1=head1 NAME 2 3perlport - Writing portable Perl 4 5=head1 DESCRIPTION 6 7Perl runs on numerous operating systems. While most of them share 8much in common, they also have their own unique features. 9 10This document is meant to help you to find out what constitutes portable 11Perl code. That way once you make a decision to write portably, 12you know where the lines are drawn, and you can stay within them. 13 14There is a tradeoff between taking full advantage of one particular 15type of computer and taking advantage of a full range of them. 16Naturally, as you broaden your range and become more diverse, the 17common factors drop, and you are left with an increasingly smaller 18area of common ground in which you can operate to accomplish a 19particular task. Thus, when you begin attacking a problem, it is 20important to consider under which part of the tradeoff curve you 21want to operate. Specifically, you must decide whether it is 22important that the task that you are coding has the full generality 23of being portable, or whether to just get the job done right now. 24This is the hardest choice to be made. The rest is easy, because 25Perl provides many choices, whichever way you want to approach your 26problem. 27 28Looking at it another way, writing portable code is usually about 29willfully limiting your available choices. Naturally, it takes 30discipline and sacrifice to do that. The product of portability 31and convenience may be a constant. You have been warned. 32 33Be aware of two important points: 34 35=over 4 36 37=item Not all Perl programs have to be portable 38 39There is no reason you should not use Perl as a language to glue Unix 40tools together, or to prototype a Macintosh application, or to manage the 41Windows registry. If it makes no sense to aim for portability for one 42reason or another in a given program, then don't bother. 43 44=item Nearly all of Perl already I<is> portable 45 46Don't be fooled into thinking that it is hard to create portable Perl 47code. It isn't. Perl tries its level-best to bridge the gaps between 48what's available on different platforms, and all the means available to 49use those features. Thus almost all Perl code runs on any machine 50without modification. But there are some significant issues in 51writing portable code, and this document is entirely about those issues. 52 53=back 54 55Here's the general rule: When you approach a task commonly done 56using a whole range of platforms, think about writing portable 57code. That way, you don't sacrifice much by way of the implementation 58choices you can avail yourself of, and at the same time you can give 59your users lots of platform choices. On the other hand, when you have to 60take advantage of some unique feature of a particular platform, as is 61often the case with systems programming (whether for Unix, Windows, 62VMS, etc.), consider writing platform-specific code. 63 64When the code will run on only two or three operating systems, you 65may need to consider only the differences of those particular systems. 66The important thing is to decide where the code will run and to be 67deliberate in your decision. 68 69The material below is separated into three main sections: main issues of 70portability (L</"ISSUES">), platform-specific issues (L</"PLATFORMS">), and 71built-in Perl functions that behave differently on various ports 72(L</"FUNCTION IMPLEMENTATIONS">). 73 74This information should not be considered complete; it includes possibly 75transient information about idiosyncrasies of some of the ports, almost 76all of which are in a state of constant evolution. Thus, this material 77should be considered a perpetual work in progress 78(C<< <IMG SRC="yellow_sign.gif" ALT="Under Construction"> >>). 79 80=head1 ISSUES 81 82=head2 Newlines 83 84In most operating systems, lines in files are terminated by newlines. 85Just what is used as a newline may vary from OS to OS. Unix 86traditionally uses C<\012>, one type of DOSish I/O uses C<\015\012>, 87S<Mac OS> uses C<\015>, and z/OS uses C<\025>. 88 89Perl uses C<\n> to represent the "logical" newline, where what is 90logical may depend on the platform in use. In MacPerl, C<\n> always 91means C<\015>. On EBCDIC platforms, C<\n> could be C<\025> or C<\045>. 92In DOSish perls, C<\n> usually means C<\012>, but when 93accessing a file in "text" mode, perl uses the C<:crlf> layer that 94translates it to (or from) C<\015\012>, depending on whether you're 95reading or writing. Unix does the same thing on ttys in canonical 96mode. C<\015\012> is commonly referred to as CRLF. 97 98To trim trailing newlines from text lines use 99L<C<chomp>|perlfunc/chomp VARIABLE>. With default settings that function 100looks for a trailing C<\n> character and thus trims in a portable way. 101 102When dealing with binary files (or text files in binary mode) be sure 103to explicitly set L<C<$E<sol>>|perlvar/$E<sol>> to the appropriate value for 104your file format before using L<C<chomp>|perlfunc/chomp VARIABLE>. 105 106Because of the "text" mode translation, DOSish perls have limitations in 107using L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> and 108L<C<tell>|perlfunc/tell FILEHANDLE> on a file accessed in "text" mode. 109Stick to L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE>-ing to 110locations you got from L<C<tell>|perlfunc/tell FILEHANDLE> (and no 111others), and you are usually free to use 112L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> and 113L<C<tell>|perlfunc/tell FILEHANDLE> even in "text" mode. Using 114L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> or 115L<C<tell>|perlfunc/tell FILEHANDLE> or other file operations may be 116non-portable. If you use L<C<binmode>|perlfunc/binmode FILEHANDLE> on a 117file, however, you can usually 118L<C<seek>|perlfunc/seek FILEHANDLE,POSITION,WHENCE> and 119L<C<tell>|perlfunc/tell FILEHANDLE> with arbitrary values safely. 120 121A common misconception in socket programming is that S<C<\n eq \012>> 122everywhere. When using protocols such as common Internet protocols, 123C<\012> and C<\015> are called for specifically, and the values of 124the logical C<\n> and C<\r> (carriage return) are not reliable. 125 126 print $socket "Hi there, client!\r\n"; # WRONG 127 print $socket "Hi there, client!\015\012"; # RIGHT 128 129However, using C<\015\012> (or C<\cM\cJ>, or C<\x0D\x0A>) can be tedious 130and unsightly, as well as confusing to those maintaining the code. As 131such, the L<C<Socket>|Socket> module supplies the Right Thing for those 132who want it. 133 134 use Socket qw(:DEFAULT :crlf); 135 print $socket "Hi there, client!$CRLF" # RIGHT 136 137When reading from a socket, remember that the default input record 138separator L<C<$E<sol>>|perlvar/$E<sol>> is C<\n>, but robust socket code 139will recognize as either C<\012> or C<\015\012> as end of line: 140 141 while (<$socket>) { # NOT ADVISABLE! 142 # ... 143 } 144 145Because both CRLF and LF end in LF, the input record separator can 146be set to LF and any CR stripped later. Better to write: 147 148 use Socket qw(:DEFAULT :crlf); 149 local($/) = LF; # not needed if $/ is already \012 150 151 while (<$socket>) { 152 s/$CR?$LF/\n/; # not sure if socket uses LF or CRLF, OK 153 # s/\015?\012/\n/; # same thing 154 } 155 156This example is preferred over the previous one--even for Unix 157platforms--because now any C<\015>'s (C<\cM>'s) are stripped out 158(and there was much rejoicing). 159 160Similarly, functions that return text data--such as a function that 161fetches a web page--should sometimes translate newlines before 162returning the data, if they've not yet been translated to the local 163newline representation. A single line of code will often suffice: 164 165 $data =~ s/\015?\012/\n/g; 166 return $data; 167 168Some of this may be confusing. Here's a handy reference to the ASCII CR 169and LF characters. You can print it out and stick it in your wallet. 170 171 LF eq \012 eq \x0A eq \cJ eq chr(10) eq ASCII 10 172 CR eq \015 eq \x0D eq \cM eq chr(13) eq ASCII 13 173 174 | Unix | DOS | Mac | 175 --------------------------- 176 \n | LF | LF | CR | 177 \r | CR | CR | LF | 178 \n * | LF | CRLF | CR | 179 \r * | CR | CR | LF | 180 --------------------------- 181 * text-mode STDIO 182 183The Unix column assumes that you are not accessing a serial line 184(like a tty) in canonical mode. If you are, then CR on input becomes 185"\n", and "\n" on output becomes CRLF. 186 187These are just the most common definitions of C<\n> and C<\r> in Perl. 188There may well be others. For example, on an EBCDIC implementation 189such as z/OS (OS/390) or OS/400 (using the ILE, the PASE is ASCII-based) 190the above material is similar to "Unix" but the code numbers change: 191 192 LF eq \025 eq \x15 eq \cU eq chr(21) eq CP-1047 21 193 LF eq \045 eq \x25 eq chr(37) eq CP-0037 37 194 CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-1047 13 195 CR eq \015 eq \x0D eq \cM eq chr(13) eq CP-0037 13 196 197 | z/OS | OS/400 | 198 ---------------------- 199 \n | LF | LF | 200 \r | CR | CR | 201 \n * | LF | LF | 202 \r * | CR | CR | 203 ---------------------- 204 * text-mode STDIO 205 206=head2 Numbers endianness and Width 207 208Different CPUs store integers and floating point numbers in different 209orders (called I<endianness>) and widths (32-bit and 64-bit being the 210most common today). This affects your programs when they attempt to transfer 211numbers in binary format from one CPU architecture to another, 212usually either "live" via network connection, or by storing the 213numbers to secondary storage such as a disk file or tape. 214 215Conflicting storage orders make an utter mess out of the numbers. If a 216little-endian host (Intel, VAX) stores 0x12345678 (305419896 in 217decimal), a big-endian host (Motorola, Sparc, PA) reads it as 2180x78563412 (2018915346 in decimal). Alpha and MIPS can be either: 219Digital/Compaq used/uses them in little-endian mode; SGI/Cray uses 220them in big-endian mode. To avoid this problem in network (socket) 221connections use the L<C<pack>|perlfunc/pack TEMPLATE,LIST> and 222L<C<unpack>|perlfunc/unpack TEMPLATE,EXPR> formats C<n> and C<N>, the 223"network" orders. These are guaranteed to be portable. 224 225As of Perl 5.10.0, you can also use the C<E<gt>> and C<E<lt>> modifiers 226to force big- or little-endian byte-order. This is useful if you want 227to store signed integers or 64-bit integers, for example. 228 229You can explore the endianness of your platform by unpacking a 230data structure packed in native format such as: 231 232 print unpack("h*", pack("s2", 1, 2)), "\n"; 233 # '10002000' on e.g. Intel x86 or Alpha 21064 in little-endian mode 234 # '00100020' on e.g. Motorola 68040 235 236If you need to distinguish between endian architectures you could use 237either of the variables set like so: 238 239 $is_big_endian = unpack("h*", pack("s", 1)) =~ /01/; 240 $is_little_endian = unpack("h*", pack("s", 1)) =~ /^1/; 241 242Differing widths can cause truncation even between platforms of equal 243endianness. The platform of shorter width loses the upper parts of the 244number. There is no good solution for this problem except to avoid 245transferring or storing raw binary numbers. 246 247One can circumnavigate both these problems in two ways. Either 248transfer and store numbers always in text format, instead of raw 249binary, or else consider using modules like 250L<C<Data::Dumper>|Data::Dumper> and L<C<Storable>|Storable> (included as 251of Perl 5.8). Keeping all data as text significantly simplifies matters. 252 253=head2 Files and Filesystems 254 255Most platforms these days structure files in a hierarchical fashion. 256So, it is reasonably safe to assume that all platforms support the 257notion of a "path" to uniquely identify a file on the system. How 258that path is really written, though, differs considerably. 