 \& \er * | CR | CR |
269 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
270 \& * text\-mode STDIO
271 .Ve
272 .SS "Numbers endianness and Width"
273 .IX Subsection "Numbers endianness and Width"
274 Different CPUs store integers and floating point numbers in different
275 orders (called \fIendianness\fR) and widths (32\-bit and 64\-bit being the
276 most common today). This affects your programs when they attempt to transfer
277 numbers in binary format from one CPU architecture to another,
278 usually either "live" via network connection, or by storing the
279 numbers to secondary storage such as a disk file or tape.
280 .PP
281 Conflicting storage orders make an utter mess out of the numbers. If a
282 little-endian host (Intel, VAX) stores 0x12345678 (305419896 in
283 decimal), a big-endian host (Motorola, Sparc, PA) reads it as
284 0x78563412 (2018915346 in decimal). Alpha and MIPS can be either:
285 Digital/Compaq used/uses them in little-endian mode; SGI/Cray uses
286 them in big-endian mode. To avoid this problem in network (socket)
287 connections use the \f(CW\*(C`pack\*(C'\fR and
288 \&\f(CW\*(C`unpack\*(C'\fR formats \f(CW\*(C`n\*(C'\fR and \f(CW\*(C`N\*(C'\fR, the
289 "network" orders. These are guaranteed to be portable.
290 .PP
291 As of Perl 5.10.0, you can also use the \f(CW\*(C`>\*(C'\fR and \f(CW\*(C`<\*(C'\fR modifiers
292 to force big\- or little-endian byte-order. This is useful if you want
293 to store signed integers or 64\-bit integers, for example.
294 .PP
295 You can explore the endianness of your platform by unpacking a
296 data structure packed in native format such as:
297 .PP
298 .Vb 3
299 \& print unpack("h*", pack("s2", 1, 2)), "\en";
300 \& # \*(Aq10002000\*(Aq on e.g. Intel x86 or Alpha 21064 in little\-endian mode
301 \& # \*(Aq00100020\*(Aq on e.g. Motorola 68040
302 .Ve
303 .PP
304 If you need to distinguish between endian architectures you could use
305 either of the variables set like so:
306 .PP
307 .Vb 2
308 \& $is_big_endian = unpack("h*", pack("s", 1)) =~ /01/;
309 \& $is_little_endian = unpack("h*", pack("s", 1)) =~ /^1/;
310 .Ve
311 .PP
312 Differing widths can cause truncation even between platforms of equal
313 endianness. The platform of shorter width loses the upper parts of the
314 number. There is no good solution for this problem except to avoid
315 transferring or storing raw binary numbers.
316 .PP
317 One can circumnavigate both these problems in two ways. Either
318 transfer and store numbers always in text format, instead of raw
319 binary, or else consider using modules like
320 \&\f(CW\*(C`Data::Dumper\*(C'\fR and \f(CW\*(C`Storable\*(C'\fR (included as
321 of Perl 5.8). Keeping all data as text significantly simplifies matters.
322 .SS "Files and Filesystems"
323 .IX Subsection "Files and Filesystems"
324 Most platforms these days structure files in a hierarchical fashion.
325 So, it is reasonably safe to assume that all platforms support the
326 notion of a "path" to uniquely identify a file on the system. How
327 that path is really written, though, differs considerably.
328 .PP
329 Although similar, file path specifications differ between Unix,
330 Windows, Mac\ OS, OS/2, VMS, VOS, RISC\ OS, and probably others.
331 Unix, for example, is one of the few OSes that has the elegant idea
332 of a single root directory.
333 .PP
334 DOS, OS/2, VMS, VOS, and Windows can work similarly to Unix with \f(CW\*(C`/\*(C'\fR
335 as path separator, or in their own idiosyncratic ways (such as having
336 several root directories and various "unrooted" device files such NIL:
337 and LPT:).
338 .PP
339 Mac\ OS 9 and earlier used \f(CW\*(C`:\*(C'\fR as a path separator instead of \f(CW\*(C`/\*(C'\fR.
340 .PP
341 The filesystem may support neither hard links
342 (\f(CW\*(C`link\*(C'\fR) nor symbolic links
343 (\f(CW\*(C`symlink\*(C'\fR,
344 \&\f(CW\*(C`readlink\*(C'\fR,
345 \&\f(CW\*(C`lstat\*(C'\fR).
346 .PP
347 The filesystem may support neither access timestamp nor change
348 timestamp (meaning that about the only portable timestamp is the
349 modification timestamp), or one second granularity of any timestamps
350 (e.g. the FAT filesystem limits the time granularity to two seconds).
351 .PP
352 The "inode change timestamp" (the \f(CW\*(C`\-C\*(C'\fR
353 filetest) may really be the "creation timestamp" (which it is not in
354 Unix).
355 .PP
356 VOS perl can emulate Unix filenames with \f(CW\*(C`/\*(C'\fR as path separator. The
357 native pathname characters greater-than, less-than, number-sign, and
358 percent-sign are always accepted.
359 .PP
360 RISC\ OS perl can emulate Unix filenames with \f(CW\*(C`/\*(C'\fR as path
361 separator, or go native and use \f(CW\*(C`.\*(C'\fR for path separator and \f(CW\*(C`:\*(C'\fR to
362 signal filesystems and disk names.
363 .PP
364 Don't assume Unix filesystem access semantics: that read, write,
365 and execute are all the permissions there are, and even if they exist,
366 that their semantics (for example what do \f(CW\*(C`r\*(C'\fR, \f(CW\*(C`w\*(C'\fR, and \f(CW\*(C`x\*(C'\fR mean on
367 a directory) are the Unix ones. The various Unix/POSIX compatibility
368 layers usually try to make interfaces like \f(CW\*(C`chmod\*(C'\fR
369 work, but sometimes there simply is no good mapping.
370 .PP
371 The \f(CW\*(C`File::Spec\*(C'\fR modules provide methods to manipulate path
372 specifications and return the results in native format for each
373 platform. This is often unnecessary as Unix-style paths are
374 understood by Perl on every supported platform, but if you need to
375 produce native paths for a native utility that does not understand
376 Unix syntax, or if you are operating on paths or path components
377 in unknown (and thus possibly native) syntax, \f(CW\*(C`File::Spec\*(C'\fR
378 is your friend. Here are two brief examples:
379 .PP
380 .Vb 2
381 \& use File::Spec::Functions;
382 \& chdir(updir()); # go up one directory
383 \&
384 \& # Concatenate a path from its components
385 \& my $file = catfile(updir(), \*(Aqtemp\*(Aq, \*(Aqfile.txt\*(Aq);
386 \& # on Unix: \*(Aq../temp/file.txt\*(Aq
387 \& # on Win32: \*(Aq..\etemp\efile.txt\*(Aq
388 \& # on VMS: \*(Aq[\-.temp]file.txt\*(Aq
389 .Ve
390 .PP
391 In general, production code should not have file paths hardcoded.
392 Making them user-supplied or read from a configuration file is
393 better, keeping in mind that file path syntax varies on different
394 machines.
395 .PP
396 This is especially noticeable in scripts like Makefiles and test suites,
397 which often assume \f(CW\*(C`/\*(C'\fR as a path separator for subdirectories.
398 .PP
399 Also of use is \f(CW\*(C`File::Basename\*(C'\fR from the standard
400 distribution, which splits a pathname into pieces (base filename, full
401 path to directory, and file suffix).
402 .PP
403 Even when on a single platform (if you can call Unix a single platform),
404 remember not to count on the existence or the contents of particular
405 system-specific files or directories, like \fI/etc/passwd\fR,
406 \&\fI/etc/sendmail.conf\fR, \fI/etc/resolv.conf\fR, or even \fI/tmp/\fR. For
407 example, \fI/etc/passwd\fR may exist but not contain the encrypted
408 passwords, because the system is using some form of enhanced security.
409 Or it may not contain all the accounts, because the system is using NIS.
410 If code does need to rely on such a file, include a description of the
411 file and its format in the code's documentation, then make it easy for
412 the user to override the default location of the file.
413 .PP
414 Don't assume a text file will end with a newline. They should,
415 but people forget.
416 .PP
417 Do not have two files or directories of the same name with different
418 case, like \fItest.pl\fR and \fITest.pl\fR, as many platforms have
419 case-insensitive (or at least case-forgiving) filenames. Also, try
420 not to have non-word characters (except for \f(CW\*(C`.\*(C'\fR) in the names, and
421 keep them to the 8.3 convention, for maximum portability, onerous a
422 burden though this may appear.
423 .PP
424 Likewise, when using the \f(CW\*(C`AutoSplit\*(C'\fR module, try to keep
425 your functions to 8.3 naming and case-insensitive conventions; or, at the
426 least, make it so the resulting files have a unique (case-insensitively)
427 first 8 characters.
428 .PP
429 Whitespace in filenames is tolerated on most systems, but not all,
430 and even on systems where it might be tolerated, some utilities
431 might become confused by such whitespace.
432 .PP
433 Many systems (DOS, VMS ODS\-2) cannot have more than one \f(CW\*(C`.\*(C'\fR in their
434 filenames.
435 .PP
436 Don't assume \f(CW\*(C`>\*(C'\fR won't be the first character of a filename.
437 Always use the three-arg version of
438 \&\f(CW\*(C`open\*(C'\fR:
439 .PP
440 .Vb 1
441 \& open my $fh, \*(Aq<\*(Aq, $existing_file) or die $!;
442 .Ve
443 .PP
444 Two-arg \f(CW\*(C`open\*(C'\fR is magic and can
445 translate characters like \f(CW\*(C`>\*(C'\fR, \f(CW\*(C`<\*(C'\fR, and \f(CW\*(C`|\*(C'\fR in filenames,
446 which is usually the wrong thing to do.
447 \&\f(CW\*(C`sysopen\*(C'\fR and three-arg
448 \&\f(CW\*(C`open\*(C'\fR don't have this problem.
449 .PP
450 Don't use \f(CW\*(C`:\*(C'\fR as a part of a filename since many systems use that for
451 their own semantics (Mac OS Classic for separating pathname components,
452 many networking schemes and utilities for separating the nodename and
453 the pathname, and so on). For the same reasons, avoid \f(CW\*(C`@\*(C'\fR, \f(CW\*(C`;\*(C'\fR and
454 \&\f(CW\*(C`|\*(C'\fR.
455 .PP
456 Don't assume that in pathnames you can collapse two leading slashes
457 \&\f(CW\*(C`//\*(C'\fR into one: some networking and clustering filesystems have special
458 semantics for that. Let the operating system sort it out.
459 .PP
460 The \fIportable filename characters\fR as defined by ANSI C are
461 .PP
462 .Vb 4
463 \& 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
464 \& 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
465 \& 0 1 2 3 4 5 6 7 8 9
466 \& . _ \-
467 .Ve
468 .PP
469 and \f(CW\*(C`\-\*(C'\fR shouldn't be the first character. If you want to be
470 hypercorrect, stay case-insensitive and within the 8.3 naming
471 convention (all the files and directories have to be unique within one
472 directory if their names are lowercased and truncated to eight
473 characters before the \f(CW\*(C`.\*(C'\fR, if any, and to three characters after the
474 \&\f(CW\*(C`.\*(C'\fR, if any). (And do not use \f(CW\*(C`.\*(C'\fRs in directory names.)
475 .PP
476 On Windows extra \f(CW\*(C`.\*(C'\fRs at the end of a file or directory name are
477 ignored in most circumstances, and a directory name containing only
478 three or more \f(CW\*(C`.\*(C'\fRs are treated as the current directory by some APIs.
479 .SS "System Interaction"
480 .IX Subsection "System Interaction"
481 Not all platforms provide a command line. These are usually platforms
482 that rely primarily on a Graphical User Interface (GUI) for user
483 interaction. A program requiring a command line interface might
484 not work everywhere. This is probably for the user of the program
485 to deal with, so don't stay up late worrying about it.
486 .PP
487 Some platforms can't delete or rename files held open by the system,
488 this limitation may also apply to changing filesystem metainformation
489 like file permissions or owners. Remember to
490 \&\f(CW\*(C`close\*(C'\fR files when you are done with them.
491 Don't \f(CW\*(C`unlink\*(C'\fR or
492 \&\f(CW\*(C`rename\*(C'\fR an open file. Don't
493 \&\f(CW\*(C`tie\*(C'\fR or
494 \&\f(CW\*(C`open\*(C'\fR a file already tied or opened;
495 \&\f(CW\*(C`untie\*(C'\fR or
496 \&\f(CW\*(C`close\*(C'\fR it first.
497 .PP
498 Don't open the same file more than once at a time for writing, as some
499 operating systems put mandatory locks on such files.
500 .PP
501 Don't assume that write/modify permission on a directory gives the
502 right to add or delete files/directories in that directory. That is
503 filesystem specific: in some filesystems you need write/modify
504 permission also (or even just) in the file/directory itself. In some
505 filesystems (AFS, DFS) the permission to add/delete directory entries
506 is a completely separate permission.
507 .PP
508 Don't assume that a single \f(CW\*(C`unlink\*(C'\fR completely
509 gets rid of the file: some filesystems (most notably the ones in VMS) have
510 versioned filesystems, and \f(CW\*(C`unlink\*(C'\fR removes only
511 the most recent one (it doesn't remove all the versions because by default
512 the native tools on those platforms remove just the most recent version,
513 too). The portable idiom to remove all the versions of a file is
514 .PP
515 .Vb 1
516 \& 1 while unlink "file";
517 .Ve
518 .PP
519 This will terminate if the file is undeletable for some reason
520 (protected, not there, and so on).
521 .PP
522 Don't count on a specific environment variable existing in
523 \&\f(CW%ENV\fR. Don't count on \f(CW%ENV\fR entries
524 being case-sensitive, or even case-preserving. Don't try to clear
525 \&\f(CW%ENV\fR by saying \f(CW\*(C`%ENV = ();\*(C'\fR, or, if you really have
526 to, make it conditional on \f(CW\*(C`$^O ne \*(AqVMS\*(Aq\*(C'\fR since in VMS the
527 \&\f(CW%ENV\fR table is much more than a per-process key-value
528 string table.
