{ "mode": "man", "parameter": "perlport", "section": "1", "url": "https://www.chedong.com/phpMan.php/man/perlport/1/json", "generated": "2026-10-09T16:51:49Z", "sections": { "NAME": { "content": "perlport - Writing portable Perl\n", "subsections": [] }, "DESCRIPTION": { "content": "Perl runs on numerous operating systems. While most of them share much in common, they also\nhave their own unique features.\n\nThis document is meant to help you to find out what constitutes portable Perl code. That way\nonce you make a decision to write portably, you know where the lines are drawn, and you can\nstay within them.\n\nThere is a tradeoff between taking full advantage of one particular type of computer and\ntaking advantage of a full range of them. Naturally, as you broaden your range and become\nmore diverse, the common factors drop, and you are left with an increasingly smaller area of\ncommon ground in which you can operate to accomplish a particular task. Thus, when you begin\nattacking a problem, it is important to consider under which part of the tradeoff curve you\nwant to operate. Specifically, you must decide whether it is important that the task that\nyou are coding has the full generality of being portable, or whether to just get the job done\nright now. This is the hardest choice to be made. The rest is easy, because Perl provides\nmany choices, whichever way you want to approach your problem.\n\nLooking at it another way, writing portable code is usually about willfully limiting your\navailable choices. Naturally, it takes discipline and sacrifice to do that. The product of\nportability and convenience may be a constant. You have been warned.\n\nBe aware of two important points:\n\nNot all Perl programs have to be portable\nThere is no reason you should not use Perl as a language to glue Unix tools together, or\nto prototype a Macintosh application, or to manage the Windows registry. If it makes no\nsense to aim for portability for one reason or another in a given program, then don't\nbother.\n\nNearly all of Perl already is portable\nDon't be fooled into thinking that it is hard to create portable Perl code. It isn't.\nPerl tries its level-best to bridge the gaps between what's available on different\nplatforms, and all the means available to use those features. Thus almost all Perl code\nruns on any machine without modification. But there are some significant issues in\nwriting portable code, and this document is entirely about those issues.\n\nHere's the general rule: When you approach a task commonly done using a whole range of\nplatforms, think about writing portable code. That way, you don't sacrifice much by way of\nthe implementation choices you can avail yourself of, and at the same time you can give your\nusers lots of platform choices. On the other hand, when you have to take advantage of some\nunique feature of a particular platform, as is often the case with systems programming\n(whether for Unix, Windows, VMS, etc.), consider writing platform-specific code.\n\nWhen the code will run on only two or three operating systems, you may need to consider only\nthe differences of those particular systems. The important thing is to decide where the code\nwill run and to be deliberate in your decision.\n\nThe material below is separated into three main sections: main issues of portability\n(\"ISSUES\"), platform-specific issues (\"PLATFORMS\"), and built-in Perl functions that behave\ndifferently on various ports (\"FUNCTION IMPLEMENTATIONS\").\n\nThis information should not be considered complete; it includes possibly transient\ninformation about idiosyncrasies of some of the ports, almost all of which are in a state of\nconstant evolution. Thus, this material should be considered a perpetual work in progress\n(\"\"Under\").\n", "subsections": [] }, "ISSUES": { "content": "", "subsections": [ { "name": "Newlines", "content": "In most operating systems, lines in files are terminated by newlines. Just what is used as a\nnewline may vary from OS to OS. Unix traditionally uses \"\\012\", one type of DOSish I/O uses\n\"\\015\\012\", Mac OS uses \"\\015\", and z/OS uses \"\\025\".