259 260Although similar, file path specifications differ between Unix, 261Windows, S<Mac OS>, OS/2, VMS, VOS, S<RISC OS>, and probably others. 262Unix, for example, is one of the few OSes that has the elegant idea 263of a single root directory. 264 265DOS, OS/2, VMS, VOS, and Windows can work similarly to Unix with C</> 266as path separator, or in their own idiosyncratic ways (such as having 267several root directories and various "unrooted" device files such NIL: 268and LPT:). 269 270S<Mac OS> 9 and earlier used C<:> as a path separator instead of C</>. 271 272The filesystem may support neither hard links 273(L<C<link>|perlfunc/link OLDFILE,NEWFILE>) nor symbolic links 274(L<C<symlink>|perlfunc/symlink OLDFILE,NEWFILE>, 275L<C<readlink>|perlfunc/readlink EXPR>, 276L<C<lstat>|perlfunc/lstat FILEHANDLE>). 277 278The filesystem may support neither access timestamp nor change 279timestamp (meaning that about the only portable timestamp is the 280modification timestamp), or one second granularity of any timestamps 281(e.g. the FAT filesystem limits the time granularity to two seconds). 282 283The "inode change timestamp" (the L<C<-C>|perlfunc/-X FILEHANDLE> 284filetest) may really be the "creation timestamp" (which it is not in 285Unix). 286 287VOS perl can emulate Unix filenames with C</> as path separator. The 288native pathname characters greater-than, less-than, number-sign, and 289percent-sign are always accepted. 290 291S<RISC OS> perl can emulate Unix filenames with C</> as path 292separator, or go native and use C<.> for path separator and C<:> to 293signal filesystems and disk names. 294 295Don't assume Unix filesystem access semantics: that read, write, 296and execute are all the permissions there are, and even if they exist, 297that their semantics (for example what do C<r>, C<w>, and C<x> mean on 298a directory) are the Unix ones. The various Unix/POSIX compatibility 299layers usually try to make interfaces like L<C<chmod>|perlfunc/chmod LIST> 300work, but sometimes there simply is no good mapping. 301 302The L<C<File::Spec>|File::Spec> modules provide methods to manipulate path 303specifications and return the results in native format for each 304platform. This is often unnecessary as Unix-style paths are 305understood by Perl on every supported platform, but if you need to 306produce native paths for a native utility that does not understand 307Unix syntax, or if you are operating on paths or path components 308in unknown (and thus possibly native) syntax, L<C<File::Spec>|File::Spec> 309is your friend. Here are two brief examples: 310 311 use File::Spec::Functions; 312 chdir(updir()); # go up one directory 313 314 # Concatenate a path from its components 315 my $file = catfile(updir(), 'temp', 'file.txt'); 316 # on Unix: '../temp/file.txt' 317 # on Win32: '..\temp\file.txt' 318 # on VMS: '[-.temp]file.txt' 319 320In general, production code should not have file paths hardcoded. 321Making them user-supplied or read from a configuration file is 322better, keeping in mind that file path syntax varies on different 323machines. 324 325This is especially noticeable in scripts like Makefiles and test suites, 326which often assume C</> as a path separator for subdirectories. 327 328Also of use is L<C<File::Basename>|File::Basename> from the standard 329distribution, which splits a pathname into pieces (base filename, full 330path to directory, and file suffix). 331 332Even when on a single platform (if you can call Unix a single platform), 333remember not to count on the existence or the contents of particular 334system-specific files or directories, like F</etc/passwd>, 335F</etc/sendmail.conf>, F</etc/resolv.conf>, or even F</tmp/>. For 336example, F</etc/passwd> may exist but not contain the encrypted 337passwords, because the system is using some form of enhanced security. 338Or it may not contain all the accounts, because the system is using NIS. 339If code does need to rely on such a file, include a description of the 340file and its format in the code's documentation, then make it easy for 341the user to override the default location of the file. 342 343Don't assume a text file will end with a newline. They should, 344but people forget. 345 346Do not have two files or directories of the same name with different 347case, like F<test.pl> and F<Test.pl>, as many platforms have 348case-insensitive (or at least case-forgiving) filenames. Also, try 349not to have non-word characters (except for C<.>) in the names, and 350keep them to the 8.3 convention, for maximum portability, onerous a 351burden though this may appear. 352 353Likewise, when using the L<C<AutoSplit>|AutoSplit> module, try to keep 354your functions to 8.3 naming and case-insensitive conventions; or, at the 355least, make it so the resulting files have a unique (case-insensitively) 356first 8 characters. 357 358Whitespace in filenames is tolerated on most systems, but not all, 359and even on systems where it might be tolerated, some utilities 360might become confused by such whitespace. 361 362Many systems (DOS, VMS ODS-2) cannot have more than one C<.> in their 363filenames. 364 365Don't assume C<< > >> won't be the first character of a filename. 366Always use the three-arg version of 367L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR>: 368 369 open my $fh, '<', $existing_file) or die $!; 370 371Two-arg L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> is magic and can 372translate characters like C<< > >>, C<< < >>, and C<|> in filenames, 373which is usually the wrong thing to do. 374L<C<sysopen>|perlfunc/sysopen FILEHANDLE,FILENAME,MODE> and three-arg 375L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> don't have this problem. 376 377Don't use C<:> as a part of a filename since many systems use that for 378their own semantics (Mac OS Classic for separating pathname components, 379many networking schemes and utilities for separating the nodename and 380the pathname, and so on). For the same reasons, avoid C<@>, C<;> and 381C<|>. 382 383Don't assume that in pathnames you can collapse two leading slashes 384C<//> into one: some networking and clustering filesystems have special 385semantics for that. Let the operating system sort it out. 386 387The I<portable filename characters> as defined by ANSI C are 388 389 a b c d e f g h i j k l m n o p q r s t u v w x y z 390 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z 391 0 1 2 3 4 5 6 7 8 9 392 . _ - 393 394and C<-> shouldn't be the first character. If you want to be 395hypercorrect, stay case-insensitive and within the 8.3 naming 396convention (all the files and directories have to be unique within one 397directory if their names are lowercased and truncated to eight 398characters before the C<.>, if any, and to three characters after the 399C<.>, if any). (And do not use C<.>s in directory names.) 400 401=head2 System Interaction 402 403Not all platforms provide a command line. These are usually platforms 404that rely primarily on a Graphical User Interface (GUI) for user 405interaction. A program requiring a command line interface might 406not work everywhere. This is probably for the user of the program 407to deal with, so don't stay up late worrying about it. 408 409Some platforms can't delete or rename files held open by the system, 410this limitation may also apply to changing filesystem metainformation 411like file permissions or owners. Remember to 412L<C<close>|perlfunc/close FILEHANDLE> files when you are done with them. 413Don't L<C<unlink>|perlfunc/unlink LIST> or 414L<C<rename>|perlfunc/rename OLDNAME,NEWNAME> an open file. Don't 415L<C<tie>|perlfunc/tie VARIABLE,CLASSNAME,LIST> or 416L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> a file already tied or opened; 417L<C<untie>|perlfunc/untie VARIABLE> or 418L<C<close>|perlfunc/close FILEHANDLE> it first. 419 420Don't open the same file more than once at a time for writing, as some 421operating systems put mandatory locks on such files. 422 423Don't assume that write/modify permission on a directory gives the 424right to add or delete files/directories in that directory. That is 425filesystem specific: in some filesystems you need write/modify 426permission also (or even just) in the file/directory itself. In some 427filesystems (AFS, DFS) the permission to add/delete directory entries 428is a completely separate permission. 429 430Don't assume that a single L<C<unlink>|perlfunc/unlink LIST> completely 431gets rid of the file: some filesystems (most notably the ones in VMS) have 432versioned filesystems, and L<C<unlink>|perlfunc/unlink LIST> removes only 433the most recent one (it doesn't remove all the versions because by default 434the native tools on those platforms remove just the most recent version, 435too). The portable idiom to remove all the versions of a file is 436 437 1 while unlink "file"; 438 439This will terminate if the file is undeletable for some reason 440(protected, not there, and so on). 441 442Don't count on a specific environment variable existing in 443L<C<%ENV>|perlvar/%ENV>. Don't count on L<C<%ENV>|perlvar/%ENV> entries 444being case-sensitive, or even case-preserving. Don't try to clear 445L<C<%ENV>|perlvar/%ENV> by saying C<%ENV = ();>, or, if you really have 446to, make it conditional on C<$^O ne 'VMS'> since in VMS the 447L<C<%ENV>|perlvar/%ENV> table is much more than a per-process key-value 448string table. 449 450On VMS, some entries in the L<C<%ENV>|perlvar/%ENV> hash are dynamically 451created when their key is used on a read if they did not previously 452exist. The values for C<$ENV{HOME}>, C<$ENV{TERM}>, C<$ENV{PATH}>, and 453C<$ENV{USER}>, are known to be dynamically generated. The specific names 454that are dynamically generated may vary with the version of the C library 455on VMS, and more may exist than are documented. 456 457On VMS by default, changes to the L<C<%ENV>|perlvar/%ENV> hash persist 458after perl exits. Subsequent invocations of perl in the same process can 459inadvertently inherit environment settings that were meant to be 460temporary. 461 462Don't count on signals or L<C<%SIG>|perlvar/%SIG> for anything. 463 464Don't count on filename globbing. Use 465L<C<opendir>|perlfunc/opendir DIRHANDLE,EXPR>, 466L<C<readdir>|perlfunc/readdir DIRHANDLE>, and 467L<C<closedir>|perlfunc/closedir DIRHANDLE> instead. 468 469Don't count on per-program environment variables, or per-program current 470directories. 471 472Don't count on specific values of L<C<$!>|perlvar/$!>, neither numeric nor 473especially the string values. Users may switch their locales causing 474error messages to be translated into their languages. If you can 475trust a POSIXish environment, you can portably use the symbols defined 476by the L<C<Errno>|Errno> module, like C<ENOENT>. And don't trust on the 477values of L<C<$!>|perlvar/$!> at all except immediately after a failed 478system call. 479 480=head2 Command names versus file pathnames 481 482Don't assume that the name used to invoke a command or program with 483L<C<system>|perlfunc/system LIST> or L<C<exec>|perlfunc/exec LIST> can 484also be used to test for the existence of the file that holds the 485executable code for that command or program. 486First, many systems have "internal" commands that are built-in to the 487shell or OS and while these commands can be invoked, there is no 488corresponding file. Second, some operating systems (e.g., Cygwin, 489DJGPP, OS/2, and VOS) have required suffixes for executable files; 490these suffixes are generally permitted on the command name but are not 491required. Thus, a command like C<perl> might exist in a file named 492F<perl>, F<perl.exe>, or F<perl.pm>, depending on the operating system. 