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, 489OS/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 EMX environment for DOS, OS/2, etc. emx@iaehv.nl, 1017L<ftp://hobbes.nmsu.edu/pub/os2/dev/emx/> Also L<perlos2>. 1018 1019=item * 1020 1021Build instructions for Win32 in L<perlwin32>, or under the Cygnus environment 1022in L<perlcygwin>. 1023 1024=item * 1025 1026The C<Win32::*> modules in L<Win32>. 1027 1028=item * 1029 1030The ActiveState Pages, L<https://www.activestate.com/> 1031 1032=item * 1033 1034The Cygwin environment for Win32; F<README.cygwin> (installed 1035as L<perlcygwin>), L<https://www.cygwin.com/> 1036 1037=item * 1038 1039Build instructions for OS/2, L<perlos2> 1040 1041=back 1042 1043=head2 VMS 1044 1045Perl on VMS is discussed in L<perlvms> in the Perl distribution. 1046 1047The official name of VMS as of this writing is OpenVMS. 1048 1049Interacting with Perl from the Digital Command Language (DCL) shell 1050often requires a different set of quotation marks than Unix shells do. 1051For example: 1052 1053 $ perl -e "print ""Hello, world.\n""" 1054 Hello, world. 1055 1056There are several ways to wrap your Perl scripts in DCL F<.COM> files, if 1057you are so inclined. For example: 1058 1059 $ write sys$output "Hello from DCL!" 1060 $ if p1 .eqs. "" 1061 $ then perl -x 'f$environment("PROCEDURE") 1062 $ else perl -x - 'p1 'p2 'p3 'p4 'p5 'p6 'p7 'p8 1063 $ deck/dollars="__END__" 1064 #!/usr/bin/perl 1065 1066 print "Hello from Perl!\n"; 1067 1068 __END__ 1069 $ endif 1070 1071Do take care with C<$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT> if your 1072Perl-in-DCL script expects to do things like C<< $read = <STDIN>; >>. 1073 1074The VMS operating system has two filesystems, designated by their 1075on-disk structure (ODS) level: ODS-2 and its successor ODS-5. The 1076initial port of Perl to VMS pre-dates ODS-5, but all current testing and 1077development assumes ODS-5 and its capabilities, including case 1078preservation, extended characters in filespecs, and names up to 8192 1079bytes long. 1080 1081Perl on VMS can accept either VMS- or Unix-style file 1082specifications as in either of the following: 1083 1084 $ perl -ne "print if /perl_setup/i" SYS$LOGIN:LOGIN.COM 1085 $ perl -ne "print if /perl_setup/i" /sys$login/login.com 1086 1087but not a mixture of both as in: 1088 1089 $ perl -ne "print if /perl_setup/i" sys$login:/login.com 1090 Can't open sys$login:/login.com: file specification syntax error 1091 1092In general, the easiest path to portability is always to specify 1093filenames in Unix format unless they will need to be processed by native 1094commands or utilities. Because of this latter consideration, the 1095L<File::Spec> module by default returns native format specifications 1096regardless of input format. This default may be reversed so that 1097filenames are always reported in Unix format by specifying the 1098C<DECC$FILENAME_UNIX_REPORT> feature logical in the environment. 1099 1100The file type, or extension, is always present in a VMS-format file 1101specification even if it's zero-length. This means that, by default, 1102L<C<readdir>|perlfunc/readdir DIRHANDLE> will return a trailing dot on a 1103file with no extension, so where you would see C<"a"> on Unix you'll see 1104C<"a."> on VMS. However, the trailing dot may be suppressed by enabling 1105the C<DECC$READDIR_DROPDOTNOTYPE> feature in the environment (see the CRTL 1106documentation on feature logical names). 1107 1108What C<\n> represents depends on the type of file opened. It usually 1109represents C<\012> but it could also be C<\015>, C<\012>, C<\015\012>, 1110C<\000>, C<\040>, or nothing depending on the file organization and 1111record format. The L<C<VMS::Stdio>|VMS::Stdio> module provides access to 1112the special C<fopen()> requirements of files with unusual attributes on 1113VMS. 1114 1115The value of L<C<$^O>|perlvar/$^O> on OpenVMS is "VMS". To determine the 1116architecture that you are running on refer to 1117L<C<$Config{archname}>|Config/C<archname>>. 1118 1119On VMS, perl determines the UTC offset from the C<SYS$TIMEZONE_DIFFERENTIAL> 1120logical name. Although the VMS epoch began at 17-NOV-1858 00:00:00.00, 1121calls to L<C<localtime>|perlfunc/localtime EXPR> are adjusted to count 1122offsets from 01-JAN-1970 00:00:00.00, just like Unix. 1123 1124Also see: 1125 1126=over 4 1127 1128=item * 1129 1130F<README.vms> (installed as F<README_vms>), L<perlvms> 1131 1132=item * 1133 1134vmsperl list, vmsperl-subscribe@perl.org 1135 1136=item * 1137 1138vmsperl on the web, L<http://www.sidhe.org/vmsperl/index.html> 1139 1140=item * 1141 1142VMS Software Inc. web site, L<http://www.vmssoftware.com> 1143 1144=back 1145 1146=head2 VOS 1147 1148Perl on VOS (also known as OpenVOS) is discussed in F<README.vos> 1149in the Perl distribution (installed as L<perlvos>). Perl on VOS 1150can accept either VOS- or Unix-style file specifications as in 1151either of the following: 1152 1153 $ perl -ne "print if /perl_setup/i" >system>notices 1154 $ perl -ne "print if /perl_setup/i" /system/notices 1155 1156or even a mixture of both as in: 1157 1158 $ perl -ne "print if /perl_setup/i" >system/notices 1159 1160Even though VOS allows the slash character to appear in object 1161names, because the VOS port of Perl interprets it as a pathname 1162delimiting character, VOS files, directories, or links whose 1163names contain a slash character cannot be processed. Such files 1164must be renamed before they can be processed by Perl. 1165 1166Older releases of VOS (prior to OpenVOS Release 17.0) limit file 1167names to 32 or fewer characters, prohibit file names from 1168starting with a C<-> character, and prohibit file names from 1169containing C< > (space) or any character from the set C<< !#%&'()*;<=>? >>. 