\n\nPerl uses \"\\n\" to represent the \"logical\" newline, where what is logical may depend on the\nplatform in use. In MacPerl, \"\\n\" always means \"\\015\". On EBCDIC platforms, \"\\n\" could be\n\"\\025\" or \"\\045\". In DOSish perls, \"\\n\" usually means \"\\012\", but when accessing a file in\n\"text\" mode, perl uses the \":crlf\" layer that translates it to (or from) \"\\015\\012\",\ndepending on whether you're reading or writing. Unix does the same thing on ttys in canonical\nmode. \"\\015\\012\" is commonly referred to as CRLF.\n\nTo trim trailing newlines from text lines use \"chomp\". With default settings that function\nlooks for a trailing \"\\n\" character and thus trims in a portable way.\n\nWhen dealing with binary files (or text files in binary mode) be sure to explicitly set $/ to\nthe appropriate value for your file format before using \"chomp\".\n\nBecause of the \"text\" mode translation, DOSish perls have limitations in using \"seek\" and\n\"tell\" on a file accessed in \"text\" mode. Stick to \"seek\"-ing to locations you got from\n\"tell\" (and no others), and you are usually free to use \"seek\" and \"tell\" even in \"text\"\nmode. Using \"seek\" or \"tell\" or other file operations may be non-portable. If you use\n\"binmode\" on a file, however, you can usually \"seek\" and \"tell\" with arbitrary values safely.\n\nA common misconception in socket programming is that \"\\n eq \\012\" everywhere. When using\nprotocols such as common Internet protocols, \"\\012\" and \"\\015\" are called for specifically,\nand the values of the logical \"\\n\" and \"\\r\" (carriage return) are not reliable.\n\nprint $socket \"Hi there, client!\\r\\n\"; # WRONG\nprint $socket \"Hi there, client!\\015\\012\"; # RIGHT\n\nHowever, using \"\\015\\012\" (or \"\\cM\\cJ\", or \"\\x0D\\x0A\") can be tedious and unsightly, as well\nas confusing to those maintaining the code. As such, the \"Socket\" module supplies the Right\nThing for those who want it.\n\nuse Socket qw(:DEFAULT :crlf);\nprint $socket \"Hi there, client!$CRLF\" # RIGHT\n\nWhen reading from a socket, remember that the default input record separator $/ is \"\\n\", but\nrobust socket code will recognize as either \"\\012\" or \"\\015\\012\" as end of line:\n\nwhile (<$socket>) { # NOT ADVISABLE!\n# ...\n}\n\nBecause both CRLF and LF end in LF, the input record separator can be set to LF and any CR\nstripped later. Better to write:\n\nuse Socket qw(:DEFAULT :crlf);\nlocal($/) = LF; # not needed if $/ is already \\012\n\nwhile (<$socket>) {\ns/$CR?$LF/\\n/; # not sure if socket uses LF or CRLF, OK\n# s/\\015?\\012/\\n/; # same thing\n}\n\nThis example is preferred over the previous one--even for Unix platforms--because now any\n\"\\015\"'s (\"\\cM\"'s) are stripped out (and there was much rejoicing).\n\nSimilarly, functions that return text data--such as a function that fetches a web\npage--should sometimes translate newlines before returning the data, if they've not yet been\ntranslated to the local newline representation. A single line of code will often suffice:\n\n$data =~ s/\\015?\\012/\\n/g;\nreturn $data;\n\nSome of this may be confusing. Here's a handy reference to the ASCII CR and LF characters.\nYou can print it out and stick it in your wallet.\n\nLF eq \\012 eq \\x0A eq \\cJ eq chr(10) eq ASCII 10\nCR eq \\015 eq \\x0D eq \\cM eq chr(13) eq ASCII 13\n\n| Unix | DOS | Mac |\n---------------------------\n\\n | LF | LF | CR |\n\\r | CR | CR | LF |\n\\n * | LF | CRLF | CR |\n\\r * | CR | CR | LF |\n---------------------------\n* text-mode STDIO\n\nThe Unix column assumes that you are not accessing a serial line (like a tty) in canonical\nmode. If you are, then CR on input becomes \"\\n\", and \"\\n\" on output becomes CRLF.\n\nThese are just the most common definitions of \"\\n\" and \"\\r\" in Perl. There may well be\nothers. For example, on an EBCDIC implementation such as z/OS (OS/390) or OS/400 (using the\nILE, the PASE is ASCII-based) the above material is similar to \"Unix\" but the code numbers\nchange:\n\nLF eq \\025 eq \\x15 eq \\cU eq chr(21) eq CP-1047 21\nLF eq \\045 eq \\x25 eq chr(37) eq CP-0037 37\nCR eq \\015 eq \\x0D eq \\cM eq chr(13) eq CP-1047 13\nCR eq \\015 eq \\x0D eq \\cM eq chr(13) eq CP-0037 13\n\n| z/OS | OS/400 |\n----------------------\n\\n | LF | LF |\n\\r | CR | CR |\n\\n * | LF | LF |\n\\r * | CR | CR |\n----------------------\n* text-mode STDIO\n" }, { "name": "Numbers endianness and Width", "content": "Different CPUs store integers and floating point numbers in different orders (called\nendianness) and widths (32-bit and 64-bit being the most common today). This affects your\nprograms when they attempt to transfer numbers in binary format from one CPU architecture to\nanother, usually either \"live\" via network connection, or by storing the numbers to secondary\nstorage such as a disk file or tape.