529 .PP
530 On VMS, some entries in the \f(CW%ENV\fR hash are dynamically
531 created when their key is used on a read if they did not previously
532 exist. The values for \f(CW$ENV{HOME}\fR, \f(CW$ENV{TERM}\fR, \f(CW$ENV{PATH}\fR, and
533 \&\f(CW$ENV{USER}\fR, are known to be dynamically generated. The specific names
534 that are dynamically generated may vary with the version of the C library
535 on VMS, and more may exist than are documented.
536 .PP
537 On VMS by default, changes to the \f(CW%ENV\fR hash persist
538 after perl exits. Subsequent invocations of perl in the same process can
539 inadvertently inherit environment settings that were meant to be
540 temporary.
541 .PP
542 Don't count on signals or \f(CW%SIG\fR for anything.
543 .PP
544 Don't count on filename globbing. Use
545 \&\f(CW\*(C`opendir\*(C'\fR,
546 \&\f(CW\*(C`readdir\*(C'\fR, and
547 \&\f(CW\*(C`closedir\*(C'\fR instead.
548 .PP
549 Don't count on per-program environment variables, or per-program current
550 directories.
551 .PP
552 Don't count on specific values of \f(CW$!\fR, neither numeric nor
553 especially the string values. Users may switch their locales causing
554 error messages to be translated into their languages. If you can
555 trust a POSIXish environment, you can portably use the symbols defined
556 by the \f(CW\*(C`Errno\*(C'\fR module, like \f(CW\*(C`ENOENT\*(C'\fR. And don't trust on the
557 values of \f(CW$!\fR at all except immediately after a failed
558 system call.
559 .SS "Command names versus file pathnames"
560 .IX Subsection "Command names versus file pathnames"
561 Don't assume that the name used to invoke a command or program with
562 \&\f(CW\*(C`system\*(C'\fR or \f(CW\*(C`exec\*(C'\fR can
563 also be used to test for the existence of the file that holds the
564 executable code for that command or program.
565 First, many systems have "internal" commands that are built-in to the
566 shell or OS and while these commands can be invoked, there is no
567 corresponding file. Second, some operating systems (e.g., Cygwin,
568 OS/2, and VOS) have required suffixes for executable files;
569 these suffixes are generally permitted on the command name but are not
570 required. Thus, a command like \f(CW\*(C`perl\*(C'\fR might exist in a file named
571 \&\fIperl\fR, \fIperl.exe\fR, or \fIperl.pm\fR, depending on the operating system.
572 The variable \f(CW$Config{_exe}\fR in the
573 \&\f(CW\*(C`Config\*(C'\fR module holds the executable suffix, if any. Third,
574 the VMS port carefully sets up \f(CW$^X\fR and
575 \&\f(CW$Config{perlpath}\fR so that no further processing
576 is required. This is just as well, because the matching regular
577 expression used below would then have to deal with a possible trailing
578 version number in the VMS file name.
579 .PP
580 To convert \f(CW$^X\fR to a file pathname, taking account of
581 the requirements of the various operating system possibilities, say:
582 .PP
583 .Vb 6
584 \& use Config;
585 \& my $thisperl = $^X;
586 \& if ($^O ne \*(AqVMS\*(Aq) {
587 \& $thisperl .= $Config{_exe}
588 \& unless $thisperl =~ m/\eQ$Config{_exe}\eE$/i;
589 \& }
590 .Ve
591 .PP
592 To convert \f(CW$Config{perlpath}\fR to a file pathname, say:
593 .PP
594 .Vb 6
595 \& use Config;
596 \& my $thisperl = $Config{perlpath};
597 \& if ($^O ne \*(AqVMS\*(Aq) {
598 \& $thisperl .= $Config{_exe}
599 \& unless $thisperl =~ m/\eQ$Config{_exe}\eE$/i;
600 \& }
601 .Ve
602 .SS Networking
603 .IX Subsection "Networking"
604 Don't assume that you can reach the public Internet.
605 .PP
606 Don't assume that there is only one way to get through firewalls
607 to the public Internet.
608 .PP
609 Don't assume that you can reach outside world through any other port
610 than 80, or some web proxy. ftp is blocked by many firewalls.
611 .PP
612 Don't assume that you can send email by connecting to the local SMTP port.
613 .PP
614 Don't assume that you can reach yourself or any node by the name
615 \&'localhost'. The same goes for '127.0.0.1'. You will have to try both.
616 .PP
617 Don't assume that the host has only one network card, or that it
618 can't bind to many virtual IP addresses.
619 .PP
620 Don't assume a particular network device name.
621 .PP
622 Don't assume a particular set of
623 \&\f(CW\*(C`ioctl\*(C'\fRs will work.
624 .PP
625 Don't assume that you can ping hosts and get replies.
626 .PP
627 Don't assume that any particular port (service) will respond.
628 .PP
629 Don't assume that \f(CW\*(C`Sys::Hostname\*(C'\fR (or any other API or
630 command) returns either a fully qualified hostname or a non-qualified
631 hostname: it all depends on how the system had been configured. Also
632 remember that for things such as DHCP and NAT, the hostname you get back
633 might not be very useful.
634 .PP
635 All the above \fIdon't\fRs may look daunting, and they are, but the key
636 is to degrade gracefully if one cannot reach the particular network
637 service one wants. Croaking or hanging do not look very professional.
638 .SS "Interprocess Communication (IPC)"
639 .IX Subsection "Interprocess Communication (IPC)"
640 In general, don't directly access the system in code meant to be
641 portable. That means, no \f(CW\*(C`system\*(C'\fR,
642 \&\f(CW\*(C`exec\*(C'\fR, \f(CW\*(C`fork\*(C'\fR,
643 \&\f(CW\*(C`pipe\*(C'\fR,
644 \&\f(CW\`\`\fR or \f(CW\*(C`qx//\*(C'\fR,
645 \&\f(CW\*(C`open\*(C'\fR with a \f(CW\*(C`|\*(C'\fR, nor any of the other
646 things that makes being a Perl hacker worth being.
647 .PP
648 Commands that launch external processes are generally supported on
649 most platforms (though many of them do not support any type of
650 forking). The problem with using them arises from what you invoke
651 them on. External tools are often named differently on different
652 platforms, may not be available in the same location, might accept
653 different arguments, can behave differently, and often present their
654 results in a platform-dependent way. Thus, you should seldom depend
655 on them to produce consistent results. (Then again, if you're calling
656 \&\f(CW\*(C`netstat \-a\*(C'\fR, you probably don't expect it to run on both Unix and CP/M.)
657 .PP
658 One especially common bit of Perl code is opening a pipe to \fBsendmail\fR:
659 .PP
660 .Vb 2
661 \& open(my $mail, \*(Aq|\-\*(Aq, \*(Aq/usr/lib/sendmail \-t\*(Aq)
662 \& or die "cannot fork sendmail: $!";
663 .Ve
664 .PP
665 This is fine for systems programming when sendmail is known to be
666 available. But it is not fine for many non-Unix systems, and even
667 some Unix systems that may not have sendmail installed. If a portable
668 solution is needed, see the various distributions on CPAN that deal
669 with it. \f(CW\*(C`Mail::Mailer\*(C'\fR and \f(CW\*(C`Mail::Send\*(C'\fR
670 in the \f(CW\*(C`MailTools\*(C'\fR distribution are commonly used, and provide several
671 mailing methods, including \f(CW\*(C`mail\*(C'\fR, \f(CW\*(C`sendmail\*(C'\fR, and direct SMTP (via
672 \&\f(CW\*(C`Net::SMTP\*(C'\fR) if a mail transfer agent is not available.
673 \&\f(CW\*(C`Mail::Sendmail\*(C'\fR is a standalone module that provides
674 simple, platform-independent mailing.
675 .PP
676 The Unix System V IPC (\f(CW\*(C`msg*(), sem*(), shm*()\*(C'\fR) is not available
677 even on all Unix platforms.
678 .PP
679 Do not use either the bare result of \f(CW\*(C`pack("N", 10, 20, 30, 40)\*(C'\fR or
680 bare v\-strings (such as \f(CW\*(C`v10.20.30.40\*(C'\fR) to represent IPv4 addresses:
681 both forms just pack the four bytes into network order. That this
682 would be equal to the C language \f(CW\*(C`in_addr\*(C'\fR struct (which is what the
683 socket code internally uses) is not guaranteed. To be portable use
684 the routines of the \f(CW\*(C`Socket\*(C'\fR module, such as
685 \&\f(CW\*(C`inet_aton\*(C'\fR,
686 \&\f(CW\*(C`inet_ntoa\*(C'\fR, and
687 \&\f(CW\*(C`sockaddr_in\*(C'\fR.
688 .PP
689 The rule of thumb for portable code is: Do it all in portable Perl, or
690 use a module (that may internally implement it with platform-specific
691 code, but exposes a common interface).
692 .SS "External Subroutines (XS)"
693 .IX Subsection "External Subroutines (XS)"
694 XS code can usually be made to work with any platform, but dependent
695 libraries, header files, etc., might not be readily available or
696 portable, or the XS code itself might be platform-specific, just as Perl
697 code might be. If the libraries and headers are portable, then it is
698 normally reasonable to make sure the XS code is portable, too.
699 .PP
700 A different type of portability issue arises when writing XS code:
701 availability of a C compiler on the end-user's system. C brings
702 with it its own portability issues, and writing XS code will expose
703 you to some of those. Writing purely in Perl is an easier way to
704 achieve portability.
705 .SS "Standard Modules"
706 .IX Subsection "Standard Modules"
707 In general, the standard modules work across platforms. Notable
708 exceptions are the \f(CW\*(C`CPAN\*(C'\fR module (which currently makes
709 connections to external programs that may not be available),
710 platform-specific modules (like \f(CW\*(C`ExtUtils::MM_VMS\*(C'\fR),
711 and DBM modules.
712 .PP
713 There is no one DBM module available on all platforms.
714 \&\f(CW\*(C`SDBM_File\*(C'\fR and the others are generally available on all
715 Unix and DOSish ports, but not in MacPerl, where only
716 \&\f(CW\*(C`NDBM_File\*(C'\fR and \f(CW\*(C`DB_File\*(C'\fR are available.
717 .PP
718 The good news is that at least some DBM module should be available, and
719 \&\f(CW\*(C`AnyDBM_File\*(C'\fR will use whichever module it can find. Of
720 course, then the code needs to be fairly strict, dropping to the greatest
721 common factor (e.g., not exceeding 1K for each record), so that it will
722 work with any DBM module. See AnyDBM_File for more details.
723 .SS "Time and Date"
724 .IX Subsection "Time and Date"
725 The system's notion of time of day and calendar date is controlled in
726 widely different ways. Don't assume the timezone is stored in \f(CW$ENV{TZ}\fR,
727 and even if it is, don't assume that you can control the timezone through
728 that variable. Don't assume anything about the three-letter timezone
729 abbreviations (for example that MST would be the Mountain Standard Time,
730 it's been known to stand for Moscow Standard Time). If you need to
731 use timezones, express them in some unambiguous format like the
732 exact number of minutes offset from UTC, or the POSIX timezone
733 format.
734 .PP
735 Don't assume that the epoch starts at 00:00:00, January 1, 1970,
736 because that is OS\- and implementation-specific. It is better to
737 store a date in an unambiguous representation. The ISO 8601 standard
738 defines YYYY-MM-DD as the date format, or YYYY\-MM\-DDTHH:MM:SS
739 (that's a literal "T" separating the date from the time).
740 Please do use the ISO 8601 instead of making us guess what
741 date 02/03/04 might be. ISO 8601 even sorts nicely as-is.
742 A text representation (like "1987\-12\-18") can be easily converted
743 into an OS-specific value using a module like
744 \&\f(CW\*(C`Time::Piece\*(C'\fR (see "Date Parsing" in Time::Piece) or
745 \&\f(CW\*(C`Date::Parse\*(C'\fR. An array of values, such as those
746 returned by \f(CW\*(C`localtime\*(C'\fR, can be converted to an OS-specific
747 representation using \f(CW\*(C`Time::Local\*(C'\fR.
748 .PP
749 When calculating specific times, such as for tests in time or date modules,
750 it may be appropriate to calculate an offset for the epoch.
751 .PP
752 .Vb 2
753 \& use Time::Local qw(timegm);
754 \& my $offset = timegm(0, 0, 0, 1, 0, 1970);
755 .Ve
756 .PP
757 The value for \f(CW$offset\fR in Unix will be \f(CW0\fR, but in Mac OS Classic
758 will be some large number. \f(CW$offset\fR can then be added to a Unix time
759 value to get what should be the proper value on any system.
760 .SS "Character sets and character encoding"
761 .IX Subsection "Character sets and character encoding"
762 Assume very little about character sets.
763 .PP
764 Assume nothing about numerical values (\f(CW\*(C`ord\*(C'\fR,
765 \&\f(CW\*(C`chr\*(C'\fR) of characters.
766 Do not use explicit code point ranges (like \f(CW\*(C`\exHH\-\exHH)\*(C'\fR. However,
767 starting in Perl v5.22, regular expression pattern bracketed character
768 class ranges specified like \f(CW\*(C`qr/[\eN{U+HH}\-\eN{U+HH}]/\*(C'\fR are portable,
769 and starting in Perl v5.24, the same ranges are portable in
770 \&\f(CW\*(C`tr///\*(C'\fR.
771 You can portably use symbolic character classes like \f(CW\*(C`[:print:]\*(C'\fR.
772 .PP
773 Do not assume that the alphabetic characters are encoded contiguously
774 (in the numeric sense). There may be gaps. Special coding in Perl,
775 however, guarantees that all subsets of \f(CW\*(C`qr/[A\-Z]/\*(C'\fR, \f(CW\*(C`qr/[a\-z]/\*(C'\fR, and
776 \&\f(CW\*(C`qr/[0\-9]/\*(C'\fR behave as expected.
777 \&\f(CW\*(C`tr///\*(C'\fR
778 behaves the same for these ranges. In patterns, any ranges specified with
779 end points using the \f(CW\*(C`\eN{...}\*(C'\fR notations ensures character set
780 portability, but it is a bug in Perl v5.22 that this isn't true of
781 \&\f(CW\*(C`tr///\*(C'\fR,
782 fixed in v5.24.
783 .PP
784 Do not assume anything about the ordering of the characters.
785 The lowercase letters may come before or after the uppercase letters;
786 the lowercase and uppercase may be interlaced so that both "a" and "A"
787 come before "b"; the accented and other international characters may
788 be interlaced so that ä comes before "b".