1170 1171Newer releases of VOS (OpenVOS Release 17.0 or later) support a 1172feature known as extended names. On these releases, file names 1173can contain up to 255 characters, are prohibited from starting 1174with a C<-> character, and the set of prohibited characters is 1175reduced to C<< #%*<>? >>. There are 1176restrictions involving spaces and apostrophes: these characters 1177must not begin or end a name, nor can they immediately precede or 1178follow a period. Additionally, a space must not immediately 1179precede another space or hyphen. Specifically, the following 1180character combinations are prohibited: space-space, 1181space-hyphen, period-space, space-period, period-apostrophe, 1182apostrophe-period, leading or trailing space, and leading or 1183trailing apostrophe. Although an extended file name is limited 1184to 255 characters, a path name is still limited to 256 1185characters. 1186 1187The value of L<C<$^O>|perlvar/$^O> on VOS is "vos". To determine the 1188architecture that you are running on refer to 1189L<C<$Config{archname}>|Config/C<archname>>. 1190 1191Also see: 1192 1193=over 4 1194 1195=item * 1196 1197F<README.vos> (installed as L<perlvos>) 1198 1199=item * 1200 1201The VOS mailing list. 1202 1203There is no specific mailing list for Perl on VOS. You can contact 1204the Stratus Technologies Customer Assistance Center (CAC) for your 1205region, or you can use the contact information located in the 1206distribution files on the Stratus Anonymous FTP site. 1207 1208=item * 1209 1210Stratus Technologies on the web at L<http://www.stratus.com> 1211 1212=item * 1213 1214VOS Open-Source Software on the web at L<http://ftp.stratus.com/pub/vos/vos.html> 1215 1216=back 1217 1218=head2 EBCDIC Platforms 1219 1220v5.22 core Perl runs on z/OS (formerly OS/390). Theoretically it could 1221run on the successors of OS/400 on AS/400 minicomputers as well as 1222VM/ESA, and BS2000 for S/390 Mainframes. Such computers use EBCDIC 1223character sets internally (usually Character Code Set ID 0037 for OS/400 1224and either 1047 or POSIX-BC for S/390 systems). 1225 1226The rest of this section may need updating, but we don't know what it 1227should say. Please submit comments to 1228L<https://github.com/Perl/perl5/issues>. 1229 1230On the mainframe Perl currently works under the "Unix system 1231services for OS/390" (formerly known as OpenEdition), VM/ESA OpenEdition, or 1232the BS200 POSIX-BC system (BS2000 is supported in Perl 5.6 and greater). 1233See L<perlos390> for details. Note that for OS/400 there is also a port of 1234Perl 5.8.1/5.10.0 or later to the PASE which is ASCII-based (as opposed to 1235ILE which is EBCDIC-based), see L<perlos400>. 1236 1237As of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix 1238sub-systems do not support the C<#!> shebang trick for script invocation. 1239Hence, on OS/390 and VM/ESA Perl scripts can be executed with a header 1240similar to the following simple script: 1241 1242 : # use perl 1243 eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}' 1244 if 0; 1245 #!/usr/local/bin/perl # just a comment really 1246 1247 print "Hello from perl!\n"; 1248 1249OS/390 will support the C<#!> shebang trick in release 2.8 and beyond. 1250Calls to L<C<system>|perlfunc/system LIST> and backticks can use POSIX 1251shell syntax on all S/390 systems. 1252 1253On the AS/400, if PERL5 is in your library list, you may need 1254to wrap your Perl scripts in a CL procedure to invoke them like so: 1255 1256 BEGIN 1257 CALL PGM(PERL5/PERL) PARM('/QOpenSys/hello.pl') 1258 ENDPGM 1259 1260This will invoke the Perl script F<hello.pl> in the root of the 1261QOpenSys file system. On the AS/400 calls to 1262L<C<system>|perlfunc/system LIST> or backticks must use CL syntax. 1263 1264On these platforms, bear in mind that the EBCDIC character set may have 1265an effect on what happens with some Perl functions (such as 1266L<C<chr>|perlfunc/chr NUMBER>, L<C<pack>|perlfunc/pack TEMPLATE,LIST>, 1267L<C<print>|perlfunc/print FILEHANDLE LIST>, 1268L<C<printf>|perlfunc/printf FILEHANDLE FORMAT, LIST>, 1269L<C<ord>|perlfunc/ord EXPR>, L<C<sort>|perlfunc/sort SUBNAME LIST>, 1270L<C<sprintf>|perlfunc/sprintf FORMAT, LIST>, 1271L<C<unpack>|perlfunc/unpack TEMPLATE,EXPR>), as 1272well as bit-fiddling with ASCII constants using operators like 1273L<C<^>, C<&> and C<|>|perlop/Bitwise String Operators>, not to mention 1274dealing with socket interfaces to ASCII computers (see L</"Newlines">). 1275 1276Fortunately, most web servers for the mainframe will correctly 1277translate the C<\n> in the following statement to its ASCII equivalent 1278(C<\r> is the same under both Unix and z/OS): 1279 1280 print "Content-type: text/html\r\n\r\n"; 1281 1282The values of L<C<$^O>|perlvar/$^O> on some of these platforms include: 1283 1284 uname $^O $Config{archname} 1285 -------------------------------------------- 1286 OS/390 os390 os390 1287 OS400 os400 os400 1288 POSIX-BC posix-bc BS2000-posix-bc 1289 1290Some simple tricks for determining if you are running on an EBCDIC 1291platform could include any of the following (perhaps all): 1292 1293 if ("\t" eq "\005") { print "EBCDIC may be spoken here!\n"; } 1294 1295 if (ord('A') == 193) { print "EBCDIC may be spoken here!\n"; } 1296 1297 if (chr(169) eq 'z') { print "EBCDIC may be spoken here!\n"; } 1298 1299One thing you may not want to rely on is the EBCDIC encoding 1300of punctuation characters since these may differ from code page to code 1301page (and once your module or script is rumoured to work with EBCDIC, 1302folks will want it to work with all EBCDIC character sets). 1303 1304Also see: 1305 1306=over 4 1307 1308=item * 1309 1310L<perlos390>, L<perlos400>, L<perlbs2000>, L<perlebcdic>. 1311 1312=item * 1313 1314The perl-mvs@perl.org list is for discussion of porting issues as well as 1315general usage issues for all EBCDIC Perls. Send a message body of 1316"subscribe perl-mvs" to majordomo@perl.org. 1317 1318=item * 1319 1320AS/400 Perl information at 1321L<http://as400.rochester.ibm.com/> 1322as well as on CPAN in the F<ports/> directory. 