493The variable L<C<$Config{_exe}>|Config/C<_exe>> in the 494L<C<Config>|Config> module holds the executable suffix, if any. Third, 495the VMS port carefully sets up L<C<$^X>|perlvar/$^X> and 496L<C<$Config{perlpath}>|Config/C<perlpath>> so that no further processing 497is required. This is just as well, because the matching regular 498expression used below would then have to deal with a possible trailing 499version number in the VMS file name. 500 501To convert L<C<$^X>|perlvar/$^X> to a file pathname, taking account of 502the requirements of the various operating system possibilities, say: 503 504 use Config; 505 my $thisperl = $^X; 506 if ($^O ne 'VMS') { 507 $thisperl .= $Config{_exe} 508 unless $thisperl =~ m/\Q$Config{_exe}\E$/i; 509 } 510 511To convert L<C<$Config{perlpath}>|Config/C<perlpath>> to a file pathname, say: 512 513 use Config; 514 my $thisperl = $Config{perlpath}; 515 if ($^O ne 'VMS') { 516 $thisperl .= $Config{_exe} 517 unless $thisperl =~ m/\Q$Config{_exe}\E$/i; 518 } 519 520=head2 Networking 521 522Don't assume that you can reach the public Internet. 523 524Don't assume that there is only one way to get through firewalls 525to the public Internet. 526 527Don't assume that you can reach outside world through any other port 528than 80, or some web proxy. ftp is blocked by many firewalls. 529 530Don't assume that you can send email by connecting to the local SMTP port. 531 532Don't assume that you can reach yourself or any node by the name 533'localhost'. The same goes for '127.0.0.1'. You will have to try both. 534 535Don't assume that the host has only one network card, or that it 536can't bind to many virtual IP addresses. 537 538Don't assume a particular network device name. 539 540Don't assume a particular set of 541L<C<ioctl>|perlfunc/ioctl FILEHANDLE,FUNCTION,SCALAR>s will work. 542 543Don't assume that you can ping hosts and get replies. 544 545Don't assume that any particular port (service) will respond. 546 547Don't assume that L<C<Sys::Hostname>|Sys::Hostname> (or any other API or 548command) returns either a fully qualified hostname or a non-qualified 549hostname: it all depends on how the system had been configured. Also 550remember that for things such as DHCP and NAT, the hostname you get back 551might not be very useful. 552 553All the above I<don't>s may look daunting, and they are, but the key 554is to degrade gracefully if one cannot reach the particular network 555service one wants. Croaking or hanging do not look very professional. 556 557=head2 Interprocess Communication (IPC) 558 559In general, don't directly access the system in code meant to be 560portable. That means, no L<C<system>|perlfunc/system LIST>, 561L<C<exec>|perlfunc/exec LIST>, L<C<fork>|perlfunc/fork>, 562L<C<pipe>|perlfunc/pipe READHANDLE,WRITEHANDLE>, 563L<C<``> or C<qxE<sol>E<sol>>|perlop/C<qxE<sol>I<STRING>E<sol>>>, 564L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> with a C<|>, nor any of the other 565things that makes being a Perl hacker worth being. 566 567Commands that launch external processes are generally supported on 568most platforms (though many of them do not support any type of 569forking). The problem with using them arises from what you invoke 570them on. External tools are often named differently on different 571platforms, may not be available in the same location, might accept 572different arguments, can behave differently, and often present their 573results in a platform-dependent way. Thus, you should seldom depend 574on them to produce consistent results. (Then again, if you're calling 575C<netstat -a>, you probably don't expect it to run on both Unix and CP/M.) 576 577One especially common bit of Perl code is opening a pipe to B<sendmail>: 578 579 open(my $mail, '|-', '/usr/lib/sendmail -t') 580 or die "cannot fork sendmail: $!"; 581 582This is fine for systems programming when sendmail is known to be 583available. But it is not fine for many non-Unix systems, and even 584some Unix systems that may not have sendmail installed. If a portable 585solution is needed, see the various distributions on CPAN that deal 586with it. L<C<Mail::Mailer>|Mail::Mailer> and L<C<Mail::Send>|Mail::Send> 587in the C<MailTools> distribution are commonly used, and provide several 588mailing methods, including C<mail>, C<sendmail>, and direct SMTP (via 589L<C<Net::SMTP>|Net::SMTP>) if a mail transfer agent is not available. 590L<C<Mail::Sendmail>|Mail::Sendmail> is a standalone module that provides 591simple, platform-independent mailing. 592 593The Unix System V IPC (C<msg*(), sem*(), shm*()>) is not available 594even on all Unix platforms. 595 596Do not use either the bare result of C<pack("N", 10, 20, 30, 40)> or 597bare v-strings (such as C<v10.20.30.40>) to represent IPv4 addresses: 598both forms just pack the four bytes into network order. That this 599would be equal to the C language C<in_addr> struct (which is what the 600socket code internally uses) is not guaranteed. To be portable use 601the routines of the L<C<Socket>|Socket> module, such as 602L<C<inet_aton>|Socket/$ip_address = inet_aton $string>, 603L<C<inet_ntoa>|Socket/$string = inet_ntoa $ip_address>, and 604L<C<sockaddr_in>|Socket/$sockaddr = sockaddr_in $port, $ip_address>. 605 606The rule of thumb for portable code is: Do it all in portable Perl, or 607use a module (that may internally implement it with platform-specific 608code, but exposes a common interface). 609 610=head2 External Subroutines (XS) 611 612XS code can usually be made to work with any platform, but dependent 613libraries, header files, etc., might not be readily available or 614portable, or the XS code itself might be platform-specific, just as Perl 615code might be. If the libraries and headers are portable, then it is 616normally reasonable to make sure the XS code is portable, too. 617 618A different type of portability issue arises when writing XS code: 619availability of a C compiler on the end-user's system. C brings 620with it its own portability issues, and writing XS code will expose 621you to some of those. Writing purely in Perl is an easier way to 622achieve portability. 623 624=head2 Standard Modules 625 626In general, the standard modules work across platforms. Notable 627exceptions are the L<C<CPAN>|CPAN> module (which currently makes 628connections to external programs that may not be available), 629platform-specific modules (like L<C<ExtUtils::MM_VMS>|ExtUtils::MM_VMS>), 630and DBM modules. 631 632There is no one DBM module available on all platforms. 633L<C<SDBM_File>|SDBM_File> and the others are generally available on all 634Unix and DOSish ports, but not in MacPerl, where only 635L<C<NDBM_File>|NDBM_File> and L<C<DB_File>|DB_File> are available. 636 637The good news is that at least some DBM module should be available, and 638L<C<AnyDBM_File>|AnyDBM_File> will use whichever module it can find. Of 639course, then the code needs to be fairly strict, dropping to the greatest 640common factor (e.g., not exceeding 1K for each record), so that it will 641work with any DBM module. See L<AnyDBM_File> for more details. 642 643=head2 Time and Date 644 645The system's notion of time of day and calendar date is controlled in 646widely different ways. Don't assume the timezone is stored in C<$ENV{TZ}>, 647and even if it is, don't assume that you can control the timezone through 648that variable. Don't assume anything about the three-letter timezone 649abbreviations (for example that MST would be the Mountain Standard Time, 650it's been known to stand for Moscow Standard Time). If you need to 651use timezones, express them in some unambiguous format like the 652exact number of minutes offset from UTC, or the POSIX timezone 653format. 654 655Don't assume that the epoch starts at 00:00:00, January 1, 1970, 656because that is OS- and implementation-specific. It is better to 657store a date in an unambiguous representation. The ISO 8601 standard 658defines YYYY-MM-DD as the date format, or YYYY-MM-DDTHH:MM:SS 659(that's a literal "T" separating the date from the time). 660Please do use the ISO 8601 instead of making us guess what 661date 02/03/04 might be. ISO 8601 even sorts nicely as-is. 662A text representation (like "1987-12-18") can be easily converted 663into an OS-specific value using a module like 664L<C<Time::Piece>|Time::Piece> (see L<Time::Piece/Date Parsing>) or 665L<C<Date::Parse>|Date::Parse>. An array of values, such as those 666returned by L<C<localtime>|perlfunc/localtime EXPR>, can be converted to an OS-specific 667representation using L<C<Time::Local>|Time::Local>. 668 669When calculating specific times, such as for tests in time or date modules, 670it may be appropriate to calculate an offset for the epoch. 671 672 use Time::Local qw(timegm); 673 my $offset = timegm(0, 0, 0, 1, 0, 1970); 674 675The value for C<$offset> in Unix will be C<0>, but in Mac OS Classic 676will be some large number. C<$offset> can then be added to a Unix time 677value to get what should be the proper value on any system. 678 679=head2 Character sets and character encoding 680 681Assume very little about character sets. 682 683Assume nothing about numerical values (L<C<ord>|perlfunc/ord EXPR>, 684L<C<chr>|perlfunc/chr NUMBER>) of characters. 685Do not use explicit code point ranges (like C<\xHH-\xHH)>. However, 686starting in Perl v5.22, regular expression pattern bracketed character 687class ranges specified like C<qr/[\N{U+HH}-\N{U+HH}]/> are portable, 688and starting in Perl v5.24, the same ranges are portable in 689L<C<trE<sol>E<sol>E<sol>>|perlop/C<trE<sol>I<SEARCHLIST>E<sol>I<REPLACEMENTLIST>E<sol>cdsr>>. 690You can portably use symbolic character classes like C<[:print:]>. 691 692Do not assume that the alphabetic characters are encoded contiguously 693(in the numeric sense). There may be gaps. Special coding in Perl, 694however, guarantees that all subsets of C<qr/[A-Z]/>, C<qr/[a-z]/>, and 695C<qr/[0-9]/> behave as expected. 696L<C<trE<sol>E<sol>E<sol>>|perlop/C<trE<sol>I<SEARCHLIST>E<sol>I<REPLACEMENTLIST>E<sol>cdsr>> 697behaves the same for these ranges. In patterns, any ranges specified with 698end points using the C<\N{...}> notations ensures character set 699portability, but it is a bug in Perl v5.22 that this isn't true of 700L<C<trE<sol>E<sol>E<sol>>|perlop/C<trE<sol>I<SEARCHLIST>E<sol>I<REPLACEMENTLIST>E<sol>cdsr>>, 701fixed in v5.24. 702 703Do not assume anything about the ordering of the characters. 704The lowercase letters may come before or after the uppercase letters; 705the lowercase and uppercase may be interlaced so that both "a" and "A" 706come before "b"; the accented and other international characters may 707be interlaced so that E<auml> comes before "b". 708L<Unicode::Collate> can be used to sort this all out. 709 710=head2 Internationalisation 711 712If you may assume POSIX (a rather large assumption), you may read 713more about the POSIX locale system from L<perllocale>. The locale 714system at least attempts to make things a little bit more portable, 715or at least more convenient and native-friendly for non-English 716users. The system affects character sets and encoding, and date 717and time formatting--amongst other things. 718 719If you really want to be international, you should consider Unicode. 720See L<perluniintro> and L<perlunicode> for more information. 721 722By default Perl assumes your source code is written in an 8-bit ASCII 723superset. To embed Unicode characters in your strings and regexes, you can 724use the L<C<\x{HH}> or (more portably) C<\N{U+HH}> 725notations|perlop/Quote and Quote-like Operators>. You can also use the 726L<C<utf8>|utf8> pragma and write your code in UTF-8, which lets you use 727Unicode characters directly (not just in quoted constructs but also in 728identifiers). 729 730=head2 System Resources 731 732If your code is destined for systems with severely constrained (or 733missing!) virtual memory systems then you want to be I<especially> mindful 734of avoiding wasteful constructs such as: 735 736 my @lines = <$very_large_file>; # bad 737 738 while (<$fh>) {$file .