789 Unicode::Collate can be used to sort this all out.
790 .SS Internationalisation
791 .IX Subsection "Internationalisation"
792 If you may assume POSIX (a rather large assumption), you may read
793 more about the POSIX locale system from perllocale. The locale
794 system at least attempts to make things a little bit more portable,
795 or at least more convenient and native-friendly for non-English
796 users. The system affects character sets and encoding, and date
797 and time formatting\-\-amongst other things.
798 .PP
799 If you really want to be international, you should consider Unicode.
800 See perluniintro and perlunicode for more information.
801 .PP
802 By default Perl assumes your source code is written in an 8\-bit ASCII
803 superset. To embed Unicode characters in your strings and regexes, you can
804 use the \f(CW\*(C`\ex{HH}\*(C'\fR or (more portably) \f(CW\*(C`\eN{U+HH}\*(C'\fR
805 notations. You can also use the
806 \&\f(CW\*(C`utf8\*(C'\fR pragma and write your code in UTF\-8, which lets you use
807 Unicode characters directly (not just in quoted constructs but also in
808 identifiers).
809 .SS "System Resources"
810 .IX Subsection "System Resources"
811 If your code is destined for systems with severely constrained (or
812 missing!) virtual memory systems then you want to be \fIespecially\fR mindful
813 of avoiding wasteful constructs such as:
814 .PP
815 .Vb 1
816 \& my @lines = <$very_large_file>; # bad
817 \&
818 \& while (<$fh>) {$file .= $_} # sometimes bad
819 \& my $file = join(\*(Aq\*(Aq, <$fh>); # better
820 .Ve
821 .PP
822 The last two constructs may appear unintuitive to most people. The
823 first repeatedly grows a string, whereas the second allocates a
824 large chunk of memory in one go. On some systems, the second is
825 more efficient than the first.
826 .SS Security
827 .IX Subsection "Security"
828 Most multi-user platforms provide basic levels of security, usually
829 implemented at the filesystem level. Some, however, unfortunately do
830 not. Thus the notion of user id, or "home" directory,
831 or even the state of being logged-in, may be unrecognizable on many
832 platforms. If you write programs that are security-conscious, it
833 is usually best to know what type of system you will be running
834 under so that you can write code explicitly for that platform (or
835 class of platforms).
836 .PP
837 Don't assume the Unix filesystem access semantics: the operating
838 system or the filesystem may be using some ACL systems, which are
839 richer languages than the usual \f(CW\*(C`rwx\*(C'\fR. Even if the \f(CW\*(C`rwx\*(C'\fR exist,
840 their semantics might be different.
841 .PP
842 (From the security viewpoint, testing for permissions before attempting to
843 do something is silly anyway: if one tries this, there is potential
844 for race conditions. Someone or something might change the
845 permissions between the permissions check and the actual operation.
846 Just try the operation.)
847 .PP
848 Don't assume the Unix user and group semantics: especially, don't
849 expect \f(CW$<\fR and \f(CW$>\fR (or
850 \&\f(CW$(\fR and \f(CW$)\fR) to work for switching
851 identities (or memberships).
852 .PP
853 Don't assume set-uid and set-gid semantics. (And even if you do,
854 think twice: set-uid and set-gid are a known can of security worms.)
855 .SS Style
856 .IX Subsection "Style"
857 For those times when it is necessary to have platform-specific code,
858 consider keeping the platform-specific code in one place, making porting
859 to other platforms easier. Use the \f(CW\*(C`Config\*(C'\fR module and the
860 special variable \f(CW$^O\fR to differentiate platforms, as
861 described in "PLATFORMS".
862 .PP
863 Beware of the "else syndrome":
864 .PP
865 .Vb 5
866 \& if ($^O eq \*(AqMSWin32\*(Aq) {
867 \& # code that assumes Windows
868 \& } else {
869 \& # code that assumes Linux
870 \& }
871 .Ve
872 .PP
873 The \f(CW\*(C`else\*(C'\fR branch should be used for the really ultimate fallback,
874 not for code specific to some platform.
875 .PP
876 Be careful in the tests you supply with your module or programs.
877 Module code may be fully portable, but its tests might not be. This
878 often happens when tests spawn off other processes or call external
879 programs to aid in the testing, or when (as noted above) the tests
880 assume certain things about the filesystem and paths. Be careful not
881 to depend on a specific output style for errors, such as when checking
882 \&\f(CW$!\fR after a failed system call. Using
883 \&\f(CW$!\fR for anything else than displaying it as output is
884 doubtful (though see the \f(CW\*(C`Errno\*(C'\fR module for testing reasonably
885 portably for error value). Some platforms expect a certain output format,
886 and Perl on those platforms may have been adjusted accordingly. Most
887 specifically, don't anchor a regex when testing an error value.
888 .SH "CPAN Testers"
889 .IX Header "CPAN Testers"
890 Modules uploaded to CPAN are tested by a variety of volunteers on
891 different platforms. These CPAN testers are notified by mail of each
892 new upload, and reply to the list with PASS, FAIL, NA (not applicable to
893 this platform), or UNKNOWN (unknown), along with any relevant notations.
894 .PP
895 The purpose of the testing is twofold: one, to help developers fix any
896 problems in their code that crop up because of lack of testing on other
897 platforms; two, to provide users with information about whether
898 a given module works on a given platform.
899 .PP
900 Also see:
901 .IP \(bu 4
902 Mailing list: cpan\-testers\[email protected]
903 .IP \(bu 4
904 Testing results: <https://www.cpantesters.org/>
905 .SH PLATFORMS
906 .IX Header "PLATFORMS"
907 Perl is built with a \f(CW$^O\fR variable that indicates the
908 operating system it was built on. This was implemented
909 to help speed up code that would otherwise have to \f(CW\*(C`use Config\*(C'\fR
910 and use the value of \f(CW$Config{osname}\fR. Of course,
911 to get more detailed information about the system, looking into
912 \&\f(CW%Config\fR is certainly recommended.
913 .PP
914 \&\f(CW%Config\fR cannot always be trusted, however,
915 because it was built at compile time. If perl was built in one place,
916 then transferred elsewhere, some values may be wrong. The values may
917 even have been edited after the fact.
918 .SS Unix
919 .IX Subsection "Unix"
920 Perl works on a bewildering variety of Unix and Unix-like platforms (see
921 e.g. most of the files in the \fIhints/\fR directory in the source code kit).
922 On most of these systems, the value of \f(CW$^O\fR (hence
923 \&\f(CW$Config{osname}\fR, too) is determined either by
924 lowercasing and stripping punctuation from the first field of the string
925 returned by typing \f(CW\*(C`uname \-a\*(C'\fR (or a similar command) at the shell prompt
926 or by testing the file system for the presence of uniquely named files
927 such as a kernel or header file. Here, for example, are a few of the
928 more popular Unix flavors:
929 .PP
930 .Vb 10
931 \& uname $^O $Config{archname}
932 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
933 \& AIX aix aix
934 \& BSD/OS bsdos i386\-bsdos
935 \& Darwin darwin darwin
936 \& DYNIX/ptx dynixptx i386\-dynixptx
937 \& FreeBSD freebsd freebsd\-i386
938 \& Haiku haiku BePC\-haiku
939 \& Linux linux arm\-linux
940 \& Linux linux armv5tel\-linux
941 \& Linux linux i386\-linux
942 \& Linux linux i586\-linux
943 \& Linux linux ppc\-linux
944 \& HP\-UX hpux PA\-RISC1.1
945 \& IRIX irix irix
946 \& Mac OS X darwin darwin
947 \& NeXT 3 next next\-fat
948 \& NeXT 4 next OPENSTEP\-Mach
949 \& openbsd openbsd i386\-openbsd
950 \& OSF1 dec_osf alpha\-dec_osf
951 \& reliantunix\-n svr4 RM400\-svr4
952 \& SCO_SV sco_sv i386\-sco_sv
953 \& SINIX\-N svr4 RM400\-svr4
954 \& sn4609 unicos CRAY_C90\-unicos
955 \& sn6521 unicosmk t3e\-unicosmk
956 \& sn9617 unicos CRAY_J90\-unicos
957 \& SunOS solaris sun4\-solaris
958 \& SunOS solaris i86pc\-solaris
959 \& SunOS4 sunos sun4\-sunos
960 .Ve
961 .PP
962 Because the value of \f(CW$Config{archname}\fR may
963 depend on the hardware architecture, it can vary more than the value of
964 \&\f(CW$^O\fR.
965 .SS "DOS and Derivatives"
966 .IX Subsection "DOS and Derivatives"
967 Perl has long been ported to Intel-style microcomputers running under
968 systems like PC-DOS, MS-DOS, OS/2, and most Windows platforms you can
969 bring yourself to mention (except for Windows CE, if you count that).
970 Users familiar with \fICOMMAND.COM\fR or \fICMD.EXE\fR style shells should
971 be aware that each of these file specifications may have subtle
972 differences:
973 .PP
974 .Vb 4
975 \& my $filespec0 = "c:/foo/bar/file.txt";
976 \& my $filespec1 = "c:\e\efoo\e\ebar\e\efile.txt";
977 \& my $filespec2 = \*(Aqc:\efoo\ebar\efile.txt\*(Aq;
978 \& my $filespec3 = \*(Aqc:\e\efoo\e\ebar\e\efile.txt\*(Aq;
979 .Ve
980 .PP
981 System calls accept either \f(CW\*(C`/\*(C'\fR or \f(CW\*(C`\e\*(C'\fR as the path separator.
982 However, many command-line utilities of DOS vintage treat \f(CW\*(C`/\*(C'\fR as
983 the option prefix, so may get confused by filenames containing \f(CW\*(C`/\*(C'\fR.
984 Aside from calling any external programs, \f(CW\*(C`/\*(C'\fR will work just fine,
985 and probably better, as it is more consistent with popular usage,
986 and avoids the problem of remembering what to backwhack and what
987 not to.
988 .PP
989 The DOS FAT filesystem can accommodate only "8.3" style filenames. Under
990 the "case-insensitive, but case-preserving" HPFS (OS/2) and NTFS (NT)
991 filesystems you may have to be careful about case returned with functions
992 like \f(CW\*(C`readdir\*(C'\fR or used with functions like
993 \&\f(CW\*(C`open\*(C'\fR or
994 \&\f(CW\*(C`opendir\*(C'\fR.
995 .PP
996 DOS also treats several filenames as special, such as \fIAUX\fR, \fIPRN\fR,
997 \&\fINUL\fR, \fICON\fR, \fICOM1\fR, \fILPT1\fR, \fILPT2\fR, etc. Unfortunately, sometimes
998 these filenames won't even work if you include an explicit directory
999 prefix. It is best to avoid such filenames, if you want your code to be
1000 portable to DOS and its derivatives. It's hard to know what these all
1001 are, unfortunately.
1002 .PP
1003 Users of these operating systems may also wish to make use of
1004 scripts such as \fIpl2bat.bat\fR to put wrappers around your scripts.
1005 .PP
1006 Newline (\f(CW\*(C`\en\*(C'\fR) is translated as \f(CW\*(C`\e015\e012\*(C'\fR by the I/O system when
1007 reading from and writing to files (see "Newlines").
1008 \&\f(CWbinmode($filehandle)\fR will keep \f(CW\*(C`\en\*(C'\fR translated as \f(CW\*(C`\e012\*(C'\fR for that
1009 filehandle.
1010 \&\f(CW\*(C`binmode\*(C'\fR should always be used for code
1011 that deals with binary data. That's assuming you realize in advance that
1012 your data is in binary. General-purpose programs should often assume
1013 nothing about their data.
1014 .PP
1015 The \f(CW$^O\fR variable and the
1016 \&\f(CW$Config{archname}\fR values for various DOSish
1017 perls are as follows:
1018 .PP
1019 .Vb 10
1020 \& OS $^O $Config{archname} ID Version
1021 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
1022 \& MS\-DOS dos ?
1023 \& PC\-DOS dos ?
1024 \& OS/2 os2 ?
1025 \& Windows 3.1 ? ? 0 3 01
1026 \& Windows 95 MSWin32 MSWin32\-x86 1 4 00
1027 \& Windows 98 MSWin32 MSWin32\-x86 1 4 10
1028 \& Windows ME MSWin32 MSWin32\-x86 1 ?
1029 \& Windows NT MSWin32 MSWin32\-x86 2 4 xx
1030 \& Windows NT MSWin32 MSWin32\-ALPHA 2 4 xx
1031 \& Windows NT MSWin32 MSWin32\-ppc 2 4 xx
1032 \& Windows 2000 MSWin32 MSWin32\-x86 2 5 00
1033 \& Windows XP MSWin32 MSWin32\-x86 2 5 01
1034 \& Windows 2003 MSWin32 MSWin32\-x86 2 5 02
1035 \& Windows Vista MSWin32 MSWin32\-x86 2 6 00
1036 \& Windows 7 MSWin32 MSWin32\-x86 2 6 01
1037 \& Windows 7 MSWin32 MSWin32\-x64 2 6 01
1038 \& Windows 2008 MSWin32 MSWin32\-x86 2 6 01
1039 \& Windows 2008 MSWin32 MSWin32\-x64 2 6 01
1040 \& Windows CE MSWin32 ? 3
1041 \& Cygwin cygwin cygwin
1042 .Ve
1043 .PP
1044 The various MSWin32 Perl's can distinguish the OS they are running on
1045 via the value of the fifth element of the list returned from
1046 \&\f(CWWin32::GetOSVersion()\fR. For example:
1047 .PP
1048 .Vb 4
1049 \& if ($^O eq \*(AqMSWin32\*(Aq) {
1050 \& my @os_version_info = Win32::GetOSVersion();
1051 \& print +(\*(Aq3.1\*(Aq,\*(Aq95\*(Aq,\*(AqNT\*(Aq)[$os_version_info[4]],"\en";
1052 \& }
1053 .Ve
1054 .PP
1055 There are also \f(CW\*(C`Win32::IsWinNT()|Win32/Win32::IsWinNT()\*(C'\fR,
1056 \&\f(CW\*(C`Win32::IsWin95()|Win32/Win32::IsWin95()\*(C'\fR, and
1057 \&\f(CWWin32::GetOSName()\fR; try
1058 \&\f(CW\*(C`perldoc Win32\*(C'\fR.