1323 1324=back 1325 1326=head2 Acorn RISC OS 1327 1328Because Acorns use ASCII with newlines (C<\n>) in text files as C<\012> like 1329Unix, and because Unix filename emulation is turned on by default, 1330most simple scripts will probably work "out of the box". The native 1331filesystem is modular, and individual filesystems are free to be 1332case-sensitive or insensitive, and are usually case-preserving. Some 1333native filesystems have name length limits, which file and directory 1334names are silently truncated to fit. Scripts should be aware that the 1335standard filesystem currently has a name length limit of B<10> 1336characters, with up to 77 items in a directory, but other filesystems 1337may not impose such limitations. 1338 1339Native filenames are of the form 1340 1341 Filesystem#Special_Field::DiskName.$.Directory.Directory.File 1342 1343where 1344 1345 Special_Field is not usually present, but may contain . and $ . 1346 Filesystem =~ m|[A-Za-z0-9_]| 1347 DsicName =~ m|[A-Za-z0-9_/]| 1348 $ represents the root directory 1349 . is the path separator 1350 @ is the current directory (per filesystem but machine global) 1351 ^ is the parent directory 1352 Directory and File =~ m|[^\0- "\.\$\%\&:\@\\^\|\177]+| 1353 1354The default filename translation is roughly C<tr|/.|./|>, swapping dots 1355and slashes. 1356 1357Note that C<"ADFS::HardDisk.$.File" ne 'ADFS::HardDisk.$.File'> and that 1358the second stage of C<$> interpolation in regular expressions will fall 1359foul of the L<C<$.>|perlvar/$.> variable if scripts are not careful. 1360 1361Logical paths specified by system variables containing comma-separated 1362search lists are also allowed; hence C<System:Modules> is a valid 1363filename, and the filesystem will prefix C<Modules> with each section of 1364C<System$Path> until a name is made that points to an object on disk. 1365Writing to a new file C<System:Modules> would be allowed only if 1366C<System$Path> contains a single item list. The filesystem will also 1367expand system variables in filenames if enclosed in angle brackets, so 1368C<< <System$Dir>.Modules >> would look for the file 1369S<C<$ENV{'System$Dir'} . 'Modules'>>. The obvious implication of this is 1370that B<fully qualified filenames can start with C<< <> >>> and the 1371three-argument form of L<C<open>|perlfunc/open FILEHANDLE,MODE,EXPR> should 1372always be used. 1373 1374Because C<.> was in use as a directory separator and filenames could not 1375be assumed to be unique after 10 characters, Acorn implemented the C 1376compiler to strip the trailing C<.c> C<.h> C<.s> and C<.o> suffix from 1377filenames specified in source code and store the respective files in 1378subdirectories named after the suffix. Hence files are translated: 1379 1380 foo.h h.foo 1381 C:foo.h C:h.foo (logical path variable) 1382 sys/os.h sys.h.os (C compiler groks Unix-speak) 1383 10charname.c c.10charname 1384 10charname.o o.10charname 1385 11charname_.c c.11charname (assuming filesystem truncates at 10) 1386 1387The Unix emulation library's translation of filenames to native assumes 1388that this sort of translation is required, and it allows a user-defined list 1389of known suffixes that it will transpose in this fashion. This may 1390seem transparent, but consider that with these rules F<foo/bar/baz.h> 1391and F<foo/bar/h/baz> both map to F<foo.bar.h.baz>, and that 1392L<C<readdir>|perlfunc/readdir DIRHANDLE> and L<C<glob>|perlfunc/glob EXPR> 1393cannot and do not attempt to emulate the reverse mapping. Other 1394C<.>'s in filenames are translated to C</>. 1395 1396As implied above, the environment accessed through 1397L<C<%ENV>|perlvar/%ENV> is global, and the convention is that program 1398specific environment variables are of the form C<Program$Name>. 1399Each filesystem maintains a current directory, 1400and the current filesystem's current directory is the B<global> current 1401directory. Consequently, sociable programs don't change the current 1402directory but rely on full pathnames, and programs (and Makefiles) cannot 1403assume that they can spawn a child process which can change the current 1404directory without affecting its parent (and everyone else for that 1405matter). 1406 1407Because native operating system filehandles are global and are currently 1408allocated down from 255, with 0 being a reserved value, the Unix emulation 1409library emulates Unix filehandles. Consequently, you can't rely on 1410passing C<STDIN>, C<STDOUT>, or C<STDERR> to your children. 1411 1412The desire of users to express filenames of the form 1413C<< <Foo$Dir>.Bar >> on the command line unquoted causes problems, 1414too: L<C<``>|perlop/C<qxE<sol>I<STRING>E<sol>>> command output capture has 1415to perform a guessing game. It assumes that a string C<< <[^<>]+\$[^<>]> >> 1416is a reference to an environment variable, whereas anything else involving 1417C<< < >> or C<< > >> is redirection, and generally manages to be 99% 1418right. Of course, the problem remains that scripts cannot rely on any 1419Unix tools being available, or that any tools found have Unix-like command 1420line arguments. 1421 1422Extensions and XS are, in theory, buildable by anyone using free 1423tools. In practice, many don't, as users of the Acorn platform are 1424used to binary distributions. MakeMaker does run, but no available 1425make currently copes with MakeMaker's makefiles; even if and when 1426this should be fixed, the lack of a Unix-like shell will cause 1427problems with makefile rules, especially lines of the form 1428C<cd sdbm && make all>, and anything using quoting. 1429 1430S<"RISC OS"> is the proper name for the operating system, but the value 1431in L<C<$^O>|perlvar/$^O> is "riscos" (because we don't like shouting). 1432 1433=head2 Other perls 1434 1435Perl has been ported to many platforms that do not fit into any of 1436the categories listed above. Some, such as AmigaOS, 1437QNX, Plan 9, and VOS, have been well-integrated into the standard 1438Perl source code kit. You may need to see the F<ports/> directory 1439on CPAN for information, and possibly binaries, for the likes of: 1440aos, Atari ST, lynxos, riscos, Novell Netware, Tandem Guardian, 1441I<etc.