\n\nConflicting storage orders make an utter mess out of the numbers. If a little-endian host\n(Intel, VAX) stores 0x12345678 (305419896 in decimal), a big-endian host (Motorola, Sparc,\nPA) reads it as 0x78563412 (2018915346 in decimal). Alpha and MIPS can be either:\nDigital/Compaq used/uses them in little-endian mode; SGI/Cray uses them in big-endian mode.\nTo avoid this problem in network (socket) connections use the \"pack\" and \"unpack\" formats \"n\"\nand \"N\", the \"network\" orders. These are guaranteed to be portable.\n\nAs of Perl 5.10.0, you can also use the \">\" and \"<\" modifiers to force big- or little-endian\nbyte-order. This is useful if you want to store signed integers or 64-bit integers, for\nexample.\n\nYou can explore the endianness of your platform by unpacking a data structure packed in\nnative format such as:\n\nprint unpack(\"h*\", pack(\"s2\", 1, 2)), \"\\n\";\n# '10002000' on e.g. Intel x86 or Alpha 21064 in little-endian mode\n# '00100020' on e.g. Motorola 68040\n\nIf you need to distinguish between endian architectures you could use either of the variables\nset like so:\n\n$isbigendian = unpack(\"h*\", pack(\"s\", 1)) =~ /01/;\n$islittleendian = unpack(\"h*\", pack(\"s\", 1)) =~ /^1/;\n\nDiffering widths can cause truncation even between platforms of equal endianness. The\nplatform of shorter width loses the upper parts of the number. There is no good solution for\nthis problem except to avoid transferring or storing raw binary numbers.\n\nOne can circumnavigate both these problems in two ways. Either transfer and store numbers\nalways in text format, instead of raw binary, or else consider using modules like\n\"Data::Dumper\" and \"Storable\" (included as of Perl 5.8). Keeping all data as text\nsignificantly simplifies matters.\n" }, { "name": "Files and Filesystems", "content": "Most platforms these days structure files in a hierarchical fashion. So, it is reasonably\nsafe to assume that all platforms support the notion of a \"path\" to uniquely identify a file\non the system. How that path is really written, though, differs considerably.\n\nAlthough similar, file path specifications differ between Unix, Windows, Mac OS, OS/2, VMS,\nVOS, RISC OS, and probably others. Unix, for example, is one of the few OSes that has the\nelegant idea of a single root directory.\n\nDOS, OS/2, VMS, VOS, and Windows can work similarly to Unix with \"/\" as path separator, or in\ntheir own idiosyncratic ways (such as having several root directories and various \"unrooted\"\ndevice files such NIL: and LPT:).\n\nMac OS 9 and earlier used \":\" as a path separator instead of \"/\".\n\nThe filesystem may support neither hard links (\"link\") nor symbolic links (\"symlink\",\n\"readlink\", \"lstat\").\n\nThe filesystem may support neither access timestamp nor change timestamp (meaning that about\nthe only portable timestamp is the modification timestamp), or one second granularity of any\ntimestamps (e.g. the FAT filesystem limits the time granularity to two seconds).\n\nThe \"inode change timestamp\" (the \"-C\" filetest) may really be the \"creation timestamp\"\n(which it is not in Unix).\n\nVOS perl can emulate Unix filenames with \"/\" as path separator. The native pathname\ncharacters greater-than, less-than, number-sign, and percent-sign are always accepted.\n\nRISC OS perl can emulate Unix filenames with \"/\" as path separator, or go native and use \".\"\nfor path separator and \":\" to signal filesystems and disk names.\n\nDon't assume Unix filesystem access semantics: that read, write, and execute are all the\npermissions there are, and even if they exist, that their semantics (for example what do \"r\",\n\"w\", and \"x\" mean on a directory) are the Unix ones. The various Unix/POSIX compatibility\nlayers usually try to make interfaces like \"chmod\" work, but sometimes there simply is no\ngood mapping.