1059 The very portable \f(CWPOSIX::uname()\fR will work too:
1060 .PP
1061 .Vb 2
1062 \& c:\e> perl \-MPOSIX \-we "print join \*(Aq|\*(Aq, uname"
1063 \& Windows NT|moonru|5.0|Build 2195 (Service Pack 2)|x86
1064 .Ve
1065 .PP
1066 Errors set by Winsock functions are now put directly into \f(CW$^E\fR,
1067 and the relevant \f(CW\*(C`WSAE*\*(C'\fR error codes are now exported from the
1068 Errno and POSIX modules for testing this against.
1069 .PP
1070 The previous behavior of putting the errors (converted to POSIX-style
1071 \&\f(CW\*(C`E*\*(C'\fR error codes since Perl 5.20.0) into \f(CW$!\fR was buggy due to
1072 the non-equivalence of like-named Winsock and POSIX error constants,
1073 a relationship between which has unfortunately been established
1074 in one way or another since Perl 5.8.0.
1075 .PP
1076 The new behavior provides a much more robust solution for checking
1077 Winsock errors in portable software without accidentally matching
1078 POSIX tests that were intended for other OSes and may have different
1079 meanings for Winsock.
1080 .PP
1081 The old behavior is currently retained, warts and all, for backwards
1082 compatibility, but users are encouraged to change any code that
1083 tests \f(CW$!\fR against \f(CW\*(C`E*\*(C'\fR constants for Winsock errors to instead
1084 test \f(CW$^E\fR against \f(CW\*(C`WSAE*\*(C'\fR constants. After a suitable deprecation
1085 period, which started with Perl 5.24, the old behavior may be
1086 removed, leaving \f(CW$!\fR unchanged after Winsock function calls, to
1087 avoid any possible confusion over which error variable to check.
1088 .PP
1089 Also see:
1090 .IP \(bu 4
1091 The EMX environment for DOS, OS/2, etc. [email protected],
1092 <ftp://hobbes.nmsu.edu/pub/os2/dev/emx/> Also perlos2.
1093 .IP \(bu 4
1094 Build instructions for Win32 in perlwin32, or under the Cygnus environment
1095 in perlcygwin.
1096 .IP \(bu 4
1097 The \f(CW\*(C`Win32::*\*(C'\fR modules in Win32.
1098 .IP \(bu 4
1099 The ActiveState Pages, <https://www.activestate.com/>
1100 .IP \(bu 4
1101 The Cygwin environment for Win32; \fIREADME.cygwin\fR (installed
1102 as perlcygwin), <https://www.cygwin.com/>
1103 .IP \(bu 4
1104 Build instructions for OS/2, perlos2
1105 .SS VMS
1106 .IX Subsection "VMS"
1107 Perl on VMS is discussed in perlvms in the Perl distribution.
1108 .PP
1109 The official name of VMS as of this writing is OpenVMS.
1110 .PP
1111 Interacting with Perl from the Digital Command Language (DCL) shell
1112 often requires a different set of quotation marks than Unix shells do.
1113 For example:
1114 .PP
1115 .Vb 2
1116 \& $ perl \-e "print ""Hello, world.\en"""
1117 \& Hello, world.
1118 .Ve
1119 .PP
1120 There are several ways to wrap your Perl scripts in DCL \fI.COM\fR files, if
1121 you are so inclined. For example:
1122 .PP
1123 .Vb 6
1124 \& $ write sys$output "Hello from DCL!"
1125 \& $ if p1 .eqs. ""
1126 \& $ then perl \-x \*(Aqf$environment("PROCEDURE")
1127 \& $ else perl \-x \- \*(Aqp1 \*(Aqp2 \*(Aqp3 \*(Aqp4 \*(Aqp5 \*(Aqp6 \*(Aqp7 \*(Aqp8
1128 \& $ deck/dollars="_\|_END_\|_"
1129 \& #!/usr/bin/perl
1130 \&
1131 \& print "Hello from Perl!\en";
1132 \&
1133 \& _\|_END_\|_
1134 \& $ endif
1135 .Ve
1136 .PP
1137 Do take care with \f(CW\*(C`$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT\*(C'\fR if your
1138 Perl-in-DCL script expects to do things like \f(CW\*(C`$read = <STDIN>;\*(C'\fR.
1139 .PP
1140 The VMS operating system has two filesystems, designated by their
1141 on-disk structure (ODS) level: ODS\-2 and its successor ODS\-5. The
1142 initial port of Perl to VMS pre-dates ODS\-5, but all current testing and
1143 development assumes ODS\-5 and its capabilities, including case
1144 preservation, extended characters in filespecs, and names up to 8192
1145 bytes long.
1146 .PP
1147 Perl on VMS can accept either VMS\- or Unix-style file
1148 specifications as in either of the following:
1149 .PP
1150 .Vb 2
1151 \& $ perl \-ne "print if /perl_setup/i" SYS$LOGIN:LOGIN.COM
1152 \& $ perl \-ne "print if /perl_setup/i" /sys$login/login.com
1153 .Ve
1154 .PP
1155 but not a mixture of both as in:
1156 .PP
1157 .Vb 2
1158 \& $ perl \-ne "print if /perl_setup/i" sys$login:/login.com
1159 \& Can\*(Aqt open sys$login:/login.com: file specification syntax error
1160 .Ve
1161 .PP
1162 In general, the easiest path to portability is always to specify
1163 filenames in Unix format unless they will need to be processed by native
1164 commands or utilities. Because of this latter consideration, the
1165 File::Spec module by default returns native format specifications
1166 regardless of input format. This default may be reversed so that
1167 filenames are always reported in Unix format by specifying the
1168 \&\f(CW\*(C`DECC$FILENAME_UNIX_REPORT\*(C'\fR feature logical in the environment.
1169 .PP
1170 The file type, or extension, is always present in a VMS-format file
1171 specification even if it's zero-length. This means that, by default,
1172 \&\f(CW\*(C`readdir\*(C'\fR will return a trailing dot on a
1173 file with no extension, so where you would see \f(CW"a"\fR on Unix you'll see
1174 \&\f(CW"a."\fR on VMS. However, the trailing dot may be suppressed by enabling
1175 the \f(CW\*(C`DECC$READDIR_DROPDOTNOTYPE\*(C'\fR feature in the environment (see the CRTL
1176 documentation on feature logical names).
1177 .PP
1178 What \f(CW\*(C`\en\*(C'\fR represents depends on the type of file opened. It usually
1179 represents \f(CW\*(C`\e012\*(C'\fR but it could also be \f(CW\*(C`\e015\*(C'\fR, \f(CW\*(C`\e012\*(C'\fR, \f(CW\*(C`\e015\e012\*(C'\fR,
1180 \&\f(CW\*(C`\e000\*(C'\fR, \f(CW\*(C`\e040\*(C'\fR, or nothing depending on the file organization and
1181 record format. The \f(CW\*(C`VMS::Stdio\*(C'\fR module provides access to
1182 the special \f(CWfopen()\fR requirements of files with unusual attributes on
1183 VMS.
1184 .PP
1185 The value of \f(CW$^O\fR on OpenVMS is "VMS". To determine the
1186 architecture that you are running on refer to
1187 \&\f(CW$Config{archname}\fR.
1188 .PP
1189 On VMS, perl determines the UTC offset from the \f(CW\*(C`SYS$TIMEZONE_DIFFERENTIAL\*(C'\fR
1190 logical name. Although the VMS epoch began at 17\-NOV\-1858 00:00:00.00,
1191 calls to \f(CW\*(C`localtime\*(C'\fR are adjusted to count
1192 offsets from 01\-JAN\-1970 00:00:00.00, just like Unix.
1193 .PP
1194 Also see:
1195 .IP \(bu 4
1196 \&\fIREADME.vms\fR (installed as \fIREADME_vms\fR), perlvms
1197 .IP \(bu 4
1198 vmsperl list, vmsperl\[email protected]
1199 .IP \(bu 4
1200 vmsperl on the web, <http://www.sidhe.org/vmsperl/index.html>
1201 .IP \(bu 4
1202 VMS Software Inc. web site, <http://www.vmssoftware.com>
1203 .SS VOS
1204 .IX Subsection "VOS"
1205 Perl on VOS (also known as OpenVOS) is discussed in \fIREADME.vos\fR
1206 in the Perl distribution (installed as perlvos). Perl on VOS
1207 can accept either VOS\- or Unix-style file specifications as in
1208 either of the following:
1209 .PP
1210 .Vb 2
1211 \& $ perl \-ne "print if /perl_setup/i" >system>notices
1212 \& $ perl \-ne "print if /perl_setup/i" /system/notices
1213 .Ve
1214 .PP
1215 or even a mixture of both as in:
1216 .PP
1217 .Vb 1
1218 \& $ perl \-ne "print if /perl_setup/i" >system/notices
1219 .Ve
1220 .PP
1221 Even though VOS allows the slash character to appear in object
1222 names, because the VOS port of Perl interprets it as a pathname
1223 delimiting character, VOS files, directories, or links whose
1224 names contain a slash character cannot be processed. Such files
1225 must be renamed before they can be processed by Perl.
1226 .PP
1227 Older releases of VOS (prior to OpenVOS Release 17.0) limit file
1228 names to 32 or fewer characters, prohibit file names from
1229 starting with a \f(CW\*(C`\-\*(C'\fR character, and prohibit file names from
1230 containing \f(CW\*(C` \*(C'\fR (space) or any character from the set \f(CW\*(C`!#%&\*(Aq()*;<=>?\*(C'\fR.
1231 .PP
1232 Newer releases of VOS (OpenVOS Release 17.0 or later) support a
1233 feature known as extended names. On these releases, file names
1234 can contain up to 255 characters, are prohibited from starting
1235 with a \f(CW\*(C`\-\*(C'\fR character, and the set of prohibited characters is
1236 reduced to \f(CW\*(C`#%*<>?\*(C'\fR. There are
1237 restrictions involving spaces and apostrophes: these characters
1238 must not begin or end a name, nor can they immediately precede or
1239 follow a period. Additionally, a space must not immediately
1240 precede another space or hyphen. Specifically, the following
1241 character combinations are prohibited: space-space,
1242 space-hyphen, period-space, space-period, period-apostrophe,
1243 apostrophe-period, leading or trailing space, and leading or
1244 trailing apostrophe. Although an extended file name is limited
1245 to 255 characters, a path name is still limited to 256
1246 characters.
1247 .PP
1248 The value of \f(CW$^O\fR on VOS is "vos". To determine the
1249 architecture that you are running on refer to
1250 \&\f(CW$Config{archname}\fR.
1251 .PP
1252 Also see:
1253 .IP \(bu 4
1254 \&\fIREADME.vos\fR (installed as perlvos)
1255 .IP \(bu 4
1256 The VOS mailing list.
1257 .Sp
1258 There is no specific mailing list for Perl on VOS. You can contact
1259 the Stratus Technologies Customer Assistance Center (CAC) for your
1260 region, or you can use the contact information located in the
1261 distribution files on the Stratus Anonymous FTP site.
1262 .IP \(bu 4
1263 Stratus Technologies on the web at <http://www.stratus.com>
1264 .IP \(bu 4
1265 VOS Open-Source Software on the web at <http://ftp.stratus.com/pub/vos/vos.html>
1266 .SS "EBCDIC Platforms"
1267 .IX Subsection "EBCDIC Platforms"
1268 v5.22 core Perl runs on z/OS (formerly OS/390). Theoretically it could
1269 run on the successors of OS/400 on AS/400 minicomputers as well as
1270 VM/ESA, and BS2000 for S/390 Mainframes. Such computers use EBCDIC
1271 character sets internally (usually Character Code Set ID 0037 for OS/400
1272 and either 1047 or POSIX-BC for S/390 systems).
1273 .PP
1274 The rest of this section may need updating, but we don't know what it
1275 should say. Please submit comments to
1276 <https://github.com/Perl/perl5/issues>.
1277 .PP
1278 On the mainframe Perl currently works under the "Unix system
1279 services for OS/390" (formerly known as OpenEdition), VM/ESA OpenEdition, or
1280 the BS200 POSIX-BC system (BS2000 is supported in Perl 5.6 and greater).
1281 See perlos390 for details. Note that for OS/400 there is also a port of
1282 Perl 5.8.1/5.10.0 or later to the PASE which is ASCII-based (as opposed to
1283 ILE which is EBCDIC-based), see perlos400.
1284 .PP
1285 As of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix
1286 sub-systems do not support the \f(CW\*(C`#!\*(C'\fR shebang trick for script invocation.
1287 Hence, on OS/390 and VM/ESA Perl scripts can be executed with a header
1288 similar to the following simple script:
1289 .PP
1290 .Vb 4
1291 \& : # use perl
1292 \& eval \*(Aqexec /usr/local/bin/perl \-S $0 ${1+"$@"}\*(Aq
1293 \& if 0;
1294 \& #!/usr/local/bin/perl # just a comment really
1295 \&
1296 \& print "Hello from perl!\en";
1297 .Ve
1298 .PP
1299 OS/390 will support the \f(CW\*(C`#!\*(C'\fR shebang trick in release 2.8 and beyond.
1300 Calls to \f(CW\*(C`system\*(C'\fR and backticks can use POSIX
1301 shell syntax on all S/390 systems.
1302 .PP
1303 On the AS/400, if PERL5 is in your library list, you may need
1304 to wrap your Perl scripts in a CL procedure to invoke them like so:
1305 .PP
1306 .Vb 3
1307 \& BEGIN
1308 \& CALL PGM(PERL5/PERL) PARM(\*(Aq/QOpenSys/hello.pl\*(Aq)
1309 \& ENDPGM
1310 .Ve
1311 .PP
1312 This will invoke the Perl script \fIhello.pl\fR in the root of the
1313 QOpenSys file system. On the AS/400 calls to
1314 \&\f(CW\*(C`system\*(C'\fR or backticks must use CL syntax.
1315 .PP
1316 On these platforms, bear in mind that the EBCDIC character set may have
1317 an effect on what happens with some Perl functions (such as
1318 \&\f(CW\*(C`chr\*(C'\fR, \f(CW\*(C`pack\*(C'\fR,
1319 \&\f(CW\*(C`print\*(C'\fR,
1320 \&\f(CW\*(C`printf\*(C'\fR,
1321 \&\f(CW\*(C`ord\*(C'\fR, \f(CW\*(C`sort\*(C'\fR,
1322 \&\f(CW\*(C`sprintf\*(C'\fR,
1323 \&\f(CW\*(C`unpack\*(C'\fR), as
1324 well as bit-fiddling with ASCII constants using operators like
1325 \&\f(CW\*(C`^\*(C'\fR, \f(CW\*(C`&\*(C'\fR and \f(CW\*(C`|\*(C'\fR, not to mention
1326 dealing with socket interfaces to ASCII computers (see "Newlines").