> (Yes, we know that some of these OSes may fall under the 1442Unix category, but we are not a standards body.) 1443 1444Some approximate operating system names and their L<C<$^O>|perlvar/$^O> 1445values in the "OTHER" category include: 1446 1447 OS $^O $Config{archname} 1448 ------------------------------------------ 1449 Amiga DOS amigaos m68k-amigos 1450 1451See also: 1452 1453=over 4 1454 1455=item * 1456 1457Amiga, F<README.amiga> (installed as L<perlamiga>). 1458 1459=item * 1460 1461S<Plan 9>, F<README.plan9> 1462 1463=back 1464 1465=head1 FUNCTION IMPLEMENTATIONS 1466 1467Listed below are functions that are either completely unimplemented 1468or else have been implemented differently on various platforms. 1469Preceding each description will be, in parentheses, a list of 1470platforms that the description applies to. 1471 1472The list may well be incomplete, or even wrong in some places. When 1473in doubt, consult the platform-specific README files in the Perl 1474source distribution, and any other documentation resources accompanying 1475a given port. 1476 1477Be aware, moreover, that even among Unix-ish systems there are variations. 1478 1479For many functions, you can also query L<C<%Config>|Config/DESCRIPTION>, 1480exported by default from the L<C<Config>|Config> module. For example, to 1481check whether the platform has the L<C<lstat>|perlfunc/lstat FILEHANDLE> 1482call, check L<C<$Config{d_lstat}>|Config/C<d_lstat>>. See L<Config> for a 1483full description of available variables. 1484 1485=head2 Alphabetical Listing of Perl Functions 1486 1487=over 8 1488 1489=item -X 1490 1491(Win32) 1492C<-w> only inspects the read-only file attribute (FILE_ATTRIBUTE_READONLY), 1493which determines whether the directory can be deleted, not whether it can 1494be written to. Directories always have read and write access unless denied 1495by discretionary access control lists (DACLs). 1496 1497(VMS) 1498C<-r>, C<-w>, C<-x>, and C<-o> tell whether the file is accessible, 1499which may not reflect UIC-based file protections. 1500 1501(S<RISC OS>) 1502C<-s> by name on an open file will return the space reserved on disk, 1503rather than the current extent. C<-s> on an open filehandle returns the 1504current size. 1505 1506(Win32, VMS, S<RISC OS>) 1507C<-R>, C<-W>, C<-X>, C<-O> are indistinguishable from C<-r>, C<-w>, 1508C<-x>, C<-o>. 1509 1510(Win32, VMS, S<RISC OS>) 1511C<-g>, C<-k>, C<-l>, C<-u>, C<-A> are not particularly meaningful. 1512 1513(Win32) 1514C<-l> returns true for both symlinks and directory junctions. 1515 1516(VMS, S<RISC OS>) 1517C<-p> is not particularly meaningful. 1518 1519(VMS) 1520C<-d> is true if passed a device spec without an explicit directory. 1521 1522(Win32) 1523C<-x> (or C<-X>) determine if a file ends in one of the executable 1524suffixes. C<-S> is meaningless. 1525 1526(S<RISC OS>) 1527C<-x> (or C<-X>) determine if a file has an executable file type. 1528 1529=item alarm 1530 1531(Win32) 1532Emulated using timers that must be explicitly polled whenever Perl 1533wants to dispatch "safe signals" and therefore cannot interrupt 1534blocking system calls. 1535 1536=item atan2 1537 1538(Tru64, HP-UX 10.20) 1539Due to issues with various CPUs, math libraries, compilers, and standards, 1540results for C<atan2> may vary depending on any combination of the above. 1541Perl attempts to conform to the Open Group/IEEE standards for the results 1542returned from C<atan2>, but cannot force the issue if the system Perl is 1543run on does not allow it. 1544 1545The current version of the standards for C<atan2> is available at 1546L<http://www.opengroup.org/onlinepubs/009695399/functions/atan2.html>. 1547 1548=item binmode 1549 1550(S<RISC OS>) 1551Meaningless. 1552 1553(VMS) 1554Reopens file and restores pointer; if function fails, underlying 1555filehandle may be closed, or pointer may be in a different position. 1556 1557(Win32) 1558The value returned by L<C<tell>|perlfunc/tell FILEHANDLE> may be affected 1559after the call, and the filehandle may be flushed. 1560 1561=item chdir 1562 1563(Win32) 1564The current directory reported by the system may include any symbolic 1565links specified to chdir(). 1566 1567=item chmod 1568 1569(Win32) 1570Only good for changing "owner" read-write access; "group" and "other" 1571bits are meaningless. 1572 1573(S<RISC OS>) 1574Only good for changing "owner" and "other" read-write access. 1575 1576(VOS) 1577Access permissions are mapped onto VOS access-control list changes. 1578 1579(Cygwin) 1580The actual permissions set depend on the value of the C<CYGWIN> variable 1581in the SYSTEM environment settings. 1582 1583(Android) 1584Setting the exec bit on some locations (generally F</sdcard>) will return true 1585but not actually set the bit. 1586 1587(VMS) 1588A mode argument of zero sets permissions to the user's default permission mask 1589rather than disabling all permissions. 1590 1591=item chown 1592 1593(S<Plan 9>, S<RISC OS>) 1594Not implemented. 1595 1596(Win32) 1597Does nothing, but won't fail. 1598 1599(VOS) 1600A little funky, because VOS's notion of ownership is a little funky. 1601 1602=item chroot 1603 1604(Win32, VMS, S<Plan 9>, S<RISC OS>, VOS) 1605Not implemented. 1606 1607=item crypt 1608 1609(Win32) 1610May not be available if library or source was not provided when building 1611perl. 1612 1613(Android) 1614Not implemented. 1615 1616=item dbmclose 1617 1618(VMS, S<Plan 9>, VOS) 1619Not implemented. 1620 1621=item dbmopen 1622 1623(VMS, S<Plan 9>, VOS) 1624Not implemented. 1625 1626=item dump 1627 1628(S<RISC OS>) 1629Not useful. 1630 1631(Cygwin, Win32) 1632Not supported. 1633 1634(VMS) 1635Invokes VMS debugger. 1636 1637=item exec 1638 1639(Win32) 1640C<exec LIST> without the use of indirect object syntax (C<exec PROGRAM LIST>) 1641may fall back to trying the shell if the first C<spawn()> fails. 1642 1643Note that the list form of exec() is emulated since the Win32 API 1644CreateProcess() accepts a simple string rather than an array of 1645command-line arguments. This may have security implications for your 1646code. 1647 1648(SunOS, Solaris, HP-UX) 1649Does not automatically flush output handles on some platforms. 