\n\nThe \"File::Spec\" modules provide methods to manipulate path specifications and return the\nresults in native format for each platform. This is often unnecessary as Unix-style paths\nare understood by Perl on every supported platform, but if you need to produce native paths\nfor a native utility that does not understand Unix syntax, or if you are operating on paths\nor path components in unknown (and thus possibly native) syntax, \"File::Spec\" is your friend.\nHere are two brief examples:\n\nuse File::Spec::Functions;\nchdir(updir()); # go up one directory\n\n# Concatenate a path from its components\nmy $file = catfile(updir(), 'temp', 'file.txt');\n# on Unix: '../temp/file.txt'\n# on Win32: '..\\temp\\file.txt'\n# on VMS: '[-.temp]file.txt'\n\nIn general, production code should not have file paths hardcoded. Making them user-supplied\nor read from a configuration file is better, keeping in mind that file path syntax varies on\ndifferent machines.\n\nThis is especially noticeable in scripts like Makefiles and test suites, which often assume\n\"/\" as a path separator for subdirectories.\n\nAlso of use is \"File::Basename\" from the standard distribution, which splits a pathname into\npieces (base filename, full path to directory, and file suffix).\n\nEven when on a single platform (if you can call Unix a single platform), remember not to\ncount on the existence or the contents of particular system-specific files or directories,\nlike /etc/passwd, /etc/sendmail.conf, /etc/resolv.conf, or even /tmp/. For example,\n/etc/passwd may exist but not contain the encrypted passwords, because the system is using\nsome form of enhanced security. Or it may not contain all the accounts, because the system\nis using NIS. If code does need to rely on such a file, include a description of the file\nand its format in the code's documentation, then make it easy for the user to override the\ndefault location of the file.\n\nDon't assume a text file will end with a newline. They should, but people forget.\n\nDo not have two files or directories of the same name with different case, like test.pl and\nTest.pl, as many platforms have case-insensitive (or at least case-forgiving) filenames.\nAlso, try not to have non-word characters (except for \".\") in the names, and keep them to the\n8.3 convention, for maximum portability, onerous a burden though this may appear.\n\nLikewise, when using the \"AutoSplit\" module, try to keep your functions to 8.3 naming and\ncase-insensitive conventions; or, at the least, make it so the resulting files have a unique\n(case-insensitively) first 8 characters.\n\nWhitespace in filenames is tolerated on most systems, but not all, and even on systems where\nit might be tolerated, some utilities might become confused by such whitespace.\n\nMany systems (DOS, VMS ODS-2) cannot have more than one \".\" in their filenames.\n\nDon't assume \">\" won't be the first character of a filename. Always use the three-arg\nversion of \"open\":\n\nopen my $fh, '<', $existingfile) or die $!;\n\nTwo-arg \"open\" is magic and can translate characters like \">\", \"<\", and \"|\" in filenames,\nwhich is usually the wrong thing to do. \"sysopen\" and three-arg \"open\" don't have this\nproblem.\n\nDon't use \":\" as a part of a filename since many systems use that for their own semantics\n(Mac OS Classic for separating pathname components, many networking schemes and utilities for\nseparating the nodename and the pathname, and so on). For the same reasons, avoid \"@\", \";\"\nand \"|\".\n\nDon't assume that in pathnames you can collapse two leading slashes \"//\" into one: some\nnetworking and clustering filesystems have special semantics for that. Let the operating\nsystem sort it out.\n\nThe portable filename characters as defined by ANSI C are\n\na 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\nA 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\n0 1 2 3 4 5 6 7 8 9\n. -\n\nand \"-\" shouldn't be the first character. If you want to be hypercorrect, stay case-\ninsensitive and within the 8.3 naming convention (all the files and directories have to be\nunique within one directory if their names are lowercased and truncated to eight characters\nbefore the \".