1327 .PP
1328 Fortunately, most web servers for the mainframe will correctly
1329 translate the \f(CW\*(C`\en\*(C'\fR in the following statement to its ASCII equivalent
1330 (\f(CW\*(C`\er\*(C'\fR is the same under both Unix and z/OS):
1331 .PP
1332 .Vb 1
1333 \& print "Content\-type: text/html\er\en\er\en";
1334 .Ve
1335 .PP
1336 The values of \f(CW$^O\fR on some of these platforms include:
1337 .PP
1338 .Vb 5
1339 \& uname $^O $Config{archname}
1340 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
1341 \& OS/390 os390 os390
1342 \& OS400 os400 os400
1343 \& POSIX\-BC posix\-bc BS2000\-posix\-bc
1344 .Ve
1345 .PP
1346 Some simple tricks for determining if you are running on an EBCDIC
1347 platform could include any of the following (perhaps all):
1348 .PP
1349 .Vb 1
1350 \& if ("\et" eq "\e005") { print "EBCDIC may be spoken here!\en"; }
1351 \&
1352 \& if (ord(\*(AqA\*(Aq) == 193) { print "EBCDIC may be spoken here!\en"; }
1353 \&
1354 \& if (chr(169) eq \*(Aqz\*(Aq) { print "EBCDIC may be spoken here!\en"; }
1355 .Ve
1356 .PP
1357 One thing you may not want to rely on is the EBCDIC encoding
1358 of punctuation characters since these may differ from code page to code
1359 page (and once your module or script is rumoured to work with EBCDIC,
1360 folks will want it to work with all EBCDIC character sets).
1361 .PP
1362 Also see:
1363 .IP \(bu 4
1364 perlos390, perlos400, perlbs2000, perlebcdic.
1365 .IP \(bu 4
1366 The perl\[email protected] list is for discussion of porting issues as well as
1367 general usage issues for all EBCDIC Perls. Send a message body of
1368 "subscribe perl-mvs" to [email protected].
1369 .IP \(bu 4
1370 AS/400 Perl information at
1371 <http://as400.rochester.ibm.com/>
1372 as well as on CPAN in the \fIports/\fR directory.
1373 .SS "Acorn RISC OS"
1374 .IX Subsection "Acorn RISC OS"
1375 Because Acorns use ASCII with newlines (\f(CW\*(C`\en\*(C'\fR) in text files as \f(CW\*(C`\e012\*(C'\fR like
1376 Unix, and because Unix filename emulation is turned on by default,
1377 most simple scripts will probably work "out of the box". The native
1378 filesystem is modular, and individual filesystems are free to be
1379 case-sensitive or insensitive, and are usually case-preserving. Some
1380 native filesystems have name length limits, which file and directory
1381 names are silently truncated to fit. Scripts should be aware that the
1382 standard filesystem currently has a name length limit of \fB10\fR
1383 characters, with up to 77 items in a directory, but other filesystems
1384 may not impose such limitations.
1385 .PP
1386 Native filenames are of the form
1387 .PP
1388 .Vb 1
1389 \& Filesystem#Special_Field::DiskName.$.Directory.Directory.File
1390 .Ve
1391 .PP
1392 where
1393 .PP
1394 .Vb 8
1395 \& Special_Field is not usually present, but may contain . and $ .
1396 \& Filesystem =~ m|[A\-Za\-z0\-9_]|
1397 \& DsicName =~ m|[A\-Za\-z0\-9_/]|
1398 \& $ represents the root directory
1399 \& . is the path separator
1400 \& @ is the current directory (per filesystem but machine global)
1401 \& ^ is the parent directory
1402 \& Directory and File =~ m|[^\e0\- "\e.\e$\e%\e&:\e@\e\e^\e|\e177]+|
1403 .Ve
1404 .PP
1405 The default filename translation is roughly \f(CW\*(C`tr|/.|./|\*(C'\fR, swapping dots
1406 and slashes.
1407 .PP
1408 Note that \f(CW\*(C`"ADFS::HardDisk.$.File" ne \*(AqADFS::HardDisk.$.File\*(Aq\*(C'\fR and that
1409 the second stage of \f(CW\*(C`$\*(C'\fR interpolation in regular expressions will fall
1410 foul of the \f(CW$.\fR variable if scripts are not careful.
1411 .PP
1412 Logical paths specified by system variables containing comma-separated
1413 search lists are also allowed; hence \f(CW\*(C`System:Modules\*(C'\fR is a valid
1414 filename, and the filesystem will prefix \f(CW\*(C`Modules\*(C'\fR with each section of
1415 \&\f(CW\*(C`System$Path\*(C'\fR until a name is made that points to an object on disk.
1416 Writing to a new file \f(CW\*(C`System:Modules\*(C'\fR would be allowed only if
1417 \&\f(CW\*(C`System$Path\*(C'\fR contains a single item list. The filesystem will also
1418 expand system variables in filenames if enclosed in angle brackets, so
1419 \&\f(CW\*(C`<System$Dir>.Modules\*(C'\fR would look for the file
1420 \&\f(CW\*(C`$ENV{\*(AqSystem$Dir\*(Aq}\ .\ \*(AqModules\*(Aq\*(C'\fR. The obvious implication of this is
1421 that \fBfully qualified filenames can start with \fR\f(CB\*(C`<>\*(C'\fR and the
1422 three-argument form of \f(CW\*(C`open\*(C'\fR should
1423 always be used.
1424 .PP
1425 Because \f(CW\*(C`.\*(C'\fR was in use as a directory separator and filenames could not
1426 be assumed to be unique after 10 characters, Acorn implemented the C
1427 compiler to strip the trailing \f(CW\*(C`.c\*(C'\fR \f(CW\*(C`.h\*(C'\fR \f(CW\*(C`.s\*(C'\fR and \f(CW\*(C`.o\*(C'\fR suffix from
1428 filenames specified in source code and store the respective files in
1429 subdirectories named after the suffix. Hence files are translated:
1430 .PP
1431 .Vb 6
1432 \& foo.h h.foo
1433 \& C:foo.h C:h.foo (logical path variable)
1434 \& sys/os.h sys.h.os (C compiler groks Unix\-speak)
1435 \& 10charname.c c.10charname
1436 \& 10charname.o o.10charname
1437 \& 11charname_.c c.11charname (assuming filesystem truncates at 10)
1438 .Ve
1439 .PP
1440 The Unix emulation library's translation of filenames to native assumes
1441 that this sort of translation is required, and it allows a user-defined list
1442 of known suffixes that it will transpose in this fashion. This may
1443 seem transparent, but consider that with these rules \fIfoo/bar/baz.h\fR
1444 and \fIfoo/bar/h/baz\fR both map to \fIfoo.bar.h.baz\fR, and that
1445 \&\f(CW\*(C`readdir\*(C'\fR and \f(CW\*(C`glob\*(C'\fR
1446 cannot and do not attempt to emulate the reverse mapping. Other
1447 \&\f(CW\*(C`.\*(C'\fR's in filenames are translated to \f(CW\*(C`/\*(C'\fR.
1448 .PP
1449 As implied above, the environment accessed through
1450 \&\f(CW%ENV\fR is global, and the convention is that program
1451 specific environment variables are of the form \f(CW\*(C`Program$Name\*(C'\fR.
1452 Each filesystem maintains a current directory,
1453 and the current filesystem's current directory is the \fBglobal\fR current
1454 directory. Consequently, sociable programs don't change the current
1455 directory but rely on full pathnames, and programs (and Makefiles) cannot
1456 assume that they can spawn a child process which can change the current
1457 directory without affecting its parent (and everyone else for that
1458 matter).
1459 .PP
1460 Because native operating system filehandles are global and are currently
1461 allocated down from 255, with 0 being a reserved value, the Unix emulation
1462 library emulates Unix filehandles. Consequently, you can't rely on
1463 passing \f(CW\*(C`STDIN\*(C'\fR, \f(CW\*(C`STDOUT\*(C'\fR, or \f(CW\*(C`STDERR\*(C'\fR to your children.
1464 .PP
1465 The desire of users to express filenames of the form
1466 \&\f(CW\*(C`<Foo$Dir>.Bar\*(C'\fR on the command line unquoted causes problems,
1467 too: \f(CW\`\`\fR command output capture has
1468 to perform a guessing game. It assumes that a string \f(CW\*(C`<[^<>]+\e$[^<>]>\*(C'\fR
1469 is a reference to an environment variable, whereas anything else involving
1470 \&\f(CW\*(C`<\*(C'\fR or \f(CW\*(C`>\*(C'\fR is redirection, and generally manages to be 99%
1471 right. Of course, the problem remains that scripts cannot rely on any
1472 Unix tools being available, or that any tools found have Unix-like command
1473 line arguments.
1474 .PP
1475 Extensions and XS are, in theory, buildable by anyone using free
1476 tools. In practice, many don't, as users of the Acorn platform are
1477 used to binary distributions. MakeMaker does run, but no available
1478 make currently copes with MakeMaker's makefiles; even if and when
1479 this should be fixed, the lack of a Unix-like shell will cause
1480 problems with makefile rules, especially lines of the form
1481 \&\f(CW\*(C`cd sdbm && make all\*(C'\fR, and anything using quoting.
1482 .PP
1483 "RISC\ OS" is the proper name for the operating system, but the value
1484 in \f(CW$^O\fR is "riscos" (because we don't like shouting).
1485 .SS "Other perls"
1486 .IX Subsection "Other perls"
1487 Perl has been ported to many platforms that do not fit into any of
1488 the categories listed above. Some, such as AmigaOS,
1489 QNX, Plan 9, and VOS, have been well-integrated into the standard
1490 Perl source code kit. You may need to see the \fIports/\fR directory
1491 on CPAN for information, and possibly binaries, for the likes of:
1492 aos, Atari ST, lynxos, riscos, Novell Netware, Tandem Guardian,
1493 \&\fIetc.\fR (Yes, we know that some of these OSes may fall under the
1494 Unix category, but we are not a standards body.)
1495 .PP
1496 Some approximate operating system names and their \f(CW$^O\fR
1497 values in the "OTHER" category include:
1498 .PP
1499 .Vb 3
1500 \& OS $^O $Config{archname}
1501 \& \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
1502 \& Amiga DOS amigaos m68k\-amigos
1503 .Ve
1504 .PP
1505 See also:
1506 .IP \(bu 4
1507 Amiga, \fIREADME.amiga\fR (installed as perlamiga).
1508 .IP \(bu 4
1509 Plan\ 9, \fIREADME.plan9\fR
1510 .SH "FUNCTION IMPLEMENTATIONS"
1511 .IX Header "FUNCTION IMPLEMENTATIONS"
1512 Listed below are functions that are either completely unimplemented
1513 or else have been implemented differently on various platforms.
1514 Preceding each description will be, in parentheses, a list of
1515 platforms that the description applies to.
1516 .PP
1517 The list may well be incomplete, or even wrong in some places. When
1518 in doubt, consult the platform-specific README files in the Perl
1519 source distribution, and any other documentation resources accompanying
1520 a given port.
1521 .PP
1522 Be aware, moreover, that even among Unix-ish systems there are variations.
1523 .PP
1524 For many functions, you can also query \f(CW%Config\fR,
1525 exported by default from the \f(CW\*(C`Config\*(C'\fR module. For example, to
1526 check whether the platform has the \f(CW\*(C`lstat\*(C'\fR
1527 call, check \f(CW$Config{d_lstat}\fR. See Config for a
1528 full description of available variables.
1529 .SS "Alphabetical Listing of Perl Functions"
1530 .IX Subsection "Alphabetical Listing of Perl Functions"
1531 .IP \-X 8
1532 .IX Item "-X"
1533 (Win32)
1534 \&\f(CW\*(C`\-w\*(C'\fR only inspects the read-only file attribute (FILE_ATTRIBUTE_READONLY),
1535 which determines whether the directory can be deleted, not whether it can
1536 be written to. Directories always have read and write access unless denied
1537 by discretionary access control lists (DACLs).
1538 .Sp
1539 (VMS)
1540 \&\f(CW\*(C`\-r\*(C'\fR, \f(CW\*(C`\-w\*(C'\fR, \f(CW\*(C`\-x\*(C'\fR, and \f(CW\*(C`\-o\*(C'\fR tell whether the file is accessible,
1541 which may not reflect UIC-based file protections.
1542 .Sp
1543 (RISC\ OS)
1544 \&\f(CW\*(C`\-s\*(C'\fR by name on an open file will return the space reserved on disk,
1545 rather than the current extent. \f(CW\*(C`\-s\*(C'\fR on an open filehandle returns the
1546 current size.
1547 .Sp
1548 (Win32, VMS, RISC\ OS)
1549 \&\f(CW\*(C`\-R\*(C'\fR, \f(CW\*(C`\-W\*(C'\fR, \f(CW\*(C`\-X\*(C'\fR, \f(CW\*(C`\-O\*(C'\fR are indistinguishable from \f(CW\*(C`\-r\*(C'\fR, \f(CW\*(C`\-w\*(C'\fR,
1550 \&\f(CW\*(C`\-x\*(C'\fR, \f(CW\*(C`\-o\*(C'\fR.
1551 .Sp
1552 (Win32, VMS, RISC\ OS)
1553 \&\f(CW\*(C`\-g\*(C'\fR, \f(CW\*(C`\-k\*(C'\fR, \f(CW\*(C`\-u\*(C'\fR, \f(CW\*(C`\-A\*(C'\fR are not particularly meaningful.
1554 .Sp
1555 (VMS, RISC\ OS)
1556 \&\f(CW\*(C`\-l\*(C'\fR is not particularly meaningful.
1557 .Sp
1558 (Win32)
1559 \&\f(CW\*(C`\-l\*(C'\fR returns true for both symlinks and directory junctions.
1560 .Sp
1561 (VMS, RISC\ OS)
1562 \&\f(CW\*(C`\-p\*(C'\fR is not particularly meaningful.