1650 1651=item exit 1652 1653(VMS) 1654Emulates Unix C<exit> (which considers C<exit 1> to indicate an error) by 1655mapping the C<1> to C<SS$_ABORT> (C<44>). This behavior may be overridden 1656with the pragma L<C<use vmsish 'exit'>|vmsish/C<vmsish exit>>. As with 1657the CRTL's C<exit()> function, C<exit 0> is also mapped to an exit status 1658of C<SS$_NORMAL> (C<1>); this mapping cannot be overridden. Any other 1659argument to C<exit> 1660is used directly as Perl's exit status. On VMS, unless the future 1661POSIX_EXIT mode is enabled, the exit code should always be a valid 1662VMS exit code and not a generic number. When the POSIX_EXIT mode is 1663enabled, a generic number will be encoded in a method compatible with 1664the C library _POSIX_EXIT macro so that it can be decoded by other 1665programs, particularly ones written in C, like the GNV package. 1666 1667(Solaris) 1668C<exit> resets file pointers, which is a problem when called 1669from a child process (created by L<C<fork>|perlfunc/fork>) in 1670L<C<BEGIN>|perlmod/BEGIN, UNITCHECK, CHECK, INIT and END>. 1671A workaround is to use L<C<POSIX::_exit>|POSIX/C<_exit>>. 1672 1673 exit unless $Config{archname} =~ /\bsolaris\b/; 1674 require POSIX; 1675 POSIX::_exit(0); 1676 1677=item fcntl 1678 1679(Win32) 1680Not implemented. 1681 1682(VMS) 1683Some functions available based on the version of VMS. 1684 1685=item flock 1686 1687(VMS, S<RISC OS>, VOS) 1688Not implemented. 1689 1690=item fork 1691 1692(AmigaOS, S<RISC OS>, VMS) 1693Not implemented. 1694 1695(Win32) 1696Emulated using multiple interpreters. See L<perlfork>. 1697 1698(SunOS, Solaris, HP-UX) 1699Does not automatically flush output handles on some platforms. 1700 1701=item getlogin 1702 1703(S<RISC OS>) 1704Not implemented. 1705 1706=item getpgrp 1707 1708(Win32, VMS, S<RISC OS>) 1709Not implemented. 1710 1711=item getppid 1712 1713(Win32, S<RISC OS>) 1714Not implemented. 1715 1716=item getpriority 1717 1718(Win32, VMS, S<RISC OS>, VOS) 1719Not implemented. 1720 1721=item getpwnam 1722 1723(Win32) 1724Not implemented. 1725 1726(S<RISC OS>) 1727Not useful. 1728 1729=item getgrnam 1730 1731(Win32, VMS, S<RISC OS>) 1732Not implemented. 1733 1734=item getnetbyname 1735 1736(Android, Win32, S<Plan 9>) 1737Not implemented. 1738 1739=item getpwuid 1740 1741(Win32) 1742Not implemented. 1743 1744(S<RISC OS>) 1745Not useful. 1746 1747=item getgrgid 1748 1749(Win32, VMS, S<RISC OS>) 1750Not implemented. 1751 1752=item getnetbyaddr 1753 1754(Android, Win32, S<Plan 9>) 1755Not implemented. 1756 1757=item getprotobynumber 1758 1759(Android) 1760Not implemented. 1761 1762=item getpwent 1763 1764(Android, Win32) 1765Not implemented. 1766 1767=item getgrent 1768 1769(Android, Win32, VMS) 1770Not implemented. 1771 1772=item gethostbyname 1773 1774(S<Irix 5>) 1775C<gethostbyname('localhost')> does not work everywhere: you may have 1776to use C<gethostbyname('127.0.0.1')>. 1777 1778=item gethostent 1779 1780(Win32) 1781Not implemented. 1782 1783=item getnetent 1784 1785(Android, Win32, S<Plan 9>) 1786Not implemented. 1787 1788=item getprotoent 1789 1790(Android, Win32, S<Plan 9>) 1791Not implemented. 1792 1793=item getservent 1794 1795(Win32, S<Plan 9>) 1796Not implemented. 1797 1798=item seekdir 1799 1800(Android) 1801Not implemented. 1802 1803=item sethostent 1804 1805(Android, Win32, S<Plan 9>, S<RISC OS>) 1806Not implemented. 1807 1808=item setnetent 1809 1810(Win32, S<Plan 9>, S<RISC OS>) 1811Not implemented. 1812 1813=item setprotoent 1814 1815(Android, Win32, S<Plan 9>, S<RISC OS>) 1816Not implemented. 1817 1818=item setservent 1819 1820(S<Plan 9>, Win32, S<RISC OS>) 1821Not implemented. 1822 1823=item endpwent 1824 1825(Win32) 1826Not implemented. 1827 1828(Android) 1829Either not implemented or a no-op. 1830 1831=item endgrent 1832 1833(Android, S<RISC OS>, VMS, Win32) 1834Not implemented. 1835 1836=item endhostent 1837 1838(Android, Win32) 1839Not implemented. 1840 1841=item endnetent 1842 1843(Android, Win32, S<Plan 9>) 1844Not implemented. 1845 1846=item endprotoent 1847 1848(Android, Win32, S<Plan 9>) 1849Not implemented. 1850 1851=item endservent 1852 1853(S<Plan 9>, Win32) 1854Not implemented. 1855 1856=item getsockopt 1857 1858(S<Plan 9>) 1859Not implemented. 1860 1861=item glob 1862 1863This operator is implemented via the L<C<File::Glob>|File::Glob> extension 1864on most platforms. See L<File::Glob> for portability information. 1865 1866=item gmtime 1867 1868In theory, C<gmtime> is reliable from -2**63 to 2**63-1. However, 1869because work-arounds in the implementation use floating point numbers, 1870it will become inaccurate as the time gets larger. This is a bug and 1871will be fixed in the future. 1872 1873(VOS) 1874Time values are 32-bit quantities. 1875 1876=item ioctl 1877 1878(VMS) 1879Not implemented. 1880 1881(Win32) 1882Available only for socket handles, and it does what the C<ioctlsocket()> call 1883in the Winsock API does. 1884 1885(S<RISC OS>) 1886Available only for socket handles. 1887 1888=item kill 1889 1890(S<RISC OS>) 1891Not implemented, hence not useful for taint checking. 1892 1893(Win32) 1894C<kill> doesn't send a signal to the identified process like it does on 1895Unix platforms. Instead C<kill($sig, $pid)> terminates the process 1896identified by C<$pid>, and makes it exit immediately with exit status 1897C<$sig>. As in Unix, if C<$sig> is 0 and the specified process exists, it 1898returns true without actually terminating it. 1899 1900(Win32) 1901C<kill(-9, $pid)> will terminate the process specified by C<$pid> and 1902recursively all child processes owned by it. This is different from 1903the Unix semantics, where the signal will be delivered to all 1904processes in the same process group as the process specified by 1905C<$pid>. 1906 1907(VMS) 1908A pid of -1 indicating all processes on the system is not currently 1909supported. 1910 1911=item link 1912 1913(S<RISC OS>, VOS) 1914Not implemented. 