= $_} # sometimes bad 739 my $file = join('', <$fh>); # better 740 741The last two constructs may appear unintuitive to most people. The 742first repeatedly grows a string, whereas the second allocates a 743large chunk of memory in one go. On some systems, the second is 744more efficient than the first. 745 746=head2 Security 747 748Most multi-user platforms provide basic levels of security, usually 749implemented at the filesystem level. Some, however, unfortunately do 750not. Thus the notion of user id, or "home" directory, 751or even the state of being logged-in, may be unrecognizable on many 752platforms. If you write programs that are security-conscious, it 753is usually best to know what type of system you will be running 754under so that you can write code explicitly for that platform (or 755class of platforms). 756 757Don't assume the Unix filesystem access semantics: the operating 758system or the filesystem may be using some ACL systems, which are 759richer languages than the usual C<rwx>. Even if the C<rwx> exist, 760their semantics might be different. 761 762(From the security viewpoint, testing for permissions before attempting to 763do something is silly anyway: if one tries this, there is potential 764for race conditions. Someone or something might change the 765permissions between the permissions check and the actual operation. 766Just try the operation.) 767 768Don't assume the Unix user and group semantics: especially, don't 769expect L<C<< $< >>|perlvar/$E<lt>> and L<C<< $> >>|perlvar/$E<gt>> (or 770L<C<$(>|perlvar/$(> and L<C<$)>|perlvar/$)>) to work for switching 771identities (or memberships). 772 773Don't assume set-uid and set-gid semantics. (And even if you do, 774think twice: set-uid and set-gid are a known can of security worms.) 775 776=head2 Style 777 778For those times when it is necessary to have platform-specific code, 779consider keeping the platform-specific code in one place, making porting 780to other platforms easier. Use the L<C<Config>|Config> module and the 781special variable L<C<$^O>|perlvar/$^O> to differentiate platforms, as 782described in L</"PLATFORMS">. 783 784Beware of the "else syndrome": 785 786 if ($^O eq 'MSWin32') { 787 # code that assumes Windows 788 } else { 789 # code that assumes Linux 790 } 791 792The C<else> branch should be used for the really ultimate fallback, 793not for code specific to some platform. 794 795Be careful in the tests you supply with your module or programs. 796Module code may be fully portable, but its tests might not be. This 797often happens when tests spawn off other processes or call external 798programs to aid in the testing, or when (as noted above) the tests 799assume certain things about the filesystem and paths. Be careful not 800to depend on a specific output style for errors, such as when checking 801L<C<$!>|perlvar/$!> after a failed system call. Using 802L<C<$!>|perlvar/$!> for anything else than displaying it as output is 803doubtful (though see the L<C<Errno>|Errno> module for testing reasonably 804portably for error value). Some platforms expect a certain output format, 805and Perl on those platforms may have been adjusted accordingly. Most 806specifically, don't anchor a regex when testing an error value. 807 808=head1 CPAN Testers 809 810Modules uploaded to CPAN are tested by a variety of volunteers on 811different platforms. These CPAN testers are notified by mail of each 812new upload, and reply to the list with PASS, FAIL, NA (not applicable to 813this platform), or UNKNOWN (unknown), along with any relevant notations. 814 815The purpose of the testing is twofold: one, to help developers fix any 816problems in their code that crop up because of lack of testing on other 817platforms; two, to provide users with information about whether 818a given module works on a given platform. 819 820Also see: 821 822=over 4 823 824=item * 825 826Mailing list: cpan-testers-discuss@perl.org 827 828=item * 829 830Testing results: L<https://www.cpantesters.org/> 831 832=back 833 834=head1 PLATFORMS 835 836Perl is built with a L<C<$^O>|perlvar/$^O> variable that indicates the 837operating system it was built on. This was implemented 838to help speed up code that would otherwise have to C<use Config> 839and use the value of L<C<$Config{osname}>|Config/C<osname>>. Of course, 840to get more detailed information about the system, looking into 841L<C<%Config>|Config/DESCRIPTION> is certainly recommended. 842 843L<C<%Config>|Config/DESCRIPTION> cannot always be trusted, however, 844because it was built at compile time. If perl was built in one place, 845then transferred elsewhere, some values may be wrong. The values may 846even have been edited after the fact. 847 848=head2 Unix 849 850Perl works on a bewildering variety of Unix and Unix-like platforms (see 851e.g. most of the files in the F<hints/> directory in the source code kit). 852On most of these systems, the value of L<C<$^O>|perlvar/$^O> (hence 853L<C<$Config{osname}>|Config/C<osname>>, too) is determined either by 854lowercasing and stripping punctuation from the first field of the string 855returned by typing C<uname -a> (or a similar command) at the shell prompt 856or by testing the file system for the presence of uniquely named files 857such as a kernel or header file. Here, for example, are a few of the 858more popular Unix flavors: 859 860 uname $^O $Config{archname} 861 -------------------------------------------- 862 AIX aix aix 863 BSD/OS bsdos i386-bsdos 864 Darwin darwin darwin 865 DYNIX/ptx dynixptx i386-dynixptx 866 FreeBSD freebsd freebsd-i386 867 Haiku haiku BePC-haiku 868 Linux linux arm-linux 869 Linux linux armv5tel-linux 870 Linux linux i386-linux 871 Linux linux i586-linux 872 Linux linux ppc-linux 873 HP-UX hpux PA-RISC1.1 874 IRIX irix irix 875 Mac OS X darwin darwin 876 NeXT 3 next next-fat 877 NeXT 4 next OPENSTEP-Mach 878 openbsd openbsd i386-openbsd 879 OSF1 dec_osf alpha-dec_osf 880 reliantunix-n svr4 RM400-svr4 881 SCO_SV sco_sv i386-sco_sv 882 SINIX-N svr4 RM400-svr4 883 sn4609 unicos CRAY_C90-unicos 884 sn6521 unicosmk t3e-unicosmk 885 sn9617 unicos CRAY_J90-unicos 886 SunOS solaris sun4-solaris 887 SunOS solaris i86pc-solaris 888 SunOS4 sunos sun4-sunos 889 890Because the value of L<C<$Config{archname}>|Config/C<archname>> may 891depend on the hardware architecture, it can vary more than the value of 892L<C<$^O>|perlvar/$^O>. 893 894=head2 DOS and Derivatives 895 896Perl has long been ported to Intel-style microcomputers running under 897systems like PC-DOS, MS-DOS, OS/2, and most Windows platforms you can 898bring yourself to mention (except for Windows CE, if you count that). 899Users familiar with I<COMMAND.COM> or I<CMD.EXE> style shells should 900be aware that each of these file specifications may have subtle 901differences: 902 903 my $filespec0 = "c:/foo/bar/file.txt"; 904 my $filespec1 = "c:\\foo\\bar\\file.txt"; 905 my $filespec2 = 'c:\foo\bar\file.txt'; 906 my $filespec3 = 'c:\\foo\\bar\\file.txt'; 907 908System calls accept either C</> or C<\> as the path separator. 909However, many command-line utilities of DOS vintage treat C</> as 910the option prefix, so may get confused by filenames containing C</>. 911Aside from calling any external programs, C</> will work just fine, 912and probably better, as it is more consistent with popular usage, 913and avoids the problem of remembering what to backwhack and what 914not to. 915 916The DOS FAT filesystem can accommodate only "8.3" style filenames. Under 917the "case-insensitive, but case-preserving" HPFS (OS/2) and NTFS (NT) 918filesystems you may have to be careful about case returned with functions 919like L<C<readdir>|perlfunc/readdir DIRHANDLE> or used with functions like 920L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> or 921L<C<opendir>|perlfunc/opendir DIRHANDLE,EXPR>. 922 923DOS also treats several filenames as special, such as F<AUX>, F<PRN>, 924F<NUL>, F<CON>, F<COM1>, F<LPT1>, F<LPT2>, etc. Unfortunately, sometimes 925these filenames won't even work if you include an explicit directory 926prefix. It is best to avoid such filenames, if you want your code to be 927portable to DOS and its derivatives. It's hard to know what these all 928are, unfortunately. 929 930Users of these operating systems may also wish to make use of 931scripts such as F<pl2bat.bat> to put wrappers around your scripts. 932 933Newline (C<\n>) is translated as C<\015\012> by the I/O system when 934reading from and writing to files (see L</"Newlines">). 935C<binmode($filehandle)> will keep C<\n> translated as C<\012> for that 936filehandle. 937L<C<binmode>|perlfunc/binmode FILEHANDLE> should always be used for code 938that deals with binary data. That's assuming you realize in advance that 939your data is in binary. General-purpose programs should often assume 940nothing about their data. 941 942The L<C<$^O>|perlvar/$^O> variable and the 943L<C<$Config{archname}>|Config/C<archname>> values for various DOSish 944perls are as follows: 945 946 OS $^O $Config{archname} ID Version 947 --------------------------------------------------------- 948 MS-DOS dos ? 949 PC-DOS dos ? 950 OS/2 os2 ? 951 Windows 3.1 ? ? 0 3 01 952 Windows 95 MSWin32 MSWin32-x86 1 4 00 953 Windows 98 MSWin32 MSWin32-x86 1 4 10 954 Windows ME MSWin32 MSWin32-x86 1 ? 955 Windows NT MSWin32 MSWin32-x86 2 4 xx 956 Windows NT MSWin32 MSWin32-ALPHA 2 4 xx 957 Windows NT MSWin32 MSWin32-ppc 2 4 xx 958 Windows 2000 MSWin32 MSWin32-x86 2 5 00 959 Windows XP MSWin32 MSWin32-x86 2 5 01 960 Windows 2003 MSWin32 MSWin32-x86 2 5 02 961 Windows Vista MSWin32 MSWin32-x86 2 6 00 962 Windows 7 MSWin32 MSWin32-x86 2 6 01 963 Windows 7 MSWin32 MSWin32-x64 2 6 01 964 Windows 2008 MSWin32 MSWin32-x86 2 6 01 965 Windows 2008 MSWin32 MSWin32-x64 2 6 01 966 Windows CE MSWin32 ? 3 967 Cygwin cygwin cygwin 968 969The various MSWin32 Perl's can distinguish the OS they are running on 970via the value of the fifth element of the list returned from 971L<C<Win32::GetOSVersion()>|Win32/Win32::GetOSVersion()>. For example: 972 973 if ($^O eq 'MSWin32') { 974 my @os_version_info = Win32::GetOSVersion(); 975 print +('3.1','95','NT')[$os_version_info[4]],"\n"; 976 } 977 978There are also C<Win32::IsWinNT()|Win32/Win32::IsWinNT()>, 979C<Win32::IsWin95()|Win32/Win32::IsWin95()>, and 980L<C<Win32::GetOSName()>|Win32/Win32::GetOSName()>; try 981L<C<perldoc Win32>|Win32>. 982The very portable L<C<POSIX::uname()>|POSIX/C<uname>> will work too: 983 984 c:\> perl -MPOSIX -we "print join '|', uname" 985 Windows NT|moonru|5.0|Build 2195 (Service Pack 2)|x86 986 987Errors set by Winsock functions are now put directly into C<$^E>, 988and the relevant C<WSAE*> error codes are now exported from the 989L<Errno> and L<POSIX> modules for testing this against. 990 991The previous behavior of putting the errors (converted to POSIX-style 992C<E*> error codes since Perl 5.20.0) into C<$!> was buggy due to 993the non-equivalence of like-named Winsock and POSIX error constants, 994a relationship between which has unfortunately been established 995in one way or another since Perl 5.8.0. 996 997The new behavior provides a much more robust solution for checking 998Winsock errors in portable software without accidentally matching 999POSIX tests that were intended for other OSes and may have different 1000meanings for Winsock. 1001 1002The old behavior is currently retained, warts and all, for backwards 1003compatibility, but users are encouraged to change any code that 1004tests C<$!> against C<E*> constants for Winsock errors to instead 1005test C<$^E> against C<WSAE*> constants. After a suitable deprecation 1006period, which started with Perl 5.24, the old behavior may be 1007removed, leaving C<$!> unchanged after Winsock function calls, to 1008avoid any possible confusion over which error variable to check. 1009 1010Also see: 1011 1012=over 4 1013 1014=item * 1015 1016The djgpp environment for DOS, L<http://www.delorie.com/djgpp/> 1017and L<perldos>. 