1563 .Sp
1564 (VMS)
1565 \&\f(CW\*(C`\-d\*(C'\fR is true if passed a device spec without an explicit directory.
1566 .Sp
1567 (Win32)
1568 \&\f(CW\*(C`\-x\*(C'\fR (or \f(CW\*(C`\-X\*(C'\fR) determine if a file ends in one of the executable
1569 suffixes.
1570 .Sp
1571 (RISC\ OS)
1572 \&\f(CW\*(C`\-x\*(C'\fR (or \f(CW\*(C`\-X\*(C'\fR) determine if a file has an executable file type.
1573 .IP alarm 8
1574 .IX Item "alarm"
1575 (Win32)
1576 Emulated using timers that must be explicitly polled whenever Perl
1577 wants to dispatch "safe signals" and therefore cannot interrupt
1578 blocking system calls.
1579 .IP atan2 8
1580 .IX Item "atan2"
1581 (Tru64, HP-UX 10.20)
1582 Due to issues with various CPUs, math libraries, compilers, and standards,
1583 results for \f(CW\*(C`atan2\*(C'\fR may vary depending on any combination of the above.
1584 Perl attempts to conform to the Open Group/IEEE standards for the results
1585 returned from \f(CW\*(C`atan2\*(C'\fR, but cannot force the issue if the system Perl is
1586 run on does not allow it.
1587 .Sp
1588 The current version of the standards for \f(CW\*(C`atan2\*(C'\fR is available at
1589 <http://www.opengroup.org/onlinepubs/009695399/functions/atan2.html>.
1590 .IP binmode 8
1591 .IX Item "binmode"
1592 (RISC\ OS)
1593 Meaningless.
1594 .Sp
1595 (VMS)
1596 Reopens file and restores pointer; if function fails, underlying
1597 filehandle may be closed, or pointer may be in a different position.
1598 .Sp
1599 (Win32)
1600 The value returned by \f(CW\*(C`tell\*(C'\fR may be affected
1601 after the call, and the filehandle may be flushed.
1602 .IP chdir 8
1603 .IX Item "chdir"
1604 (Win32)
1605 The current directory reported by the system may include any symbolic
1606 links specified to \fBchdir()\fR.
1607 .IP chmod 8
1608 .IX Item "chmod"
1609 (Win32)
1610 Only good for changing "owner" read-write access; "group" and "other"
1611 bits are meaningless.
1612 .Sp
1613 (RISC\ OS)
1614 Only good for changing "owner" and "other" read-write access.
1615 .Sp
1616 (VOS)
1617 Access permissions are mapped onto VOS access-control list changes.
1618 .Sp
1619 (Cygwin)
1620 The actual permissions set depend on the value of the \f(CW\*(C`CYGWIN\*(C'\fR variable
1621 in the SYSTEM environment settings.
1622 .Sp
1623 (Android)
1624 Setting the exec bit on some locations (generally \fI/sdcard\fR) will return true
1625 but not actually set the bit.
1626 .Sp
1627 (VMS)
1628 A mode argument of zero sets permissions to the user's default permission mask
1629 rather than disabling all permissions.
1630 .IP chown 8
1631 .IX Item "chown"
1632 (Plan\ 9, RISC\ OS)
1633 Not implemented.
1634 .Sp
1635 (Win32)
1636 Does nothing, but won't fail.
1637 .Sp
1638 (VOS)
1639 A little funky, because VOS's notion of ownership is a little funky.
1640 .IP chroot 8
1641 .IX Item "chroot"
1642 (Win32, VMS, Plan\ 9, RISC\ OS, VOS)
1643 Not implemented.
1644 .IP crypt 8
1645 .IX Item "crypt"
1646 (Win32)
1647 May not be available if library or source was not provided when building
1648 perl.
1649 .Sp
1650 (Android)
1651 Not implemented.
1652 .IP dbmclose 8
1653 .IX Item "dbmclose"
1654 (VMS, Plan\ 9, VOS)
1655 Not implemented.
1656 .IP dbmopen 8
1657 .IX Item "dbmopen"
1658 (VMS, Plan\ 9, VOS)
1659 Not implemented.
1660 .IP dump 8
1661 .IX Item "dump"
1662 (RISC\ OS)
1663 Not useful.
1664 .Sp
1665 (Cygwin, Win32)
1666 Not supported.
1667 .Sp
1668 (VMS)
1669 Invokes VMS debugger.
1670 .IP exec 8
1671 .IX Item "exec"
1672 (Win32)
1673 \&\f(CW\*(C`exec LIST\*(C'\fR without the use of indirect object syntax (\f(CW\*(C`exec PROGRAM LIST\*(C'\fR)
1674 may fall back to trying the shell if the first \f(CWspawn()\fR fails.
1675 .Sp
1676 Note that the list form of \fBexec()\fR is emulated since the Win32 API
1677 \&\fBCreateProcess()\fR accepts a simple string rather than an array of
1678 command-line arguments. This may have security implications for your
1679 code.
1680 .Sp
1681 (SunOS, Solaris, HP-UX)
1682 Does not automatically flush output handles on some platforms.
1683 .IP exit 8
1684 .IX Item "exit"
1685 (VMS)
1686 Emulates Unix \f(CW\*(C`exit\*(C'\fR (which considers \f(CW\*(C`exit 1\*(C'\fR to indicate an error) by
1687 mapping the \f(CW1\fR to \f(CW\*(C`SS$_ABORT\*(C'\fR (\f(CW44\fR). This behavior may be overridden
1688 with the pragma \f(CW\*(C`use vmsish \*(Aqexit\*(Aq\*(C'\fR. As with
1689 the CRTL's \f(CWexit()\fR function, \f(CW\*(C`exit 0\*(C'\fR is also mapped to an exit status
1690 of \f(CW\*(C`SS$_NORMAL\*(C'\fR (\f(CW1\fR); this mapping cannot be overridden. Any other
1691 argument to \f(CW\*(C`exit\*(C'\fR
1692 is used directly as Perl's exit status. On VMS, unless the future
1693 POSIX_EXIT mode is enabled, the exit code should always be a valid
1694 VMS exit code and not a generic number. When the POSIX_EXIT mode is
1695 enabled, a generic number will be encoded in a method compatible with
1696 the C library _POSIX_EXIT macro so that it can be decoded by other
1697 programs, particularly ones written in C, like the GNV package.
1698 .Sp
1699 (Solaris)
1700 \&\f(CW\*(C`exit\*(C'\fR resets file pointers, which is a problem when called
1701 from a child process (created by \f(CW\*(C`fork\*(C'\fR) in
1702 \&\f(CW\*(C`BEGIN\*(C'\fR.
1703 A workaround is to use \f(CW\*(C`POSIX::_exit\*(C'\fR.
1704 .Sp
1705 .Vb 3
1706 \& exit unless $Config{archname} =~ /\ebsolaris\eb/;
1707 \& require POSIX;
1708 \& POSIX::_exit(0);
1709 .Ve
1710 .IP fcntl 8
1711 .IX Item "fcntl"
1712 (Win32)
1713 Not implemented.
1714 .Sp
1715 (VMS)
1716 Some functions available based on the version of VMS.
1717 .IP flock 8
1718 .IX Item "flock"
1719 (VMS, RISC\ OS, VOS)
1720 Not implemented.
1721 .IP fork 8
1722 .IX Item "fork"
1723 (AmigaOS, RISC\ OS, VMS)
1724 Not implemented.
1725 .Sp
1726 (Win32)
1727 Emulated using multiple interpreters. See perlfork.
1728 .Sp
1729 (SunOS, Solaris, HP-UX)
1730 Does not automatically flush output handles on some platforms.
1731 .IP getlogin 8
1732 .IX Item "getlogin"
1733 (RISC\ OS)
1734 Not implemented.
1735 .IP getpgrp 8
1736 .IX Item "getpgrp"
1737 (Win32, VMS, RISC\ OS)
1738 Not implemented.
1739 .IP getppid 8
1740 .IX Item "getppid"
1741 (Win32, RISC\ OS)
1742 Not implemented.
1743 .IP getpriority 8
1744 .IX Item "getpriority"
1745 (Win32, VMS, RISC\ OS, VOS)
1746 Not implemented.
1747 .IP getpwnam 8
1748 .IX Item "getpwnam"
1749 (Win32)
1750 Not implemented.
1751 .Sp
1752 (RISC\ OS)
1753 Not useful.
1754 .IP getgrnam 8
1755 .IX Item "getgrnam"
1756 (Win32, VMS, RISC\ OS)
1757 Not implemented.
1758 .IP getnetbyname 8
1759 .IX Item "getnetbyname"
1760 (Android, Win32, Plan\ 9)
1761 Not implemented.
1762 .IP getpwuid 8
1763 .IX Item "getpwuid"
1764 (Win32)
1765 Not implemented.
1766 .Sp
1767 (RISC\ OS)
1768 Not useful.
1769 .IP getgrgid 8
1770 .IX Item "getgrgid"
1771 (Win32, VMS, RISC\ OS)
1772 Not implemented.
1773 .IP getnetbyaddr 8
1774 .IX Item "getnetbyaddr"
1775 (Android, Win32, Plan\ 9)
1776 Not implemented.
1777 .IP getprotobynumber 8
1778 .IX Item "getprotobynumber"
1779 (Android)
1780 Not implemented.
1781 .IP getpwent 8
1782 .IX Item "getpwent"
1783 (Android, Win32)
1784 Not implemented.
1785 .IP getgrent 8
1786 .IX Item "getgrent"
1787 (Android, Win32, VMS)
1788 Not implemented.
1789 .IP gethostbyname 8
1790 .IX Item "gethostbyname"
1791 (Irix\ 5)
1792 \&\f(CWgethostbyname(\*(Aqlocalhost\*(Aq)\fR does not work everywhere: you may have
1793 to use \f(CWgethostbyname(\*(Aq127.0.0.1\*(Aq)\fR.
1794 .IP gethostent 8
1795 .IX Item "gethostent"
1796 (Win32)
1797 Not implemented.
1798 .IP getnetent 8
1799 .IX Item "getnetent"
1800 (Android, Win32, Plan\ 9)
1801 Not implemented.
1802 .IP getprotoent 8
1803 .IX Item "getprotoent"
1804 (Android, Win32, Plan\ 9)
1805 Not implemented.
1806 .IP getservent 8
1807 .IX Item "getservent"
1808 (Win32, Plan\ 9)
1809 Not implemented.
1810 .IP seekdir 8
1811 .IX Item "seekdir"
1812 (Android)
1813 Not implemented.
1814 .IP sethostent 8
1815 .IX Item "sethostent"
1816 (Android, Win32, Plan\ 9, RISC\ OS)
1817 Not implemented.
1818 .IP setnetent 8
1819 .IX Item "setnetent"
1820 (Win32, Plan\ 9, RISC\ OS)
1821 Not implemented.
1822 .IP setprotoent 8
1823 .IX Item "setprotoent"
1824 (Android, Win32, Plan\ 9, RISC\ OS)
1825 Not implemented.
1826 .IP setservent 8
1827 .IX Item "setservent"
1828 (Plan\ 9, Win32, RISC\ OS)
1829 Not implemented.
1830 .IP endpwent 8
1831 .IX Item "endpwent"
1832 (Win32)
1833 Not implemented.
1834 .Sp
1835 (Android)
1836 Either not implemented or a no-op.
1837 .IP endgrent 8
1838 .IX Item "endgrent"
1839 (Android, RISC\ OS, VMS, Win32)
1840 Not implemented.
1841 .IP endhostent 8
1842 .IX Item "endhostent"
1843 (Android, Win32)
1844 Not implemented.
1845 .IP endnetent 8
1846 .IX Item "endnetent"
1847 (Android, Win32, Plan\ 9)
1848 Not implemented.
1849 .IP endprotoent 8
1850 .IX Item "endprotoent"
1851 (Android, Win32, Plan\ 9)
1852 Not implemented.
1853 .IP endservent 8
1854 .IX Item "endservent"
1855 (Plan\ 9, Win32)
1856 Not implemented.
1857 .IP getsockopt 8
1858 .IX Item "getsockopt"
1859 (Plan\ 9)
1860 Not implemented.
1861 .IP glob 8
1862 .IX Item "glob"
1863 This operator is implemented via the \f(CW\*(C`File::Glob\*(C'\fR extension
1864 on most platforms. See File::Glob for portability information.
1865 .IP gmtime 8
1866 .IX Item "gmtime"
1867 In theory, \f(CW\*(C`gmtime\*(C'\fR is reliable from \-2**63 to 2**63\-1. However,
1868 because work-arounds in the implementation use floating point numbers,
1869 it will become inaccurate as the time gets larger. This is a bug and
1870 will be fixed in the future.
1871 .Sp
1872 (VOS)
1873 Time values are 32\-bit quantities.
1874 .IP ioctl 8
1875 .IX Item "ioctl"
1876 (VMS)
1877 Not implemented.
1878 .Sp
1879 (Win32)
1880 Available only for socket handles, and it does what the \f(CWioctlsocket()\fR call
1881 in the Winsock API does.
1882 .Sp
1883 (RISC\ OS)
1884 Available only for socket handles.
1885 .IP kill 8
1886 .IX Item "kill"
1887 (RISC\ OS)
1888 Not implemented, hence not useful for taint checking.
1889 .Sp
1890 (Win32)
1891 \&\f(CW\*(C`kill\*(C'\fR doesn't send a signal to the identified process like it does on
1892 Unix platforms. Instead \f(CW\*(C`kill($sig, $pid)\*(C'\fR terminates the process
1893 identified by \f(CW$pid\fR, and makes it exit immediately with exit status
1894 \&\f(CW$sig\fR. As in Unix, if \f(CW$sig\fR is 0 and the specified process exists, it
1895 returns true without actually terminating it.
1896 .Sp
1897 (Win32)
1898 \&\f(CW\*(C`kill(\-9, $pid)\*(C'\fR will terminate the process specified by \f(CW$pid\fR and
1899 recursively all child processes owned by it. This is different from
1900 the Unix semantics, where the signal will be delivered to all
1901 processes in the same process group as the process specified by
1902 \&\f(CW$pid\fR.
1903 .Sp
1904 (VMS)
1905 A pid of \-1 indicating all processes on the system is not currently
1906 supported.