1915 1916(AmigaOS) 1917Link count not updated because hard links are not quite that hard 1918(They are sort of half-way between hard and soft links). 1919 1920(Win32) 1921Hard links are implemented on Win32 under NTFS only. They are 1922natively supported on Windows 2000 and later. On Windows NT they 1923are implemented using the Windows POSIX subsystem support and the 1924Perl process will need Administrator or Backup Operator privileges 1925to create hard links. 1926 1927(VMS) 1928Available on 64 bit OpenVMS 8.2 and later. 1929 1930=item localtime 1931 1932C<localtime> has the same range as L</gmtime>, but because time zone 1933rules change, its accuracy for historical and future times may degrade 1934but usually by no more than an hour. 1935 1936=item lstat 1937 1938(S<RISC OS>) 1939Not implemented. 1940 1941(Win32) 1942Treats directory junctions as symlinks. 1943 1944=item msgctl 1945 1946=item msgget 1947 1948=item msgsnd 1949 1950=item msgrcv 1951 1952(Android, Win32, VMS, S<Plan 9>, S<RISC OS>, VOS) 1953Not implemented. 1954 1955=item open 1956 1957(S<RISC OS>) 1958Open modes C<|-> and C<-|> are unsupported. 1959 1960(SunOS, Solaris, HP-UX) 1961Opening a process does not automatically flush output handles on some 1962platforms. 1963 1964(Win32) 1965Both of modes C<|-> and C<-|> are supported, but the list form is 1966emulated since the Win32 API CreateProcess() accepts a simple string 1967rather than an array of arguments. This may have security 1968implications for your code. 1969 1970=item readlink 1971 1972(VMS, S<RISC OS>) 1973Not implemented. 1974 1975(Win32) 1976readlink() on a directory junction returns the object name, not a 1977simple path. 1978 1979=item rename 1980 1981(Win32) 1982Can't move directories between directories on different logical volumes. 1983 1984=item rewinddir 1985 1986(Win32) 1987Will not cause L<C<readdir>|perlfunc/readdir DIRHANDLE> to re-read the 1988directory stream. The entries already read before the C<rewinddir> call 1989will just be returned again from a cache buffer. 1990 1991=item select 1992 1993(Win32, VMS) 1994Only implemented on sockets. 1995 1996(S<RISC OS>) 1997Only reliable on sockets. 1998 1999Note that the L<C<select FILEHANDLE>|perlfunc/select FILEHANDLE> form is 2000generally portable. 2001 2002=item semctl 2003 2004=item semget 2005 2006=item semop 2007 2008(Android, Win32, VMS, S<RISC OS>) 2009Not implemented. 2010 2011=item setgrent 2012 2013(Android, VMS, Win32, S<RISC OS>) 2014Not implemented. 2015 2016=item setpgrp 2017 2018(Win32, VMS, S<RISC OS>, VOS) 2019Not implemented. 2020 2021=item setpriority 2022 2023(Win32, VMS, S<RISC OS>, VOS) 2024Not implemented. 2025 2026=item setpwent 2027 2028(Android, Win32, S<RISC OS>) 2029Not implemented. 2030 2031=item setsockopt 2032 2033(S<Plan 9>) 2034Not implemented. 2035 2036=item shmctl 2037 2038=item shmget 2039 2040=item shmread 2041 2042=item shmwrite 2043 2044(Android, Win32, VMS, S<RISC OS>) 2045Not implemented. 2046 2047=item sleep 2048 2049(Win32) 2050Emulated using synchronization functions such that it can be 2051interrupted by L<C<alarm>|perlfunc/alarm SECONDS>, and limited to a 2052maximum of 4294967 seconds, approximately 49 days. 2053 2054=item socketpair 2055 2056(S<RISC OS>) 2057Not implemented. 2058 2059(VMS) 2060Available on 64 bit OpenVMS 8.2 and later. 2061 2062=item stat 2063 2064Platforms that do not have C<rdev>, C<blksize>, or C<blocks> will return 2065these as C<''>, so numeric comparison or manipulation of these fields may 2066cause 'not numeric' warnings. 2067 2068(S<Mac OS X>) 2069C<ctime> not supported on UFS. 2070 2071(Win32) 2072C<ctime> is creation time instead of inode change time. 2073 2074(VMS) 2075C<dev> and C<ino> are not necessarily reliable. 2076 2077(S<RISC OS>) 2078C<mtime>, C<atime> and C<ctime> all return the last modification time. 2079C<dev> and C<ino> are not necessarily reliable. 2080 2081(OS/2) 2082C<dev>, C<rdev>, C<blksize>, and C<blocks> are not available. C<ino> is not 2083meaningful and will differ between stat calls on the same file. 2084 2085(Cygwin) 2086Some versions of cygwin when doing a C<stat("foo")> and not finding it 2087may then attempt to C<stat("foo.exe")>. 2088 2089=item symlink 2090 2091(S<RISC OS>) 2092Not implemented. 2093 2094(Win32) 2095Requires either elevated permissions or developer mode and a 2096sufficiently recent version of Windows 10. You can check whether the current 2097process has the required privileges using the 2098L<Win32::IsSymlinkCreationAllowed()|Win32/Win32::IsSymlinkCreationAllowed()> 2099function. 2100 2101Since Windows needs to know whether the target is a directory or not when 2102creating the link the target Perl will only create the link as a directory 2103link when the target exists and is a directory. 2104 2105Windows does not recognize forward slashes as path separators in 2106symbolic links. Hence on Windows, any C</> in the I<OLDFILE> 2107parameter to symlink() are converted to C<\>. This is reflected in 2108the result returned by readlink(), the C<\> in the result are not 2109converted back to C</>. 2110 2111(VMS) 2112Implemented on 64 bit VMS 8.3. VMS requires the symbolic link to be in Unix 2113syntax if it is intended to resolve to a valid path. 2114 2115=item syscall 2116 2117(Win32, VMS, S<RISC OS>, VOS) 2118Not implemented. 2119 2120=item sysopen 2121 2122(S<Mac OS>, OS/390) 2123The traditional C<0>, C<1>, and C<2> MODEs are implemented with different 2124numeric values on some systems. The flags exported by L<C<Fcntl>|Fcntl> 2125(C<O_RDONLY>, C<O_WRONLY>, C<O_RDWR>) should work everywhere though. 2126 2127=item system 2128 2129(Win32) 2130As an optimization, may not call the command shell specified in 2131C<$ENV{PERL5SHELL}>. C<system(1, @args)> spawns an external 2132process and immediately returns its process designator, without 2133waiting for it to terminate. Return value may be used subsequently 2134in L<C<wait>|perlfunc/wait> or L<C<waitpid>|perlfunc/waitpid PID,FLAGS>. 2135Failure to C<spawn()> a subprocess is indicated by setting 2136L<C<$?