1018 1019=item * 1020 1021The EMX environment for DOS, OS/2, etc. emx@iaehv.nl, 1022L<ftp://hobbes.nmsu.edu/pub/os2/dev/emx/> Also L<perlos2>. 1023 1024=item * 1025 1026Build instructions for Win32 in L<perlwin32>, or under the Cygnus environment 1027in L<perlcygwin>. 1028 1029=item * 1030 1031The C<Win32::*> modules in L<Win32>. 1032 1033=item * 1034 1035The ActiveState Pages, L<https://www.activestate.com/> 1036 1037=item * 1038 1039The Cygwin environment for Win32; F<README.cygwin> (installed 1040as L<perlcygwin>), L<https://www.cygwin.com/> 1041 1042=item * 1043 1044The U/WIN environment for Win32, 1045L<http://www.research.att.com/sw/tools/uwin/> 1046 1047=item * 1048 1049Build instructions for OS/2, L<perlos2> 1050 1051=back 1052 1053=head2 VMS 1054 1055Perl on VMS is discussed in L<perlvms> in the Perl distribution. 1056 1057The official name of VMS as of this writing is OpenVMS. 1058 1059Interacting with Perl from the Digital Command Language (DCL) shell 1060often requires a different set of quotation marks than Unix shells do. 1061For example: 1062 1063 $ perl -e "print ""Hello, world.\n""" 1064 Hello, world. 1065 1066There are several ways to wrap your Perl scripts in DCL F<.COM> files, if 1067you are so inclined. For example: 1068 1069 $ write sys$output "Hello from DCL!" 1070 $ if p1 .eqs. "" 1071 $ then perl -x 'f$environment("PROCEDURE") 1072 $ else perl -x - 'p1 'p2 'p3 'p4 'p5 'p6 'p7 'p8 1073 $ deck/dollars="__END__" 1074 #!/usr/bin/perl 1075 1076 print "Hello from Perl!\n"; 1077 1078 __END__ 1079 $ endif 1080 1081Do take care with C<$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT> if your 1082Perl-in-DCL script expects to do things like C<< $read = <STDIN>; >>. 1083 1084The VMS operating system has two filesystems, designated by their 1085on-disk structure (ODS) level: ODS-2 and its successor ODS-5. The 1086initial port of Perl to VMS pre-dates ODS-5, but all current testing and 1087development assumes ODS-5 and its capabilities, including case 1088preservation, extended characters in filespecs, and names up to 8192 1089bytes long. 1090 1091Perl on VMS can accept either VMS- or Unix-style file 1092specifications as in either of the following: 1093 1094 $ perl -ne "print if /perl_setup/i" SYS$LOGIN:LOGIN.COM 1095 $ perl -ne "print if /perl_setup/i" /sys$login/login.com 1096 1097but not a mixture of both as in: 1098 1099 $ perl -ne "print if /perl_setup/i" sys$login:/login.com 1100 Can't open sys$login:/login.com: file specification syntax error 1101 1102In general, the easiest path to portability is always to specify 1103filenames in Unix format unless they will need to be processed by native 1104commands or utilities. Because of this latter consideration, the 1105L<File::Spec> module by default returns native format specifications 1106regardless of input format. This default may be reversed so that 1107filenames are always reported in Unix format by specifying the 1108C<DECC$FILENAME_UNIX_REPORT> feature logical in the environment. 1109 1110The file type, or extension, is always present in a VMS-format file 1111specification even if it's zero-length. This means that, by default, 1112L<C<readdir>|perlfunc/readdir DIRHANDLE> will return a trailing dot on a 1113file with no extension, so where you would see C<"a"> on Unix you'll see 1114C<"a."> on VMS. However, the trailing dot may be suppressed by enabling 1115the C<DECC$READDIR_DROPDOTNOTYPE> feature in the environment (see the CRTL 1116documentation on feature logical names). 1117 1118What C<\n> represents depends on the type of file opened. It usually 1119represents C<\012> but it could also be C<\015>, C<\012>, C<\015\012>, 1120C<\000>, C<\040>, or nothing depending on the file organization and 1121record format. The L<C<VMS::Stdio>|VMS::Stdio> module provides access to 1122the special C<fopen()> requirements of files with unusual attributes on 1123VMS. 1124 1125The value of L<C<$^O>|perlvar/$^O> on OpenVMS is "VMS". To determine the 1126architecture that you are running on refer to 1127L<C<$Config{archname}>|Config/C<archname>>. 1128 1129On VMS, perl determines the UTC offset from the C<SYS$TIMEZONE_DIFFERENTIAL> 1130logical name. Although the VMS epoch began at 17-NOV-1858 00:00:00.00, 1131calls to L<C<localtime>|perlfunc/localtime EXPR> are adjusted to count 1132offsets from 01-JAN-1970 00:00:00.00, just like Unix. 1133 1134Also see: 1135 1136=over 4 1137 1138=item * 1139 1140F<README.vms> (installed as F<README_vms>), L<perlvms> 1141 1142=item * 1143 1144vmsperl list, vmsperl-subscribe@perl.org 1145 1146=item * 1147 1148vmsperl on the web, L<http://www.sidhe.org/vmsperl/index.html> 1149 1150=item * 1151 1152VMS Software Inc. web site, L<http://www.vmssoftware.com> 1153 1154=back 1155 1156=head2 VOS 1157 1158Perl on VOS (also known as OpenVOS) is discussed in F<README.vos> 1159in the Perl distribution (installed as L<perlvos>). Perl on VOS 1160can accept either VOS- or Unix-style file specifications as in 1161either of the following: 1162 1163 $ perl -ne "print if /perl_setup/i" >system>notices 1164 $ perl -ne "print if /perl_setup/i" /system/notices 1165 1166or even a mixture of both as in: 1167 1168 $ perl -ne "print if /perl_setup/i" >system/notices 1169 1170Even though VOS allows the slash character to appear in object 1171names, because the VOS port of Perl interprets it as a pathname 1172delimiting character, VOS files, directories, or links whose 1173names contain a slash character cannot be processed. Such files 1174must be renamed before they can be processed by Perl. 1175 1176Older releases of VOS (prior to OpenVOS Release 17.0) limit file 1177names to 32 or fewer characters, prohibit file names from 1178starting with a C<-> character, and prohibit file names from 1179containing C< > (space) or any character from the set C<< !#%&'()*;<=>? >>. 1180 1181Newer releases of VOS (OpenVOS Release 17.0 or later) support a 1182feature known as extended names. On these releases, file names 1183can contain up to 255 characters, are prohibited from starting 1184with a C<-> character, and the set of prohibited characters is 1185reduced to C<< #%*<>? >>. There are 1186restrictions involving spaces and apostrophes: these characters 1187must not begin or end a name, nor can they immediately precede or 1188follow a period. Additionally, a space must not immediately 1189precede another space or hyphen. Specifically, the following 1190character combinations are prohibited: space-space, 1191space-hyphen, period-space, space-period, period-apostrophe, 1192apostrophe-period, leading or trailing space, and leading or 1193trailing apostrophe. Although an extended file name is limited 1194to 255 characters, a path name is still limited to 256 1195characters. 1196 1197The value of L<C<$^O>|perlvar/$^O> on VOS is "vos". To determine the 1198architecture that you are running on refer to 1199L<C<$Config{archname}>|Config/C<archname>>. 1200 1201Also see: 1202 1203=over 4 1204 1205=item * 1206 1207F<README.vos> (installed as L<perlvos>) 1208 1209=item * 1210 1211The VOS mailing list. 1212 1213There is no specific mailing list for Perl on VOS. You can contact 1214the Stratus Technologies Customer Assistance Center (CAC) for your 1215region, or you can use the contact information located in the 1216distribution files on the Stratus Anonymous FTP site. 1217 1218=item * 1219 1220Stratus Technologies on the web at L<http://www.stratus.com> 1221 1222=item * 1223 1224VOS Open-Source Software on the web at L<http://ftp.stratus.com/pub/vos/vos.html> 1225 1226=back 1227 1228=head2 EBCDIC Platforms 1229 1230v5.22 core Perl runs on z/OS (formerly OS/390). Theoretically it could 1231run on the successors of OS/400 on AS/400 minicomputers as well as 1232VM/ESA, and BS2000 for S/390 Mainframes. Such computers use EBCDIC 1233character sets internally (usually Character Code Set ID 0037 for OS/400 1234and either 1047 or POSIX-BC for S/390 systems). 1235 1236The rest of this section may need updating, but we don't know what it 1237should say. Please submit comments to 1238L<https://github.com/Perl/perl5/issues>. 1239 1240On the mainframe Perl currently works under the "Unix system 1241services for OS/390" (formerly known as OpenEdition), VM/ESA OpenEdition, or 1242the BS200 POSIX-BC system (BS2000 is supported in Perl 5.6 and greater). 1243See L<perlos390> for details. Note that for OS/400 there is also a port of 1244Perl 5.8.1/5.10.0 or later to the PASE which is ASCII-based (as opposed to 1245ILE which is EBCDIC-based), see L<perlos400>. 1246 1247As of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix 1248sub-systems do not support the C<#!> shebang trick for script invocation. 1249Hence, on OS/390 and VM/ESA Perl scripts can be executed with a header 1250similar to the following simple script: 1251 1252 : # use perl 1253 eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}' 1254 if 0; 1255 #!/usr/local/bin/perl # just a comment really 1256 1257 print "Hello from perl!\n"; 1258 1259OS/390 will support the C<#!> shebang trick in release 2.8 and beyond. 1260Calls to L<C<system>|perlfunc/system LIST> and backticks can use POSIX 1261shell syntax on all S/390 systems. 1262 1263On the AS/400, if PERL5 is in your library list, you may need 1264to wrap your Perl scripts in a CL procedure to invoke them like so: 1265 1266 BEGIN 1267 CALL PGM(PERL5/PERL) PARM('/QOpenSys/hello.pl') 1268 ENDPGM 1269 1270This will invoke the Perl script F<hello.pl> in the root of the 1271QOpenSys file system. On the AS/400 calls to 1272L<C<system>|perlfunc/system LIST> or backticks must use CL syntax. 1273 1274On these platforms, bear in mind that the EBCDIC character set may have 1275an effect on what happens with some Perl functions (such as 1276L<C<chr>|perlfunc/chr NUMBER>, L<C<pack>|perlfunc/pack TEMPLATE,LIST>, 1277L<C<print>|perlfunc/print FILEHANDLE LIST>, 1278L<C<printf>|perlfunc/printf FILEHANDLE FORMAT, LIST>, 1279L<C<ord>|perlfunc/ord EXPR>, L<C<sort>|perlfunc/sort SUBNAME LIST>, 1280L<C<sprintf>|perlfunc/sprintf FORMAT, LIST>, 1281L<C<unpack>|perlfunc/unpack TEMPLATE,EXPR>), as 1282well as bit-fiddling with ASCII constants using operators like 1283L<C<^>, C<&> and C<|>|perlop/Bitwise String Operators>, not to mention 1284dealing with socket interfaces to ASCII computers (see L</"Newlines">). 1285 1286Fortunately, most web servers for the mainframe will correctly 1287translate the C<\n> in the following statement to its ASCII equivalent 1288(C<\r> is the same under both Unix and z/OS): 1289 1290 print "Content-type: text/html\r\n\r\n"; 1291 1292The values of L<C<$^O>|perlvar/$^O> on some of these platforms include: 1293 1294 uname $^O $Config{archname} 1295 -------------------------------------------- 1296 OS/390 os390 os390 1297 OS400 os400 os400 1298 POSIX-BC posix-bc BS2000-posix-bc 1299 1300Some simple tricks for determining if you are running on an EBCDIC 1301platform could include any of the following (perhaps all): 1302 1303 if ("\t" eq "\005") { print "EBCDIC may be spoken here!\n"; } 1304 1305 if (ord('A') == 193) { print "EBCDIC may be spoken here!\n"; } 1306 1307 if (chr(169) eq 'z') { print "EBCDIC may be spoken here!\n"; } 1308 1309One thing you may not want to rely on is the EBCDIC encoding 1310of punctuation characters since these may differ from code page to code 1311page (and once your module or script is rumoured to work with EBCDIC, 1312folks will want it to work with all EBCDIC character sets). 1313 1314Also see: 1315 1316=over 4 1317 1318=item * 1319 1320L<perlos390>, L<perlos400>, L<perlbs2000>, L<perlebcdic>. 1321 1322=item * 1323 1324The perl-mvs@perl.org list is for discussion of porting issues as well as 1325general usage issues for all EBCDIC Perls. Send a message body of 1326"subscribe perl-mvs" to majordomo@perl.org. 1327 1328=item * 1329 1330AS/400 Perl information at 1331L<http://as400.rochester.ibm.com/> 1332as well as on CPAN in the F<ports/> directory. 