1907 .IP link 8
1908 .IX Item "link"
1909 (RISC\ OS, VOS)
1910 Not implemented.
1911 .Sp
1912 (AmigaOS)
1913 Link count not updated because hard links are not quite that hard
1914 (They are sort of half-way between hard and soft links).
1915 .Sp
1916 (Win32)
1917 Hard links are implemented on Win32 under NTFS only. They are
1918 natively supported on Windows 2000 and later. On Windows NT they
1919 are implemented using the Windows POSIX subsystem support and the
1920 Perl process will need Administrator or Backup Operator privileges
1921 to create hard links.
1922 .Sp
1923 (VMS)
1924 Available on 64 bit OpenVMS 8.2 and later.
1925 .IP localtime 8
1926 .IX Item "localtime"
1927 \&\f(CW\*(C`localtime\*(C'\fR has the same range as "gmtime", but because time zone
1928 rules change, its accuracy for historical and future times may degrade
1929 but usually by no more than an hour.
1930 .IP lstat 8
1931 .IX Item "lstat"
1932 (RISC\ OS)
1933 Not implemented.
1934 .Sp
1935 (Win32)
1936 Treats directory junctions as symlinks.
1937 .IP msgctl 8
1938 .IX Item "msgctl"
1939 .PD 0
1940 .IP msgget 8
1941 .IX Item "msgget"
1942 .IP msgsnd 8
1943 .IX Item "msgsnd"
1944 .IP msgrcv 8
1945 .IX Item "msgrcv"
1946 .PD
1947 (Android, Win32, VMS, Plan\ 9, RISC\ OS, VOS)
1948 Not implemented.
1949 .IP open 8
1950 .IX Item "open"
1951 (RISC\ OS)
1952 Open modes \f(CW\*(C`|\-\*(C'\fR and \f(CW\*(C`\-|\*(C'\fR are unsupported.
1953 .Sp
1954 (SunOS, Solaris, HP-UX)
1955 Opening a process does not automatically flush output handles on some
1956 platforms.
1957 .Sp
1958 (Win32)
1959 Both of modes \f(CW\*(C`|\-\*(C'\fR and \f(CW\*(C`\-|\*(C'\fR are supported, but the list form is
1960 emulated since the Win32 API \fBCreateProcess()\fR accepts a simple string
1961 rather than an array of arguments. This may have security
1962 implications for your code.
1963 .IP readlink 8
1964 .IX Item "readlink"
1965 (VMS, RISC\ OS)
1966 Not implemented.
1967 .Sp
1968 (Win32)
1969 \&\fBreadlink()\fR on a directory junction returns the object name, not a
1970 simple path.
1971 .IP rename 8
1972 .IX Item "rename"
1973 (Win32)
1974 Can't move directories between directories on different logical volumes.
1975 .IP rewinddir 8
1976 .IX Item "rewinddir"
1977 (Win32)
1978 Will not cause \f(CW\*(C`readdir\*(C'\fR to re-read the
1979 directory stream. The entries already read before the \f(CW\*(C`rewinddir\*(C'\fR call
1980 will just be returned again from a cache buffer.
1981 .IP select 8
1982 .IX Item "select"
1983 (Win32, VMS)
1984 Only implemented on sockets.
1985 .Sp
1986 (RISC\ OS)
1987 Only reliable on sockets.
1988 .Sp
1989 Note that the \f(CW\*(C`select FILEHANDLE\*(C'\fR form is
1990 generally portable.
1991 .IP semctl 8
1992 .IX Item "semctl"
1993 .PD 0
1994 .IP semget 8
1995 .IX Item "semget"
1996 .IP semop 8
1997 .IX Item "semop"
1998 .PD
1999 (Android, Win32, VMS, RISC\ OS)
2000 Not implemented.
2001 .IP setgrent 8
2002 .IX Item "setgrent"
2003 (Android, VMS, Win32, RISC\ OS)
2004 Not implemented.
2005 .IP setpgrp 8
2006 .IX Item "setpgrp"
2007 (Win32, VMS, RISC\ OS, VOS)
2008 Not implemented.
2009 .IP setpriority 8
2010 .IX Item "setpriority"
2011 (Win32, VMS, RISC\ OS, VOS)
2012 Not implemented.
2013 .IP setpwent 8
2014 .IX Item "setpwent"
2015 (Android, Win32, RISC\ OS)
2016 Not implemented.
2017 .IP setsockopt 8
2018 .IX Item "setsockopt"
2019 (Plan\ 9)
2020 Not implemented.
2021 .IP shmctl 8
2022 .IX Item "shmctl"
2023 .PD 0
2024 .IP shmget 8
2025 .IX Item "shmget"
2026 .IP shmread 8
2027 .IX Item "shmread"
2028 .IP shmwrite 8
2029 .IX Item "shmwrite"
2030 .PD
2031 (Android, Win32, VMS, RISC\ OS)
2032 Not implemented.
2033 .IP sleep 8
2034 .IX Item "sleep"
2035 (Win32)
2036 Emulated using synchronization functions such that it can be
2037 interrupted by \f(CW\*(C`alarm\*(C'\fR, and limited to a
2038 maximum of 4294967 seconds, approximately 49 days.
2039 .IP socketpair 8
2040 .IX Item "socketpair"
2041 (RISC\ OS)
2042 Not implemented.
2043 .Sp
2044 (VMS)
2045 Available on 64 bit OpenVMS 8.2 and later.
2046 .IP stat 8
2047 .IX Item "stat"
2048 Platforms that do not have \f(CW\*(C`rdev\*(C'\fR, \f(CW\*(C`blksize\*(C'\fR, or \f(CW\*(C`blocks\*(C'\fR will return
2049 these as \f(CW\*(Aq\*(Aq\fR, so numeric comparison or manipulation of these fields may
2050 cause 'not numeric' warnings.
2051 .Sp
2052 (Mac\ OS\ X)
2053 \&\f(CW\*(C`ctime\*(C'\fR not supported on UFS.
2054 .Sp
2055 (Win32)
2056 \&\f(CW\*(C`ctime\*(C'\fR is creation time instead of inode change time.
2057 .Sp
2058 (VMS)
2059 \&\f(CW\*(C`dev\*(C'\fR and \f(CW\*(C`ino\*(C'\fR are not necessarily reliable.
2060 .Sp
2061 (RISC\ OS)
2062 \&\f(CW\*(C`mtime\*(C'\fR, \f(CW\*(C`atime\*(C'\fR and \f(CW\*(C`ctime\*(C'\fR all return the last modification time.
2063 \&\f(CW\*(C`dev\*(C'\fR and \f(CW\*(C`ino\*(C'\fR are not necessarily reliable.
2064 .Sp
2065 (OS/2)
2066 \&\f(CW\*(C`dev\*(C'\fR, \f(CW\*(C`rdev\*(C'\fR, \f(CW\*(C`blksize\*(C'\fR, and \f(CW\*(C`blocks\*(C'\fR are not available. \f(CW\*(C`ino\*(C'\fR is not
2067 meaningful and will differ between stat calls on the same file.
2068 .Sp
2069 (Cygwin)
2070 Some versions of cygwin when doing a \f(CWstat("foo")\fR and not finding it
2071 may then attempt to \f(CWstat("foo.exe")\fR.
2072 .IP symlink 8
2073 .IX Item "symlink"
2074 (RISC\ OS)
2075 Not implemented.
2076 .Sp
2077 (Win32)
2078 Requires either elevated permissions or developer mode and a
2079 sufficiently recent version of Windows 10. You can check whether the current
2080 process has the required privileges using the
2081 \&\fBWin32::IsSymlinkCreationAllowed()\fR
2082 function.
2083 .Sp
2084 Since Windows needs to know whether the target is a directory or not when
2085 creating the link the target Perl will only create the link as a directory
2086 link when the target exists and is a directory.
2087 .Sp
2088 Windows does not recognize forward slashes as path separators in
2089 symbolic links. Hence on Windows, any \f(CW\*(C`/\*(C'\fR in the \fIOLDFILE\fR
2090 parameter to \fBsymlink()\fR are converted to \f(CW\*(C`\e\*(C'\fR. This is reflected in
2091 the result returned by \fBreadlink()\fR, the \f(CW\*(C`\e\*(C'\fR in the result are not
2092 converted back to \f(CW\*(C`/\*(C'\fR.
2093 .Sp
2094 (VMS)
2095 Implemented on 64 bit VMS 8.3. VMS requires the symbolic link to be in Unix
2096 syntax if it is intended to resolve to a valid path.
2097 .IP syscall 8
2098 .IX Item "syscall"
2099 (Win32, VMS, RISC\ OS, VOS)
2100 Not implemented.
2101 .IP sysopen 8
2102 .IX Item "sysopen"
2103 (Mac\ OS, OS/390)
2104 The traditional \f(CW0\fR, \f(CW1\fR, and \f(CW2\fR MODEs are implemented with different
2105 numeric values on some systems. The flags exported by \f(CW\*(C`Fcntl\*(C'\fR
2106 (\f(CW\*(C`O_RDONLY\*(C'\fR, \f(CW\*(C`O_WRONLY\*(C'\fR, \f(CW\*(C`O_RDWR\*(C'\fR) should work everywhere though.
2107 .IP system 8
2108 .IX Item "system"
2109 (Win32)
2110 As an optimization, may not call the command shell specified in
2111 \&\f(CW$ENV{PERL5SHELL}\fR. \f(CW\*(C`system(1, @args)\*(C'\fR spawns an external
2112 process and immediately returns its process designator, without
2113 waiting for it to terminate. Return value may be used subsequently
2114 in \f(CW\*(C`wait\*(C'\fR or \f(CW\*(C`waitpid\*(C'\fR.
2115 Failure to \f(CWspawn()\fR a subprocess is indicated by setting
2116 \&\f(CW$?\fR to \f(CW\*(C`255 << 8\*(C'\fR. \f(CW$?\fR is set in a
2117 way compatible with Unix (i.e. the exit status of the subprocess is
2118 obtained by \f(CW\*(C`$? >> 8\*(C'\fR, as described in the documentation).
2119 .Sp
2120 Note that the list form of \fBsystem()\fR is emulated since the Win32 API
2121 \&\fBCreateProcess()\fR accepts a simple string rather than an array of
2122 command-line arguments. This may have security implications for your
2123 code.
2124 .Sp
2125 (RISC\ OS)
2126 There is no shell to process metacharacters, and the native standard is
2127 to pass a command line terminated by "\en" "\er" or "\e0" to the spawned
2128 program. Redirection such as \f(CW\*(C`> foo\*(C'\fR is performed (if at all) by
2129 the run time library of the spawned program. \f(CW\*(C`system LIST\*(C'\fR will call
2130 the Unix emulation library's \f(CW\*(C`exec\*(C'\fR emulation,
2131 which attempts to provide emulation of the stdin, stdout, stderr in force
2132 in the parent, provided the child program uses a compatible version of the
2133 emulation library. \f(CW\*(C`system SCALAR\*(C'\fR will call the native command line
2134 directly and no such emulation of a child Unix program will occur.
2135 Mileage \fBwill\fR vary.
2136 .Sp
2137 (Win32)
2138 \&\f(CW\*(C`system LIST\*(C'\fR without the use of indirect object syntax (\f(CW\*(C`system PROGRAM LIST\*(C'\fR)
2139 may fall back to trying the shell if the first \f(CWspawn()\fR fails.
2140 .Sp
2141 (SunOS, Solaris, HP-UX)
2142 Does not automatically flush output handles on some platforms.
2143 .Sp
2144 (VMS)
2145 As with Win32, \f(CW\*(C`system(1, @args)\*(C'\fR spawns an external process and
2146 immediately returns its process designator without waiting for the
2147 process to terminate. In this case the return value may be used subsequently
2148 in \f(CW\*(C`wait\*(C'\fR or \f(CW\*(C`waitpid\*(C'\fR.
2149 Otherwise the return value is POSIX-like (shifted up by 8 bits), which only
2150 allows room for a made-up value derived from the severity bits of the native
2151 32\-bit condition code (unless overridden by
2152 \&\f(CW\*(C`use vmsish \*(Aqstatus\*(Aq\*(C'\fR). If the native
2153 condition code is one that has a POSIX value encoded, the POSIX value will
2154 be decoded to extract the expected exit value. For more details see
2155 "$?" in perlvms.
2156 .IP telldir 8
2157 .IX Item "telldir"
2158 (Android)
2159 Not implemented.
2160 .IP times 8
2161 .IX Item "times"
2162 (Win32)
2163 "Cumulative" times will be bogus. On anything other than Windows NT
2164 or Windows 2000, "system" time will be bogus, and "user" time is
2165 actually the time returned by the \f(CWclock()\fR function in the C
2166 runtime library.
2167 .Sp
2168 (RISC\ OS)
2169 Not useful.
2170 .IP truncate 8
2171 .IX Item "truncate"
2172 (Older versions of VMS)
2173 Not implemented.
2174 .Sp
2175 (VOS)
2176 Truncation to same-or-shorter lengths only.
2177 .Sp
2178 (Win32)
2179 If a FILEHANDLE is supplied, it must be writable and opened in append
2180 mode (i.e., use \f(CW\*(C`open(my $fh, \*(Aq>>\*(Aq, \*(Aqfilename\*(Aq)\*(C'\fR
2181 or \f(CW\*(C`sysopen(my $fh, ..., O_APPEND|O_RDWR)\*(C'\fR. If a filename is supplied, it
2182 should not be held open elsewhere.
2183 .IP umask 8
2184 .IX Item "umask"
2185 Returns \f(CW\*(C`undef\*(C'\fR where unavailable.
2186 .Sp
2187 (AmigaOS)
2188 \&\f(CW\*(C`umask\*(C'\fR works but the correct permissions are set only when the file
2189 is finally closed.
2190 .IP utime 8
2191 .IX Item "utime"
2192 (VMS, RISC\ OS)
2193 Only the modification time is updated.
2194 .Sp
2195 (Win32)
2196 May not behave as expected. Behavior depends on the C runtime
2197 library's implementation of \f(CWutime()\fR, and the filesystem
2198 being used. The FAT filesystem typically does not support an "access
2199 time" field, and it may limit timestamps to a granularity of two seconds.