>|perlvar/$?> to C<<< 255 << 8 >>>. L<C<$?>|perlvar/$?> is set in a 2137way compatible with Unix (i.e. the exit status of the subprocess is 2138obtained by C<<< $? >> 8 >>>, as described in the documentation). 2139 2140Note that the list form of system() is emulated since the Win32 API 2141CreateProcess() accepts a simple string rather than an array of 2142command-line arguments. This may have security implications for your 2143code. 2144 2145(S<RISC OS>) 2146There is no shell to process metacharacters, and the native standard is 2147to pass a command line terminated by "\n" "\r" or "\0" to the spawned 2148program. Redirection such as C<< > foo >> is performed (if at all) by 2149the run time library of the spawned program. C<system LIST> will call 2150the Unix emulation library's L<C<exec>|perlfunc/exec LIST> emulation, 2151which attempts to provide emulation of the stdin, stdout, stderr in force 2152in the parent, provided the child program uses a compatible version of the 2153emulation library. C<system SCALAR> will call the native command line 2154directly and no such emulation of a child Unix program will occur. 2155Mileage B<will> vary. 2156 2157(Win32) 2158C<system LIST> without the use of indirect object syntax (C<system PROGRAM LIST>) 2159may fall back to trying the shell if the first C<spawn()> fails. 2160 2161(SunOS, Solaris, HP-UX) 2162Does not automatically flush output handles on some platforms. 2163 2164(VMS) 2165As with Win32, C<system(1, @args)> spawns an external process and 2166immediately returns its process designator without waiting for the 2167process to terminate. In this case the return value may be used subsequently 2168in L<C<wait>|perlfunc/wait> or L<C<waitpid>|perlfunc/waitpid PID,FLAGS>. 2169Otherwise the return value is POSIX-like (shifted up by 8 bits), which only 2170allows room for a made-up value derived from the severity bits of the native 217132-bit condition code (unless overridden by 2172L<C<use vmsish 'status'>|vmsish/C<vmsish status>>). If the native 2173condition code is one that has a POSIX value encoded, the POSIX value will 2174be decoded to extract the expected exit value. For more details see 2175L<perlvms/$?>. 2176 2177=item telldir 2178 2179(Android) 2180Not implemented. 2181 2182=item times 2183 2184(Win32) 2185"Cumulative" times will be bogus. On anything other than Windows NT 2186or Windows 2000, "system" time will be bogus, and "user" time is 2187actually the time returned by the L<C<clock()>|clock(3)> function in the C 2188runtime library. 2189 2190(S<RISC OS>) 2191Not useful. 2192 2193=item truncate 2194 2195(Older versions of VMS) 2196Not implemented. 2197 2198(VOS) 2199Truncation to same-or-shorter lengths only. 2200 2201(Win32) 2202If a FILEHANDLE is supplied, it must be writable and opened in append 2203mode (i.e., use C<<< open(my $fh, '>>', 'filename') >>> 2204or C<sysopen(my $fh, ..., O_APPEND|O_RDWR)>. If a filename is supplied, it 2205should not be held open elsewhere. 2206 2207=item umask 2208 2209Returns C<undef> where unavailable. 2210 2211(AmigaOS) 2212C<umask> works but the correct permissions are set only when the file 2213is finally closed. 2214 2215=item utime 2216 2217(VMS, S<RISC OS>) 2218Only the modification time is updated. 2219 2220(Win32) 2221May not behave as expected. Behavior depends on the C runtime 2222library's implementation of L<C<utime()>|utime(2)>, and the filesystem 2223being used. The FAT filesystem typically does not support an "access 2224time" field, and it may limit timestamps to a granularity of two seconds. 2225 2226=item wait 2227 2228=item waitpid 2229 2230(Win32) 2231Can only be applied to process handles returned for processes spawned 2232using C<system(1, ...)> or pseudo processes created with 2233L<C<fork>|perlfunc/fork>. 2234 2235(S<RISC OS>) 2236Not useful. 2237 2238=back 2239 2240 2241=head1 Supported Platforms 2242 2243The following platforms are known to build Perl 5.12 (as of April 2010, 2244its release date) from the standard source code distribution available 2245at L<http://www.cpan.org/src> 2246 2247=over 2248 2249=item Linux (x86, ARM, IA64) 2250 2251=item HP-UX 2252 2253=item AIX 2254 2255=item Win32 2256 2257=over 2258 2259=item Windows 2000 2260 2261=item Windows XP 2262 2263=item Windows Server 2003 2264 2265=item Windows Vista 2266 2267=item Windows Server 2008 2268 2269=item Windows 7 2270 2271=back 2272 2273=item Cygwin 2274 2275Some tests are known to fail: 2276 2277=over 2278 2279=item * 2280 2281F<ext/XS-APItest/t/call_checker.t> - see 2282L<https://github.com/Perl/perl5/issues/10750> 2283 2284=item * 2285 2286F<dist/I18N-Collate/t/I18N-Collate.t> 2287 2288=item * 2289 2290F<ext/Win32CORE/t/win32core.t> - may fail on recent cygwin installs. 2291 2292=back 2293 2294=item Solaris (x86, SPARC) 2295 2296=item OpenVMS 2297 2298=over 2299 2300=item Alpha (7.2 and later) 2301 2302=item I64 (8.2 and later) 2303 2304=back 2305 2306=item NetBSD 2307 2308=item FreeBSD 2309 2310=item Debian GNU/kFreeBSD 2311 2312=item Haiku 2313 2314=item Irix (6.5. What else?) 2315 2316=item OpenBSD 2317 2318=item Dragonfly BSD 2319 2320=item Midnight BSD 2321 2322=item QNX Neutrino RTOS (6.5.0) 2323 2324=item MirOS BSD 2325 2326=item Stratus OpenVOS (17.0 or later) 2327 2328Caveats: 2329 2330=over 2331 2332=item time_t issues that may or may not be fixed 2333 2334=back 2335 2336=item Stratus VOS / OpenVOS 2337 2338=item AIX 2339 2340=item Android 2341 2342=item FreeMINT 2343 2344Perl now builds with FreeMiNT/Atari. It fails a few tests, that needs 2345some investigation. 