1333 1334=back 1335 1336=head2 Acorn RISC OS 1337 1338Because Acorns use ASCII with newlines (C<\n>) in text files as C<\012> like 1339Unix, and because Unix filename emulation is turned on by default, 1340most simple scripts will probably work "out of the box". The native 1341filesystem is modular, and individual filesystems are free to be 1342case-sensitive or insensitive, and are usually case-preserving. Some 1343native filesystems have name length limits, which file and directory 1344names are silently truncated to fit. Scripts should be aware that the 1345standard filesystem currently has a name length limit of B<10> 1346characters, with up to 77 items in a directory, but other filesystems 1347may not impose such limitations. 1348 1349Native filenames are of the form 1350 1351 Filesystem#Special_Field::DiskName.$.Directory.Directory.File 1352 1353where 1354 1355 Special_Field is not usually present, but may contain . and $ . 1356 Filesystem =~ m|[A-Za-z0-9_]| 1357 DsicName =~ m|[A-Za-z0-9_/]| 1358 $ represents the root directory 1359 . is the path separator 1360 @ is the current directory (per filesystem but machine global) 1361 ^ is the parent directory 1362 Directory and File =~ m|[^\0- "\.\$\%\&:\@\\^\|\177]+| 1363 1364The default filename translation is roughly C<tr|/.|./|>, swapping dots 1365and slashes. 1366 1367Note that C<"ADFS::HardDisk.$.File" ne 'ADFS::HardDisk.$.File'> and that 1368the second stage of C<$> interpolation in regular expressions will fall 1369foul of the L<C<$.>|perlvar/$.> variable if scripts are not careful. 1370 1371Logical paths specified by system variables containing comma-separated 1372search lists are also allowed; hence C<System:Modules> is a valid 1373filename, and the filesystem will prefix C<Modules> with each section of 1374C<System$Path> until a name is made that points to an object on disk. 1375Writing to a new file C<System:Modules> would be allowed only if 1376C<System$Path> contains a single item list. The filesystem will also 1377expand system variables in filenames if enclosed in angle brackets, so 1378C<< <System$Dir>.Modules >> would look for the file 1379S<C<$ENV{'System$Dir'} . 'Modules'>>. The obvious implication of this is 1380that B<fully qualified filenames can start with C<< <> >>> and the 1381three-argument form of L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> should 1382always be used. 1383 1384Because C<.> was in use as a directory separator and filenames could not 1385be assumed to be unique after 10 characters, Acorn implemented the C 1386compiler to strip the trailing C<.c> C<.h> C<.s> and C<.o> suffix from 1387filenames specified in source code and store the respective files in 1388subdirectories named after the suffix. Hence files are translated: 1389 1390 foo.h h.foo 1391 C:foo.h C:h.foo (logical path variable) 1392 sys/os.h sys.h.os (C compiler groks Unix-speak) 1393 10charname.c c.10charname 1394 10charname.o o.10charname 1395 11charname_.c c.11charname (assuming filesystem truncates at 10) 1396 1397The Unix emulation library's translation of filenames to native assumes 1398that this sort of translation is required, and it allows a user-defined list 1399of known suffixes that it will transpose in this fashion. This may 1400seem transparent, but consider that with these rules F<foo/bar/baz.h> 1401and F<foo/bar/h/baz> both map to F<foo.bar.h.baz>, and that 1402L<C<readdir>|perlfunc/readdir DIRHANDLE> and L<C<glob>|perlfunc/glob EXPR> 1403cannot and do not attempt to emulate the reverse mapping. Other 1404C<.>'s in filenames are translated to C</>. 1405 1406As implied above, the environment accessed through 1407L<C<%ENV>|perlvar/%ENV> is global, and the convention is that program 1408specific environment variables are of the form C<Program$Name>. 1409Each filesystem maintains a current directory, 1410and the current filesystem's current directory is the B<global> current 1411directory. Consequently, sociable programs don't change the current 1412directory but rely on full pathnames, and programs (and Makefiles) cannot 1413assume that they can spawn a child process which can change the current 1414directory without affecting its parent (and everyone else for that 1415matter). 1416 1417Because native operating system filehandles are global and are currently 1418allocated down from 255, with 0 being a reserved value, the Unix emulation 1419library emulates Unix filehandles. Consequently, you can't rely on 1420passing C<STDIN>, C<STDOUT>, or C<STDERR> to your children. 1421 1422The desire of users to express filenames of the form 1423C<< <Foo$Dir>.Bar >> on the command line unquoted causes problems, 1424too: L<C<``>|perlop/C<qxE<sol>I<STRING>E<sol>>> command output capture has 1425to perform a guessing game. It assumes that a string C<< <[^<>]+\$[^<>]> >> 1426is a reference to an environment variable, whereas anything else involving 1427C<< < >> or C<< > >> is redirection, and generally manages to be 99% 1428right. Of course, the problem remains that scripts cannot rely on any 1429Unix tools being available, or that any tools found have Unix-like command 1430line arguments. 1431 1432Extensions and XS are, in theory, buildable by anyone using free 1433tools. In practice, many don't, as users of the Acorn platform are 1434used to binary distributions. MakeMaker does run, but no available 1435make currently copes with MakeMaker's makefiles; even if and when 1436this should be fixed, the lack of a Unix-like shell will cause 1437problems with makefile rules, especially lines of the form 1438C<cd sdbm && make all>, and anything using quoting. 1439 1440S<"RISC OS"> is the proper name for the operating system, but the value 1441in L<C<$^O>|perlvar/$^O> is "riscos" (because we don't like shouting). 1442 1443=head2 Other perls 1444 1445Perl has been ported to many platforms that do not fit into any of 1446the categories listed above. Some, such as AmigaOS, 1447QNX, Plan 9, and VOS, have been well-integrated into the standard 1448Perl source code kit. You may need to see the F<ports/> directory 1449on CPAN for information, and possibly binaries, for the likes of: 1450aos, Atari ST, lynxos, riscos, Novell Netware, Tandem Guardian, 1451I<etc.> (Yes, we know that some of these OSes may fall under the 1452Unix category, but we are not a standards body.) 1453 1454Some approximate operating system names and their L<C<$^O>|perlvar/$^O> 1455values in the "OTHER" category include: 1456 1457 OS $^O $Config{archname} 1458 ------------------------------------------ 1459 Amiga DOS amigaos m68k-amigos 1460 1461See also: 1462 1463=over 4 1464 1465=item * 1466 1467Amiga, F<README.amiga> (installed as L<perlamiga>). 1468 1469=item * 1470 1471A free perl5-based PERL.NLM for Novell Netware is available in 1472precompiled binary and source code form from L<http://www.novell.com/> 1473as well as from CPAN. 1474 1475=item * 1476 1477S<Plan 9>, F<README.plan9> 1478 1479=back 1480 1481=head1 FUNCTION IMPLEMENTATIONS 1482 1483Listed below are functions that are either completely unimplemented 1484or else have been implemented differently on various platforms. 1485Preceding each description will be, in parentheses, a list of 1486platforms that the description applies to. 1487 1488The list may well be incomplete, or even wrong in some places. When 1489in doubt, consult the platform-specific README files in the Perl 1490source distribution, and any other documentation resources accompanying 1491a given port. 1492 1493Be aware, moreover, that even among Unix-ish systems there are variations. 1494 1495For many functions, you can also query L<C<%Config>|Config/DESCRIPTION>, 1496exported by default from the L<C<Config>|Config> module. For example, to 1497check whether the platform has the L<C<lstat>|perlfunc/lstat FILEHANDLE> 1498call, check L<C<$Config{d_lstat}>|Config/C<d_lstat>>. See L<Config> for a 1499full description of available variables. 1500 1501=head2 Alphabetical Listing of Perl Functions 1502 1503=over 8 1504 1505=item -X 1506 1507(Win32) 1508C<-w> only inspects the read-only file attribute (FILE_ATTRIBUTE_READONLY), 1509which determines whether the directory can be deleted, not whether it can 1510be written to. Directories always have read and write access unless denied 1511by discretionary access control lists (DACLs). 1512 1513(VMS) 1514C<-r>, C<-w>, C<-x>, and C<-o> tell whether the file is accessible, 1515which may not reflect UIC-based file protections. 1516 1517(S<RISC OS>) 1518C<-s> by name on an open file will return the space reserved on disk, 1519rather than the current extent. C<-s> on an open filehandle returns the 1520current size. 1521 1522(Win32, VMS, S<RISC OS>) 1523C<-R>, C<-W>, C<-X>, C<-O> are indistinguishable from C<-r>, C<-w>, 1524C<-x>, C<-o>. 1525 1526(Win32, VMS, S<RISC OS>) 1527C<-g>, C<-k>, C<-l>, C<-u>, C<-A> are not particularly meaningful. 1528 1529(Win32) 1530C<-l> returns true for both symlinks and directory junctions. 1531 1532(VMS, S<RISC OS>) 1533C<-p> is not particularly meaningful. 1534 1535(VMS) 1536C<-d> is true if passed a device spec without an explicit directory. 1537 1538(Win32) 1539C<-x> (or C<-X>) determine if a file ends in one of the executable 1540suffixes. C<-S> is meaningless. 1541 1542(S<RISC OS>) 1543C<-x> (or C<-X>) determine if a file has an executable file type. 1544 1545=item alarm 1546 1547(Win32) 1548Emulated using timers that must be explicitly polled whenever Perl 1549wants to dispatch "safe signals" and therefore cannot interrupt 1550blocking system calls. 1551 1552=item atan2 1553 1554(Tru64, HP-UX 10.20) 1555Due to issues with various CPUs, math libraries, compilers, and standards, 1556results for C<atan2> may vary depending on any combination of the above. 1557Perl attempts to conform to the Open Group/IEEE standards for the results 1558returned from C<atan2>, but cannot force the issue if the system Perl is 1559run on does not allow it. 1560 1561The current version of the standards for C<atan2> is available at 1562L<http://www.opengroup.org/onlinepubs/009695399/functions/atan2.html>. 1563 1564=item binmode 1565 1566(S<RISC OS>) 1567Meaningless. 1568 1569(VMS) 1570Reopens file and restores pointer; if function fails, underlying 1571filehandle may be closed, or pointer may be in a different position. 1572 1573(Win32) 1574The value returned by L<C<tell>|perlfunc/tell FILEHANDLE> may be affected 1575after the call, and the filehandle may be flushed. 1576 1577=item chdir 1578 1579(Win32) 1580The current directory reported by the system may include any symbolic 1581links specified to chdir(). 1582 1583=item chmod 1584 1585(Win32) 1586Only good for changing "owner" read-write access; "group" and "other" 1587bits are meaningless. 1588 1589(S<RISC OS>) 1590Only good for changing "owner" and "other" read-write access. 1591 1592(VOS) 1593Access permissions are mapped onto VOS access-control list changes. 1594 1595(Cygwin) 1596The actual permissions set depend on the value of the C<CYGWIN> variable 1597in the SYSTEM environment settings. 1598 1599(Android) 1600Setting the exec bit on some locations (generally F</sdcard>) will return true 1601but not actually set the bit. 1602 1603(VMS) 1604A mode argument of zero sets permissions to the user's default permission mask 1605rather than disabling all permissions. 1606 1607=item chown 1608 1609(S<Plan 9>, S<RISC OS>) 1610Not implemented. 1611 1612(Win32) 1613Does nothing, but won't fail. 1614 1615(VOS) 1616A little funky, because VOS's notion of ownership is a little funky. 1617 1618=item chroot 1619 1620(Win32, VMS, S<Plan 9>, S<RISC OS>, VOS) 1621Not implemented. 1622 1623=item crypt 1624 1625(Win32) 1626May not be available if library or source was not provided when building 1627perl. 1628 1629(Android) 1630Not implemented. 1631 1632=item dbmclose 1633 1634(VMS, S<Plan 9>, VOS) 1635Not implemented. 1636 1637=item dbmopen 1638 1639(VMS, S<Plan 9>, VOS) 1640Not implemented. 1641 1642=item dump 1643 1644(S<RISC OS>) 1645Not useful. 1646 1647(Cygwin, Win32) 1648Not supported. 