2200 .IP wait 8
2201 .IX Item "wait"
2202 .PD 0
2203 .IP waitpid 8
2204 .IX Item "waitpid"
2205 .PD
2206 (Win32)
2207 Can only be applied to process handles returned for processes spawned
2208 using \f(CW\*(C`system(1, ...)\*(C'\fR or pseudo processes created with
2209 \&\f(CW\*(C`fork\*(C'\fR.
2210 .Sp
2211 (RISC\ OS)
2212 Not useful.
2213 .SH "Supported Platforms"
2214 .IX Header "Supported Platforms"
2215 The following platforms are known to build Perl 5.12 (as of April 2010,
2216 its release date) from the standard source code distribution available
2217 at <http://www.cpan.org/src>
2218 .IP "Linux (x86, ARM, IA64)" 4
2219 .IX Item "Linux (x86, ARM, IA64)"
2220 .PD 0
2221 .IP HP-UX 4
2222 .IX Item "HP-UX"
2223 .IP AIX 4
2224 .IX Item "AIX"
2225 .IP Win32 4
2226 .IX Item "Win32"
2227 .RS 4
2228 .IP "Windows 2000" 4
2229 .IX Item "Windows 2000"
2230 .IP "Windows XP" 4
2231 .IX Item "Windows XP"
2232 .IP "Windows Server 2003" 4
2233 .IX Item "Windows Server 2003"
2234 .IP "Windows Vista" 4
2235 .IX Item "Windows Vista"
2236 .IP "Windows Server 2008" 4
2237 .IX Item "Windows Server 2008"
2238 .IP "Windows 7" 4
2239 .IX Item "Windows 7"
2240 .RE
2241 .RS 4
2242 .RE
2243 .IP Cygwin 4
2244 .IX Item "Cygwin"
2245 .PD
2246 Some tests are known to fail:
2247 .RS 4
2248 .IP \(bu 4
2249 \&\fIext/XS\-APItest/t/call_checker.t\fR \- see
2250 <https://github.com/Perl/perl5/issues/10750>
2251 .IP \(bu 4
2252 \&\fIdist/I18N\-Collate/t/I18N\-Collate.t\fR
2253 .IP \(bu 4
2254 \&\fIext/Win32CORE/t/win32core.t\fR \- may fail on recent cygwin installs.
2255 .RE
2256 .RS 4
2257 .RE
2258 .IP "Solaris (x86, SPARC)" 4
2259 .IX Item "Solaris (x86, SPARC)"
2260 .PD 0
2261 .IP OpenVMS 4
2262 .IX Item "OpenVMS"
2263 .RS 4
2264 .IP "Alpha (7.2 and later)" 4
2265 .IX Item "Alpha (7.2 and later)"
2266 .IP "I64 (8.2 and later)" 4
2267 .IX Item "I64 (8.2 and later)"
2268 .RE
2269 .RS 4
2270 .RE
2271 .IP NetBSD 4
2272 .IX Item "NetBSD"
2273 .IP FreeBSD 4
2274 .IX Item "FreeBSD"
2275 .IP "Debian GNU/kFreeBSD" 4
2276 .IX Item "Debian GNU/kFreeBSD"
2277 .IP Haiku 4
2278 .IX Item "Haiku"
2279 .IP "Irix (6.5. What else?)" 4
2280 .IX Item "Irix (6.5. What else?)"
2281 .IP OpenBSD 4
2282 .IX Item "OpenBSD"
2283 .IP "Dragonfly BSD" 4
2284 .IX Item "Dragonfly BSD"
2285 .IP "Midnight BSD" 4
2286 .IX Item "Midnight BSD"
2287 .IP "QNX Neutrino RTOS (6.5.0)" 4
2288 .IX Item "QNX Neutrino RTOS (6.5.0)"
2289 .IP "MirOS BSD" 4
2290 .IX Item "MirOS BSD"
2291 .IP "Stratus OpenVOS (17.0 or later)" 4
2292 .IX Item "Stratus OpenVOS (17.0 or later)"
2293 .PD
2294 Caveats:
2295 .RS 4
2296 .IP "time_t issues that may or may not be fixed" 4
2297 .IX Item "time_t issues that may or may not be fixed"
2298 .RE
2299 .RS 4
2300 .RE
2301 .PD 0
2302 .IP "Stratus VOS / OpenVOS" 4
2303 .IX Item "Stratus VOS / OpenVOS"
2304 .IP AIX 4
2305 .IX Item "AIX"
2306 .IP Android 4
2307 .IX Item "Android"
2308 .IP FreeMINT 4
2309 .IX Item "FreeMINT"
2310 .PD
2311 Perl now builds with FreeMiNT/Atari. It fails a few tests, that needs
2312 some investigation.
2313 .Sp
2314 The FreeMiNT port uses GNU dld for loadable module capabilities. So
2315 ensure you have that library installed when building perl.
2316 .SH "EOL Platforms"
2317 .IX Header "EOL Platforms"
2318 .SS "(Perl 5.37.1)"
2319 .IX Subsection "(Perl 5.37.1)"
2320 The following platforms were supported by a previous version of
2321 Perl but have been officially removed from Perl's source code
2322 as of 5.37.1:
2323 .IP Ultrix 4
2324 .IX Item "Ultrix"
2325 .SS "(Perl 5.36)"
2326 .IX Subsection "(Perl 5.36)"
2327 The following platforms were supported by a previous version of
2328 Perl but have been officially removed from Perl's source code
2329 as of 5.36:
2330 .IP NetWare 4
2331 .IX Item "NetWare"
2332 .PD 0
2333 .IP DOS/DJGPP 4
2334 .IX Item "DOS/DJGPP"
2335 .IP "AT&T UWIN" 4
2336 .IX Item "AT&T UWIN"
2337 .PD
2338 .SS "(Perl 5.20)"
2339 .IX Subsection "(Perl 5.20)"
2340 The following platforms were supported by a previous version of
2341 Perl but have been officially removed from Perl's source code
2342 as of 5.20:
2343 .IP "AT&T 3b1" 4
2344 .IX Item "AT&T 3b1"
2345 .SS "(Perl 5.14)"
2346 .IX Subsection "(Perl 5.14)"
2347 The following platforms were supported up to 5.10. They may still
2348 have worked in 5.12, but supporting code has been removed for 5.14:
2349 .IP "Windows 95" 4
2350 .IX Item "Windows 95"
2351 .PD 0
2352 .IP "Windows 98" 4
2353 .IX Item "Windows 98"
2354 .IP "Windows ME" 4
2355 .IX Item "Windows ME"
2356 .IP "Windows NT4" 4
2357 .IX Item "Windows NT4"
2358 .PD
2359 .SS "(Perl 5.12)"
2360 .IX Subsection "(Perl 5.12)"
2361 The following platforms were supported by a previous version of
2362 Perl but have been officially removed from Perl's source code
2363 as of 5.12:
2364 .IP "Atari MiNT" 4
2365 .IX Item "Atari MiNT"
2366 .PD 0
2367 .IP "Apollo Domain/OS" 4
2368 .IX Item "Apollo Domain/OS"
2369 .IP "Apple Mac OS 8/9" 4
2370 .IX Item "Apple Mac OS 8/9"
2371 .IP "Tenon Machten" 4
2372 .IX Item "Tenon Machten"
2373 .PD
2374 .SH "Supported Platforms (Perl 5.8)"
2375 .IX Header "Supported Platforms (Perl 5.8)"
2376 As of July 2002 (the Perl release 5.8.0), the following platforms were
2377 able to build Perl from the standard source code distribution
2378 available at <http://www.cpan.org/src/>
2379 .PP
2380 .Vb 10
2381 \& AIX
2382 \& BeOS
2383 \& BSD/OS (BSDi)
2384 \& Cygwin
2385 \& DG/UX
2386 \& DOS DJGPP 1)
2387 \& DYNIX/ptx
2388 \& EPOC R5
2389 \& FreeBSD
2390 \& HI\-UXMPP (Hitachi) (5.8.0 worked but we didn\*(Aqt know it)
2391 \& HP\-UX
2392 \& IRIX
2393 \& Linux
2394 \& Mac OS Classic
2395 \& Mac OS X (Darwin)
2396 \& MPE/iX
2397 \& NetBSD
2398 \& NetWare
2399 \& NonStop\-UX
2400 \& ReliantUNIX (formerly SINIX)
2401 \& OpenBSD
2402 \& OpenVMS (formerly VMS)
2403 \& Open UNIX (Unixware) (since Perl 5.8.1/5.9.0)
2404 \& OS/2
2405 \& OS/400 (using the PASE) (since Perl 5.8.1/5.9.0)
2406 \& POSIX\-BC (formerly BS2000)
2407 \& QNX
2408 \& Solaris
2409 \& SunOS 4
2410 \& SUPER\-UX (NEC)
2411 \& Tru64 UNIX (formerly DEC OSF/1, Digital UNIX)
2412 \& UNICOS
2413 \& UNICOS/mk
2414 \& UTS
2415 \& VOS / OpenVOS
2416 \& Win95/98/ME/2K/XP 2)
2417 \& WinCE
2418 \& z/OS (formerly OS/390)
2419 \& VM/ESA
2420 \&
2421 \& 1) in DOS mode either the DOS or OS/2 ports can be used
2422 \& 2) compilers: Borland, MinGW (GCC), VC6
2423 .Ve
2424 .PP
2425 The following platforms worked with the previous releases (5.6 and
2426 5.7), but we did not manage either to fix or to test these in time
2427 for the 5.8.0 release. There is a very good chance that many of these
2428 will work fine with the 5.8.0.
2429 .PP
2430 .Vb 10
2431 \& BSD/OS
2432 \& DomainOS
2433 \& Hurd
2434 \& LynxOS
2435 \& MachTen
2436 \& PowerMAX
2437 \& SCO SV
2438 \& SVR4
2439 \& Unixware
2440 \& Windows 3.1
2441 .Ve
2442 .PP
2443 Known to be broken for 5.8.0 (but 5.6.1 and 5.7.2 can be used):
2444 .PP
2445 .Vb 1
2446 \& AmigaOS 3
2447 .Ve
2448 .PP
2449 The following platforms have been known to build Perl from source in
2450 the past (5.005_03 and earlier), but we haven't been able to verify
2451 their status for the current release, either because the
2452 hardware/software platforms are rare or because we don't have an
2453 active champion on these platforms\-\-or both. They used to work,
2454 though, so go ahead and try compiling them, and let
2455 <https://github.com/Perl/perl5/issues> know
2456 of any trouble.
2457 .PP
2458 .Vb 10
2459 \& 3b1
2460 \& A/UX
2461 \& ConvexOS
2462 \& CX/UX
2463 \& DC/OSx
2464 \& DDE SMES
2465 \& DOS EMX
2466 \& Dynix
2467 \& EP/IX
2468 \& ESIX
2469 \& FPS
2470 \& GENIX
2471 \& Greenhills
2472 \& ISC
2473 \& MachTen 68k
2474 \& MPC
2475 \& NEWS\-OS
2476 \& NextSTEP
2477 \& OpenSTEP
2478 \& Opus
2479 \& Plan 9
2480 \& RISC/os
2481 \& SCO ODT/OSR
2482 \& Stellar
2483 \& SVR2
2484 \& TI1500
2485 \& TitanOS
2486 \& Unisys Dynix
2487 .Ve
2488 .PP
2489 The following platforms have their own source code distributions and
2490 binaries available via <http://www.cpan.org/ports/>
2491 .PP
2492 .Vb 1
2493 \& Perl release
2494 \&
2495 \& OS/400 (ILE) 5.005_02
2496 \& Tandem Guardian 5.004
2497 .Ve
2498 .PP
2499 The following platforms have only binaries available via
2500 <http://www.cpan.org/ports/index.html> :
2501 .PP
2502 .Vb 1
2503 \& Perl release
2504 \&
2505 \& Acorn RISCOS 5.005_02
2506 \& AOS 5.002
2507 \& LynxOS 5.004_02
2508 .Ve
2509 .PP
2510 Although we do suggest that you always build your own Perl from
2511 the source code, both for maximal configurability and for security,
2512 in case you are in a hurry you can check
2513 <http://www.cpan.org/ports/index.html> for binary distributions.
2514 .SH "SEE ALSO"
2515 .IX Header "SEE ALSO"
2516 perlaix, perlamiga, perlbs2000,
2517 perlcygwin,
2518 perlebcdic, perlfreebsd, perlhurd, perlhpux, perlirix,
2519 perlmacosx,
2520 perlos2, perlos390, perlos400,
2521 perlplan9, perlqnx, perlsolaris, perltru64,
2522 perlunicode, perlvms, perlvos, perlwin32, and Win32.
2523 .SH "AUTHORS / CONTRIBUTORS"
2524 .IX Header "AUTHORS / CONTRIBUTORS"
2525 Abigail <[email protected]>,
2526 Charles Bailey <[email protected]>,
2527 Graham Barr <[email protected]>,
2528 Tom Christiansen <[email protected]>,
2529 Nicholas Clark <[email protected]>,
2530 Thomas Dorner <[email protected]>,
2531 Andy Dougherty <[email protected]>,
2532 Dominic Dunlop <[email protected]>,
2533 Neale Ferguson <[email protected]>,
2534 David J. Fiander <[email protected]>,
2535 Paul Green <[email protected]>,
2536 M.J.T. Guy <[email protected]>,
2537 Jarkko Hietaniemi <[email protected]>,
2538 Luther Huffman <[email protected]>,
2539 Nick Ing-Simmons <nick@ing\-simmons.net>,
2540 Andreas J. König <[email protected]>,
2541 Markus Laker <[email protected]>,
2542 Andrew M. Langmead <[email protected]>,
2543 Lukas Mai <[email protected]>,
2544 Larry Moore <[email protected]>,
2545 Paul Moore <[email protected]\-it.com>,
2546 Chris Nandor <[email protected]>,
2547 Matthias Neeracher <[email protected]>,
2548 Philip Newton <[email protected]>,
2549 Gary Ng <[email protected]>,
2550 Tom Phoenix <[email protected]>,
2551 André Pirard <[email protected]>,
2552 Peter Prymmer <[email protected]>,
2553 Hugo van der Sanden <[email protected]>,
2554 Gurusamy Sarathy <[email protected]>,
2555 Paul J. Schinder <[email protected]>,
2556 Michael G Schwern <[email protected]>,
2557 Dan Sugalski <[email protected]>,
2558 Nathan Torkington <[email protected]>,
2559 John Malmberg <[email protected]>