2346 2347The FreeMiNT port uses GNU dld for loadable module capabilities. So 2348ensure you have that library installed when building perl. 2349 2350=back 2351 2352=head1 EOL Platforms 2353 2354=head2 (Perl 5.37.1) 2355 2356The following platforms were supported by a previous version of 2357Perl but have been officially removed from Perl's source code 2358as of 5.37.1: 2359 2360=over 2361 2362=item Ultrix 2363 2364=back 2365 2366=head2 (Perl 5.36) 2367 2368The following platforms were supported by a previous version of 2369Perl but have been officially removed from Perl's source code 2370as of 5.36: 2371 2372=over 2373 2374=item NetWare 2375 2376=item DOS/DJGPP 2377 2378=item AT&T UWIN 2379 2380=back 2381 2382=head2 (Perl 5.20) 2383 2384The following platforms were supported by a previous version of 2385Perl but have been officially removed from Perl's source code 2386as of 5.20: 2387 2388=over 2389 2390=item AT&T 3b1 2391 2392=back 2393 2394=head2 (Perl 5.14) 2395 2396The following platforms were supported up to 5.10. They may still 2397have worked in 5.12, but supporting code has been removed for 5.14: 2398 2399=over 2400 2401=item Windows 95 2402 2403=item Windows 98 2404 2405=item Windows ME 2406 2407=item Windows NT4 2408 2409=back 2410 2411=head2 (Perl 5.12) 2412 2413The following platforms were supported by a previous version of 2414Perl but have been officially removed from Perl's source code 2415as of 5.12: 2416 2417=over 2418 2419=item Atari MiNT 2420 2421=item Apollo Domain/OS 2422 2423=item Apple Mac OS 8/9 2424 2425=item Tenon Machten 2426 2427=back 2428 2429 2430=head1 Supported Platforms (Perl 5.8) 2431 2432As of July 2002 (the Perl release 5.8.0), the following platforms were 2433able to build Perl from the standard source code distribution 2434available at L<http://www.cpan.org/src/> 2435 2436 AIX 2437 BeOS 2438 BSD/OS (BSDi) 2439 Cygwin 2440 DG/UX 2441 DOS DJGPP 1) 2442 DYNIX/ptx 2443 EPOC R5 2444 FreeBSD 2445 HI-UXMPP (Hitachi) (5.8.0 worked but we didn't know it) 2446 HP-UX 2447 IRIX 2448 Linux 2449 Mac OS Classic 2450 Mac OS X (Darwin) 2451 MPE/iX 2452 NetBSD 2453 NetWare 2454 NonStop-UX 2455 ReliantUNIX (formerly SINIX) 2456 OpenBSD 2457 OpenVMS (formerly VMS) 2458 Open UNIX (Unixware) (since Perl 5.8.1/5.9.0) 2459 OS/2 2460 OS/400 (using the PASE) (since Perl 5.8.1/5.9.0) 2461 POSIX-BC (formerly BS2000) 2462 QNX 2463 Solaris 2464 SunOS 4 2465 SUPER-UX (NEC) 2466 Tru64 UNIX (formerly DEC OSF/1, Digital UNIX) 2467 UNICOS 2468 UNICOS/mk 2469 UTS 2470 VOS / OpenVOS 2471 Win95/98/ME/2K/XP 2) 2472 WinCE 2473 z/OS (formerly OS/390) 2474 VM/ESA 2475 2476 1) in DOS mode either the DOS or OS/2 ports can be used 2477 2) compilers: Borland, MinGW (GCC), VC6 2478 2479The following platforms worked with the previous releases (5.6 and 24805.7), but we did not manage either to fix or to test these in time 2481for the 5.8.0 release. There is a very good chance that many of these 2482will work fine with the 5.8.0. 2483 2484 BSD/OS 2485 DomainOS 2486 Hurd 2487 LynxOS 2488 MachTen 2489 PowerMAX 2490 SCO SV 2491 SVR4 2492 Unixware 2493 Windows 3.1 2494 2495Known to be broken for 5.8.0 (but 5.6.1 and 5.7.2 can be used): 2496 2497 AmigaOS 3 2498 2499The following platforms have been known to build Perl from source in 2500the past (5.005_03 and earlier), but we haven't been able to verify 2501their status for the current release, either because the 2502hardware/software platforms are rare or because we don't have an 2503active champion on these platforms--or both. They used to work, 2504though, so go ahead and try compiling them, and let 2505L<https://github.com/Perl/perl5/issues> know 2506of any trouble. 2507 2508 3b1 2509 A/UX 2510 ConvexOS 2511 CX/UX 2512 DC/OSx 2513 DDE SMES 2514 DOS EMX 2515 Dynix 2516 EP/IX 2517 ESIX 2518 FPS 2519 GENIX 2520 Greenhills 2521 ISC 2522 MachTen 68k 2523 MPC 2524 NEWS-OS 2525 NextSTEP 2526 OpenSTEP 2527 Opus 2528 Plan 9 2529 RISC/os 2530 SCO ODT/OSR 2531 Stellar 2532 SVR2 2533 TI1500 2534 TitanOS 2535 Unisys Dynix 2536 2537The following platforms have their own source code distributions and 2538binaries available via L<http://www.cpan.org/ports/> 2539 2540 Perl release 2541 2542 OS/400 (ILE) 5.005_02 2543 Tandem Guardian 5.004 2544 2545The following platforms have only binaries available via 2546L<http://www.cpan.org/ports/index.html> : 2547 2548 Perl release 2549 2550 Acorn RISCOS 5.005_02 2551 AOS 5.002 2552 LynxOS 5.004_02 2553 2554Although we do suggest that you always build your own Perl from 2555the source code, both for maximal configurability and for security, 2556in case you are in a hurry you can check 2557L<http://www.cpan.org/ports/index.html> for binary distributions. 2558 2559=head1 SEE ALSO 2560 2561L<perlaix>, L<perlamiga>, L<perlbs2000>, 2562L<perlcygwin>, 2563L<perlebcdic>, L<perlfreebsd>, L<perlhurd>, L<perlhpux>, L<perlirix>, 2564L<perlmacosx>, 2565L<perlos2>, L<perlos390>, L<perlos400>, 2566L<perlplan9>, L<perlqnx>, L<perlsolaris>, L<perltru64>, 2567L<perlunicode>, L<perlvms>, L<perlvos>, L<perlwin32>, and L<Win32>. 2568 2569=head1 AUTHORS / CONTRIBUTORS 2570 2571Abigail <abigail@abigail.be>, 2572Charles Bailey <bailey@newman.upenn.edu>, 2573Graham Barr <gbarr@pobox.com>, 2574Tom Christiansen <tchrist@perl.com>, 2575Nicholas Clark <nick@ccl4.org>, 2576Thomas Dorner <Thomas.Dorner@start.de>, 2577Andy Dougherty <doughera@lafayette.edu>, 2578Dominic Dunlop <domo@computer.org>, 2579Neale Ferguson <neale@vma.tabnsw.com.au>, 2580David J. Fiander <davidf@mks.com>, 2581Paul Green <Paul.Green@stratus.com>, 2582M.J.T. Guy <mjtg@cam.ac.uk>, 2583Jarkko Hietaniemi <jhi@iki.fi>, 2584Luther Huffman <lutherh@stratcom.com>, 2585Nick Ing-Simmons <nick@ing-simmons.net>, 2586Andreas J. KE<ouml>nig <a.koenig@mind.de>, 2587Markus Laker <mlaker@contax.co.uk>, 2588Andrew M. Langmead <aml@world.std.com>, 2589Lukas Mai <l.mai@web.de>, 2590Larry Moore <ljmoore@freespace.net>, 2591Paul Moore <Paul.Moore@uk.origin-it.com>, 2592Chris Nandor <pudge@pobox.com>, 2593Matthias Neeracher <neeracher@mac.com>, 2594Philip Newton <pne@cpan.org>, 2595Gary Ng <71564.1743@CompuServe.COM>, 2596Tom Phoenix <rootbeer@teleport.com>, 2597AndrE<eacute> Pirard <A.Pirard@ulg.ac.be>, 2598Peter Prymmer <pvhp@forte.com>, 2599Hugo van der Sanden <hv@crypt0.demon.co.uk>, 2600Gurusamy Sarathy <gsar@activestate.com>, 2601Paul J. Schinder <schinder@pobox.com>, 2602Michael G Schwern <schwern@pobox.com>, 2603Dan Sugalski <dan@sidhe.org>, 2604Nathan Torkington <gnat@frii.com>, 2605John Malmberg <wb8tyw@qsl.net> 2606