1649 1650(VMS) 1651Invokes VMS debugger. 1652 1653=item exec 1654 1655(Win32) 1656C<exec LIST> without the use of indirect object syntax (C<exec PROGRAM LIST>) 1657may fall back to trying the shell if the first C<spawn()> fails. 1658 1659Note that the list form of exec() is emulated since the Win32 API 1660CreateProcess() accepts a simple string rather than an array of 1661command-line arguments. This may have security implications for your 1662code. 1663 1664(SunOS, Solaris, HP-UX) 1665Does not automatically flush output handles on some platforms. 1666 1667=item exit 1668 1669(VMS) 1670Emulates Unix C<exit> (which considers C<exit 1> to indicate an error) by 1671mapping the C<1> to C<SS$_ABORT> (C<44>). This behavior may be overridden 1672with the pragma L<C<use vmsish 'exit'>|vmsish/C<vmsish exit>>. As with 1673the CRTL's C<exit()> function, C<exit 0> is also mapped to an exit status 1674of C<SS$_NORMAL> (C<1>); this mapping cannot be overridden. Any other 1675argument to C<exit> 1676is used directly as Perl's exit status. On VMS, unless the future 1677POSIX_EXIT mode is enabled, the exit code should always be a valid 1678VMS exit code and not a generic number. When the POSIX_EXIT mode is 1679enabled, a generic number will be encoded in a method compatible with 1680the C library _POSIX_EXIT macro so that it can be decoded by other 1681programs, particularly ones written in C, like the GNV package. 1682 1683(Solaris) 1684C<exit> resets file pointers, which is a problem when called 1685from a child process (created by L<C<fork>|perlfunc/fork>) in 1686L<C<BEGIN>|perlmod/BEGIN, UNITCHECK, CHECK, INIT and END>. 1687A workaround is to use L<C<POSIX::_exit>|POSIX/C<_exit>>. 1688 1689 exit unless $Config{archname} =~ /\bsolaris\b/; 1690 require POSIX; 1691 POSIX::_exit(0); 1692 1693=item fcntl 1694 1695(Win32) 1696Not implemented. 1697 1698(VMS) 1699Some functions available based on the version of VMS. 1700 1701=item flock 1702 1703(VMS, S<RISC OS>, VOS) 1704Not implemented. 1705 1706=item fork 1707 1708(AmigaOS, S<RISC OS>, VMS) 1709Not implemented. 1710 1711(Win32) 1712Emulated using multiple interpreters. See L<perlfork>. 1713 1714(SunOS, Solaris, HP-UX) 1715Does not automatically flush output handles on some platforms. 1716 1717=item getlogin 1718 1719(S<RISC OS>) 1720Not implemented. 1721 1722=item getpgrp 1723 1724(Win32, VMS, S<RISC OS>) 1725Not implemented. 1726 1727=item getppid 1728 1729(Win32, S<RISC OS>) 1730Not implemented. 1731 1732=item getpriority 1733 1734(Win32, VMS, S<RISC OS>, VOS) 1735Not implemented. 1736 1737=item getpwnam 1738 1739(Win32) 1740Not implemented. 1741 1742(S<RISC OS>) 1743Not useful. 1744 1745=item getgrnam 1746 1747(Win32, VMS, S<RISC OS>) 1748Not implemented. 1749 1750=item getnetbyname 1751 1752(Android, Win32, S<Plan 9>) 1753Not implemented. 1754 1755=item getpwuid 1756 1757(Win32) 1758Not implemented. 1759 1760(S<RISC OS>) 1761Not useful. 1762 1763=item getgrgid 1764 1765(Win32, VMS, S<RISC OS>) 1766Not implemented. 1767 1768=item getnetbyaddr 1769 1770(Android, Win32, S<Plan 9>) 1771Not implemented. 1772 1773=item getprotobynumber 1774 1775(Android) 1776Not implemented. 1777 1778=item getpwent 1779 1780(Android, Win32) 1781Not implemented. 1782 1783=item getgrent 1784 1785(Android, Win32, VMS) 1786Not implemented. 1787 1788=item gethostbyname 1789 1790(S<Irix 5>) 1791C<gethostbyname('localhost')> does not work everywhere: you may have 1792to use C<gethostbyname('127.0.0.1')>. 1793 1794=item gethostent 1795 1796(Win32) 1797Not implemented. 1798 1799=item getnetent 1800 1801(Android, Win32, S<Plan 9>) 1802Not implemented. 1803 1804=item getprotoent 1805 1806(Android, Win32, S<Plan 9>) 1807Not implemented. 1808 1809=item getservent 1810 1811(Win32, S<Plan 9>) 1812Not implemented. 1813 1814=item seekdir 1815 1816(Android) 1817Not implemented. 1818 1819=item sethostent 1820 1821(Android, Win32, S<Plan 9>, S<RISC OS>) 1822Not implemented. 1823 1824=item setnetent 1825 1826(Win32, S<Plan 9>, S<RISC OS>) 1827Not implemented. 1828 1829=item setprotoent 1830 1831(Android, Win32, S<Plan 9>, S<RISC OS>) 1832Not implemented. 1833 1834=item setservent 1835 1836(S<Plan 9>, Win32, S<RISC OS>) 1837Not implemented. 1838 1839=item endpwent 1840 1841(Win32) 1842Not implemented. 1843 1844(Android) 1845Either not implemented or a no-op. 1846 1847=item endgrent 1848 1849(Android, S<RISC OS>, VMS, Win32) 1850Not implemented. 1851 1852=item endhostent 1853 1854(Android, Win32) 1855Not implemented. 1856 1857=item endnetent 1858 1859(Android, Win32, S<Plan 9>) 1860Not implemented. 1861 1862=item endprotoent 1863 1864(Android, Win32, S<Plan 9>) 1865Not implemented. 1866 1867=item endservent 1868 1869(S<Plan 9>, Win32) 1870Not implemented. 1871 1872=item getsockopt 1873 1874(S<Plan 9>) 1875Not implemented. 1876 1877=item glob 1878 1879This operator is implemented via the L<C<File::Glob>|File::Glob> extension 1880on most platforms. See L<File::Glob> for portability information. 1881 1882=item gmtime 1883 1884In theory, C<gmtime> is reliable from -2**63 to 2**63-1. However, 1885because work-arounds in the implementation use floating point numbers, 1886it will become inaccurate as the time gets larger. This is a bug and 1887will be fixed in the future. 1888 1889(VOS) 1890Time values are 32-bit quantities. 1891 1892=item ioctl 1893 1894(VMS) 1895Not implemented. 1896 1897(Win32) 1898Available only for socket handles, and it does what the C<ioctlsocket()> call 1899in the Winsock API does. 1900 1901(S<RISC OS>) 1902Available only for socket handles. 1903 1904=item kill 1905 1906(S<RISC OS>) 1907Not implemented, hence not useful for taint checking. 1908 1909(Win32) 1910C<kill> doesn't send a signal to the identified process like it does on 1911Unix platforms. Instead C<kill($sig, $pid)> terminates the process 1912identified by C<$pid>, and makes it exit immediately with exit status 1913C<$sig>. As in Unix, if C<$sig> is 0 and the specified process exists, it 1914returns true without actually terminating it. 1915 1916(Win32) 1917C<kill(-9, $pid)> will terminate the process specified by C<$pid> and 1918recursively all child processes owned by it. This is different from 1919the Unix semantics, where the signal will be delivered to all 1920processes in the same process group as the process specified by 1921C<$pid>. 1922 1923(VMS) 1924A pid of -1 indicating all processes on the system is not currently 1925supported. 1926 1927=item link 1928 1929(S<RISC OS>, VOS) 1930Not implemented. 1931 1932(AmigaOS) 1933Link count not updated because hard links are not quite that hard 1934(They are sort of half-way between hard and soft links). 1935 1936(Win32) 1937Hard links are implemented on Win32 under NTFS only. They are 1938natively supported on Windows 2000 and later. On Windows NT they 1939are implemented using the Windows POSIX subsystem support and the 1940Perl process will need Administrator or Backup Operator privileges 1941to create hard links. 1942 1943(VMS) 1944Available on 64 bit OpenVMS 8.2 and later. 1945 1946=item localtime 1947 1948C<localtime> has the same range as L</gmtime>, but because time zone 1949rules change, its accuracy for historical and future times may degrade 1950but usually by no more than an hour. 1951 1952=item lstat 1953 1954(S<RISC OS>) 1955Not implemented. 1956 1957(Win32) 1958Treats directory junctions as symlinks. 1959 1960=item msgctl 1961 1962=item msgget 1963 1964=item msgsnd 1965 1966=item msgrcv 1967 1968(Android, Win32, VMS, S<Plan 9>, S<RISC OS>, VOS) 1969Not implemented. 1970 1971=item open 1972 1973(S<RISC OS>) 1974Open modes C<|-> and C<-|> are unsupported. 1975 1976(SunOS, Solaris, HP-UX) 1977Opening a process does not automatically flush output handles on some 1978platforms. 1979 1980(Win32) 1981Both of modes C<|-> and C<-|> are supported, but the list form is 1982emulated since the Win32 API CreateProcess() accepts a simple string 1983rather than an array of arguments. This may have security 1984implications for your code. 1985 1986=item readlink 1987 1988(VMS, S<RISC OS>) 1989Not implemented. 1990 1991(Win32) 1992readlink() on a directory junction returns the object name, not a 1993simple path. 1994 1995=item rename 1996 1997(Win32) 1998Can't move directories between directories on different logical volumes. 1999 2000=item rewinddir 2001 2002(Win32) 2003Will not cause L<C<readdir>|perlfunc/readdir DIRHANDLE> to re-read the 2004directory stream. The entries already read before the C<rewinddir> call 2005will just be returned again from a cache buffer. 2006 2007=item select 2008 2009(Win32, VMS) 2010Only implemented on sockets. 2011 2012(S<RISC OS>) 2013Only reliable on sockets. 2014 2015Note that the L<C<select FILEHANDLE>|perlfunc/select FILEHANDLE> form is 2016generally portable. 2017 2018=item semctl 2019 2020=item semget 2021 2022=item semop 2023 2024(Android, Win32, VMS, S<RISC OS>) 2025Not implemented. 2026 2027=item setgrent 2028 2029(Android, VMS, Win32, S<RISC OS>) 2030Not implemented. 2031 2032=item setpgrp 2033 2034(Win32, VMS, S<RISC OS>, VOS) 2035Not implemented. 2036 2037=item setpriority 2038 2039(Win32, VMS, S<RISC OS>, VOS) 2040Not implemented. 2041 2042=item setpwent 2043 2044(Android, Win32, S<RISC OS>) 2045Not implemented. 2046 2047=item setsockopt 2048 2049(S<Plan 9>) 2050Not implemented. 2051 2052=item shmctl 2053 2054=item shmget 2055 2056=item shmread 2057 2058=item shmwrite 2059 2060(Android, Win32, VMS, S<RISC OS>) 2061Not implemented. 2062 2063=item sleep 2064 2065(Win32) 2066Emulated using synchronization functions such that it can be 2067interrupted by L<C<alarm>|perlfunc/alarm SECONDS>, and limited to a 2068maximum of 4294967 seconds, approximately 49 days. 2069 2070=item socketpair 2071 2072(S<RISC OS>) 2073Not implemented. 2074 2075(VMS) 2076Available on 64 bit OpenVMS 8.2 and later. 2077 2078=item stat 2079 2080Platforms that do not have C<rdev>, C<blksize>, or C<blocks> will return 2081these as C<''>, so numeric comparison or manipulation of these fields may 2082cause 'not numeric' warnings. 2083 2084(S<Mac OS X>) 2085C<ctime> not supported on UFS. 2086 2087(Win32) 2088C<ctime> is creation time instead of inode change time. 2089 2090(VMS) 2091C<dev> and C<ino> are not necessarily reliable. 2092 2093(S<RISC OS>) 2094C<mtime>, C<atime> and C<ctime> all return the last modification time. 2095C<dev> and C<ino> are not necessarily reliable. 2096 2097(OS/2) 2098C<dev>, C<rdev>, C<blksize>, and C<blocks> are not available. C<ino> is not 2099meaningful and will differ between stat calls on the same file. 2100 2101(Cygwin) 2102Some versions of cygwin when doing a C<stat("foo")> and not finding it 2103may then attempt to C<stat("foo.exe")>. 2104 2105=item symlink 2106 2107(S<RISC OS>) 2108Not implemented. 2109 2110(Win32) 2111Requires either elevated permissions or developer mode and a 2112sufficiently recent version of Windows 10. You can check whether the current 2113process has the required privileges using the 2114L<Win32::IsSymlinkCreationAllowed()|Win32/Win32::IsSymlinkCreationAllowed()> 2115function. 2116 2117Since Windows needs to know whether the target is a directory or not when 2118creating the link the target Perl will only create the link as a directory 2119link when the target exists and is a directory. 2120 2121(VMS) 2122Implemented on 64 bit VMS 8.3. VMS requires the symbolic link to be in Unix 2123syntax if it is intended to resolve to a valid path. 2124 2125=item syscall 2126 2127(Win32, VMS, S<RISC OS>, VOS) 2128Not implemented. 2129 2130=item sysopen 2131 2132(S<Mac OS>, OS/390) 2133The traditional C<0>, C<1>, and C<2> MODEs are implemented with different 2134numeric values on some systems. The flags exported by L<C<Fcntl>|Fcntl> 2135(C<O_RDONLY>, C<O_WRONLY>, C<O_RDWR>) should work everywhere though. 2136 2137=item system 2138 2139(Win32) 2140As an optimization, may not call the command shell specified in 2141C<$ENV{PERL5SHELL}>. C<system(1, @args)> spawns an external 2142process and immediately returns its process designator, without 2143waiting for it to terminate. Return value may be used subsequently 2144in L<C<wait>|perlfunc/wait> or L<C<waitpid>|perlfunc/waitpid PID,FLAGS>. 2145Failure to C<spawn()> a subprocess is indicated by setting 2146L<C<$?>|perlvar/$?> to C<<< 255 << 8 >>>. L<C<$?>|perlvar/$?> is set in a 2147way compatible with Unix (i.e. the exit status of the subprocess is 2148obtained by C<<< $? >> 8 >>>, as described in the documentation). 2149 2150Note that the list form of system() is emulated since the Win32 API 2151CreateProcess() accepts a simple string rather than an array of 2152command-line arguments. This may have security implications for your 2153code. 2154 2155(S<RISC OS>) 2156There is no shell to process metacharacters, and the native standard is 2157to pass a command line terminated by "\n" "\r" or "\0" to the spawned 2158program. Redirection such as C<< > foo >> is performed (if at all) by 2159the run time library of the spawned program. C<system LIST> will call 2160the Unix emulation library's L<C<exec>|perlfunc/exec LIST> emulation, 2161which attempts to provide emulation of the stdin, stdout, stderr in force 2162in the parent, provided the child program uses a compatible version of the 2163emulation library. C<system SCALAR> will call the native command line 2164directly and no such emulation of a child Unix program will occur. 2165Mileage B<will> vary. 2166 2167(Win32) 2168C<system LIST> without the use of indirect object syntax (C<system PROGRAM LIST>) 2169may fall back to trying the shell if the first C<spawn()> fails. 2170 2171(SunOS, Solaris, HP-UX) 2172Does not automatically flush output handles on some platforms. 2173 2174(VMS) 2175As with Win32, C<system(1, @args)> spawns an external process and 2176immediately returns its process designator without waiting for the 2177process to terminate. In this case the return value may be used subsequently 2178in L<C<wait>|perlfunc/wait> or L<C<waitpid>|perlfunc/waitpid PID,FLAGS>. 2179Otherwise the return value is POSIX-like (shifted up by 8 bits), which only 2180allows room for a made-up value derived from the severity bits of the native 218132-bit condition code (unless overridden by 2182L<C<use vmsish 'status'>|vmsish/C<vmsish status>>). If the native 2183condition code is one that has a POSIX value encoded, the POSIX value will 2184be decoded to extract the expected exit value. For more details see 2185L<perlvms/$?>. 2186 2187=item telldir 2188 2189(Android) 2190Not implemented. 2191 2192=item times 2193 2194(Win32) 2195"Cumulative" times will be bogus. On anything other than Windows NT 2196or Windows 2000, "system" time will be bogus, and "user" time is 2197actually the time returned by the L<C<clock()>|clock(3)> function in the C 2198runtime library. 2199 2200(S<RISC OS>) 2201Not useful. 2202 2203=item truncate 2204 2205(Older versions of VMS) 2206Not implemented. 2207 2208(VOS) 2209Truncation to same-or-shorter lengths only. 2210 2211(Win32) 2212If a FILEHANDLE is supplied, it must be writable and opened in append 2213mode (i.e., use C<<< open(my $fh, '>>', 'filename') >>> 2214or C<sysopen(my $fh, ..., O_APPEND|O_RDWR)>. If a filename is supplied, it 2215should not be held open elsewhere. 2216 2217=item umask 2218 2219Returns C<undef> where unavailable. 2220 2221(AmigaOS) 2222C<umask> works but the correct permissions are set only when the file 2223is finally closed. 2224 2225=item utime 2226 2227(VMS, S<RISC OS>) 2228Only the modification time is updated. 2229 2230(Win32) 2231May not behave as expected. Behavior depends on the C runtime 2232library's implementation of L<C<utime()>|utime(2)>, and the filesystem 2233being used. The FAT filesystem typically does not support an "access 2234time" field, and it may limit timestamps to a granularity of two seconds. 2235 2236=item wait 2237 2238=item waitpid 2239 2240(Win32) 2241Can only be applied to process handles returned for processes spawned 2242using C<system(1, ...)> or pseudo processes created with 2243L<C<fork>|perlfunc/fork>. 2244 2245(S<RISC OS>) 2246Not useful. 2247 2248=back 2249 2250 2251=head1 Supported Platforms 2252 2253The following platforms are known to build Perl 5.12 (as of April 2010, 2254its release date) from the standard source code distribution available 2255at L<http://www.cpan.org/src> 2256 2257=over 2258 2259=item Linux (x86, ARM, IA64) 2260 2261=item HP-UX 2262 2263=item AIX 2264 2265=item Win32 2266 2267=over 2268 2269=item Windows 2000 2270 2271=item Windows XP 2272 2273=item Windows Server 2003 2274 2275=item Windows Vista 2276 2277=item Windows Server 2008 2278 2279=item Windows 7 2280 2281=back 2282 2283=item Cygwin 2284 2285Some tests are known to fail: 2286 2287=over 2288 2289=item * 2290 2291F<ext/XS-APItest/t/call_checker.t> - see 2292L<https://github.com/Perl/perl5/issues/10750> 2293 2294=item * 2295 2296F<dist/I18N-Collate/t/I18N-Collate.t> 2297 2298=item * 2299 2300F<ext/Win32CORE/t/win32core.t> - may fail on recent cygwin installs. 2301 2302=back 2303 2304=item Solaris (x86, SPARC) 2305 2306=item OpenVMS 2307 2308=over 2309 2310=item Alpha (7.2 and later) 2311 2312=item I64 (8.2 and later) 2313 2314=back 2315 2316=item NetBSD 2317 2318=item FreeBSD 2319 2320=item Debian GNU/kFreeBSD 2321 2322=item Haiku 2323 2324=item Irix (6.5. What else?) 2325 2326=item OpenBSD 2327 2328=item Dragonfly BSD 2329 2330=item Midnight BSD 2331 2332=item QNX Neutrino RTOS (6.5.0) 2333 2334=item MirOS BSD 2335 2336=item Stratus OpenVOS (17.0 or later) 2337 2338Caveats: 2339 2340=over 2341 2342=item time_t issues that may or may not be fixed 2343 2344=back 2345 2346=item Stratus VOS / OpenVOS 2347 2348=item AIX 2349 2350=item Android 2351 2352=item FreeMINT 2353 2354Perl now builds with FreeMiNT/Atari. It fails a few tests, that needs 2355some investigation. 2356 2357The FreeMiNT port uses GNU dld for loadable module capabilities. So 2358ensure you have that library installed when building perl. 2359 2360=back 2361 2362=head1 EOL Platforms 2363 2364=head2 (Perl 5.20) 2365 2366The following platforms were supported by a previous version of 2367Perl but have been officially removed from Perl's source code 2368as of 5.20: 2369 2370=over 2371 2372=item AT&T 3b1 2373 2374=back 2375 2376=head2 (Perl 5.14) 2377 2378The following platforms were supported up to 5.10. They may still 2379have worked in 5.12, but supporting code has been removed for 5.14: 2380 2381=over 2382 2383=item Windows 95 2384 2385=item Windows 98 2386 2387=item Windows ME 2388 2389=item Windows NT4 2390 2391=back 2392 2393=head2 (Perl 5.12) 2394 2395The following platforms were supported by a previous version of 2396Perl but have been officially removed from Perl's source code 2397as of 5.12: 2398 2399=over 2400 2401=item Atari MiNT 2402 2403=item Apollo Domain/OS 2404 2405=item Apple Mac OS 8/9 2406 2407=item Tenon Machten 2408 2409=back 2410 2411 2412=head1 Supported Platforms (Perl 5.8) 2413 2414As of July 2002 (the Perl release 5.8.0), the following platforms were 2415able to build Perl from the standard source code distribution 2416available at L<http://www.cpan.org/src/> 2417 2418 AIX 2419 BeOS 2420 BSD/OS (BSDi) 2421 Cygwin 2422 DG/UX 2423 DOS DJGPP 1) 2424 DYNIX/ptx 2425 EPOC R5 2426 FreeBSD 2427 HI-UXMPP (Hitachi) (5.8.0 worked but we didn't know it) 2428 HP-UX 2429 IRIX 2430 Linux 2431 Mac OS Classic 2432 Mac OS X (Darwin) 2433 MPE/iX 2434 NetBSD 2435 NetWare 2436 NonStop-UX 2437 ReliantUNIX (formerly SINIX) 2438 OpenBSD 2439 OpenVMS (formerly VMS) 2440 Open UNIX (Unixware) (since Perl 5.8.1/5.9.0) 2441 OS/2 2442 OS/400 (using the PASE) (since Perl 5.8.1/5.9.0) 2443 POSIX-BC (formerly BS2000) 2444 QNX 2445 Solaris 2446 SunOS 4 2447 SUPER-UX (NEC) 2448 Tru64 UNIX (formerly DEC OSF/1, Digital UNIX) 2449 UNICOS 2450 UNICOS/mk 2451 UTS 2452 VOS / OpenVOS 2453 Win95/98/ME/2K/XP 2) 2454 WinCE 2455 z/OS (formerly OS/390) 2456 VM/ESA 2457 2458 1) in DOS mode either the DOS or OS/2 ports can be used 2459 2) compilers: Borland, MinGW (GCC), VC6 2460 2461The following platforms worked with the previous releases (5.6 and 24625.7), but we did not manage either to fix or to test these in time 2463for the 5.8.0 release. There is a very good chance that many of these 2464will work fine with the 5.8.0. 2465 2466 BSD/OS 2467 DomainOS 2468 Hurd 2469 LynxOS 2470 MachTen 2471 PowerMAX 2472 SCO SV 2473 SVR4 2474 Unixware 2475 Windows 3.1 2476 2477Known to be broken for 5.8.0 (but 5.6.1 and 5.7.2 can be used): 2478 2479 AmigaOS 3 2480 2481The following platforms have been known to build Perl from source in 2482the past (5.005_03 and earlier), but we haven't been able to verify 2483their status for the current release, either because the 2484hardware/software platforms are rare or because we don't have an 2485active champion on these platforms--or both. They used to work, 2486though, so go ahead and try compiling them, and let 2487L<https://github.com/Perl/perl5/issues> know 2488of any trouble. 2489 2490 3b1 2491 A/UX 2492 ConvexOS 2493 CX/UX 2494 DC/OSx 2495 DDE SMES 2496 DOS EMX 2497 Dynix 2498 EP/IX 2499 ESIX 2500 FPS 2501 GENIX 2502 Greenhills 2503 ISC 2504 MachTen 68k 2505 MPC 2506 NEWS-OS 2507 NextSTEP 2508 OpenSTEP 2509 Opus 2510 Plan 9 2511 RISC/os 2512 SCO ODT/OSR 2513 Stellar 2514 SVR2 2515 TI1500 2516 TitanOS 2517 Ultrix 2518 Unisys Dynix 2519 2520The following platforms have their own source code distributions and 2521binaries available via L<http://www.cpan.org/ports/> 2522 2523 Perl release 2524 2525 OS/400 (ILE) 5.005_02 2526 Tandem Guardian 5.004 2527 2528The following platforms have only binaries available via 2529L<http://www.cpan.org/ports/index.html> : 2530 2531 Perl release 2532 2533 Acorn RISCOS 5.005_02 2534 AOS 5.002 2535 LynxOS 5.004_02 2536 2537Although we do suggest that you always build your own Perl from 2538the source code, both for maximal configurability and for security, 2539in case you are in a hurry you can check 2540L<http://www.cpan.org/ports/index.html> for binary distributions. 2541 2542=head1 SEE ALSO 2543 2544L<perlaix>, L<perlamiga>, L<perlbs2000>, 2545L<perlcygwin>, L<perldos>, 2546L<perlebcdic>, L<perlfreebsd>, L<perlhurd>, L<perlhpux>, L<perlirix>, 2547L<perlmacos>, L<perlmacosx>, 2548L<perlnetware>, L<perlos2>, L<perlos390>, L<perlos400>, 2549L<perlplan9>, L<perlqnx>, L<perlsolaris>, L<perltru64>, 2550L<perlunicode>, L<perlvms>, L<perlvos>, L<perlwin32>, and L<Win32>. 2551 2552=head1 AUTHORS / CONTRIBUTORS 2553 2554Abigail <abigail@abigail.be>, 2555Charles Bailey <bailey@newman.upenn.edu>, 2556Graham Barr <gbarr@pobox.com>, 2557Tom Christiansen <tchrist@perl.com>, 2558Nicholas Clark <nick@ccl4.org>, 2559Thomas Dorner <Thomas.Dorner@start.de>, 2560Andy Dougherty <doughera@lafayette.edu>, 2561Dominic Dunlop <domo@computer.org>, 2562Neale Ferguson <neale@vma.tabnsw.com.au>, 2563David J. Fiander <davidf@mks.com>, 2564Paul Green <Paul.Green@stratus.com>, 2565M.J.T. Guy <mjtg@cam.ac.uk>, 2566Jarkko Hietaniemi <jhi@iki.fi>, 2567Luther Huffman <lutherh@stratcom.com>, 2568Nick Ing-Simmons <nick@ing-simmons.net>, 2569Andreas J. KE<ouml>nig <a.koenig@mind.de>, 2570Markus Laker <mlaker@contax.co.uk>, 2571Andrew M. Langmead <aml@world.std.com>, 2572Lukas Mai <l.mai@web.de>, 2573Larry Moore <ljmoore@freespace.net>, 2574Paul Moore <Paul.Moore@uk.origin-it.com>, 2575Chris Nandor <pudge@pobox.com>, 2576Matthias Neeracher <neeracher@mac.com>, 2577Philip Newton <pne@cpan.org>, 2578Gary Ng <71564.1743@CompuServe.COM>, 2579Tom Phoenix <rootbeer@teleport.com>, 2580AndrE<eacute> Pirard <A.Pirard@ulg.ac.be>, 2581Peter Prymmer <pvhp@forte.com>, 2582Hugo van der Sanden <hv@crypt0.demon.co.uk>, 2583Gurusamy Sarathy <gsar@activestate.com>, 2584Paul J. Schinder <schinder@pobox.com>, 2585Michael G Schwern <schwern@pobox.com>, 2586Dan Sugalski <dan@sidhe.org>, 2587Nathan Torkington <gnat@frii.com>, 2588John Malmberg <wb8tyw@qsl.net> 2589