\", if any, and to three characters after the \".\", if any). (And do not use \".\"s\nin directory names.)\n" }, { "name": "System Interaction", "content": "Not all platforms provide a command line. These are usually platforms that rely primarily on\na Graphical User Interface (GUI) for user interaction. A program requiring a command line\ninterface might not work everywhere. This is probably for the user of the program to deal\nwith, so don't stay up late worrying about it.\n\nSome platforms can't delete or rename files held open by the system, this limitation may also\napply to changing filesystem metainformation like file permissions or owners. Remember to\n\"close\" files when you are done with them. Don't \"unlink\" or \"rename\" an open file. Don't\n\"tie\" or \"open\" a file already tied or opened; \"untie\" or \"close\" it first.\n\nDon't open the same file more than once at a time for writing, as some operating systems put\nmandatory locks on such files.\n\nDon't assume that write/modify permission on a directory gives the right to add or delete\nfiles/directories in that directory. That is filesystem specific: in some filesystems you\nneed write/modify permission also (or even just) in the file/directory itself. In some\nfilesystems (AFS, DFS) the permission to add/delete directory entries is a completely\nseparate permission.\n\nDon't assume that a single \"unlink\" completely gets rid of the file: some filesystems (most\nnotably the ones in VMS) have versioned filesystems, and \"unlink\" removes only the most\nrecent one (it doesn't remove all the versions because by default the native tools on those\nplatforms remove just the most recent version, too). The portable idiom to remove all the\nversions of a file is\n\n1 while unlink \"file\";\n\nThis will terminate if the file is undeletable for some reason (protected, not there, and so\non).\n\nDon't count on a specific environment variable existing in %ENV. Don't count on %ENV entries\nbeing case-sensitive, or even case-preserving. Don't try to clear %ENV by saying \"%ENV =\n();\", or, if you really have to, make it conditional on \"$^O ne 'VMS'\" since in VMS the %ENV\ntable is much more than a per-process key-value string table.\n\nOn VMS, some entries in the %ENV hash are dynamically created when their key is used on a\nread if they did not previously exist. The values for $ENV{HOME}, $ENV{TERM}, $ENV{PATH},\nand $ENV{USER}, are known to be dynamically generated. The specific names that are\ndynamically generated may vary with the version of the C library on VMS, and more may exist\nthan are documented.\n\nOn VMS by default, changes to the %ENV hash persist after perl exits. Subsequent invocations\nof perl in the same process can inadvertently inherit environment settings that were meant to\nbe temporary.\n\nDon't count on signals or %SIG for anything.\n\nDon't count on filename globbing. Use \"opendir\", \"readdir\", and \"closedir\" instead.\n\nDon't count on per-program environment variables, or per-program current directories.\n\nDon't count on specific values of $!, neither numeric nor especially the string values. Users\nmay switch their locales causing error messages to be translated into their languages. If\nyou can trust a POSIXish environment, you can portably use the symbols defined by the \"Errno\"\nmodule, like \"ENOENT\". And don't trust on the values of $! at all except immediately after a\nfailed system call.\n" }, { "name": "Command names versus file pathnames", "content": "Don't assume that the name used to invoke a command or program with \"system\" or \"exec\" can\nalso be used to test for the existence of the file that holds the executable code for that\ncommand or program. First, many systems have \"internal\" commands that are built-in to the\nshell or OS and while these commands can be invoked, there is no corresponding file. Second,\nsome operating systems (e.g., Cygwin, OS/2, and VOS) have required suffixes for executable\nfiles; these suffixes are generally permitted on the command name but are not required.\nThus, a command like \"perl\" might exist in a file named perl, perl.exe, or perl.pm, depending\non the operating system. The variable $Config{exe} in the \"Config\" module holds the\nexecutable suffix, if any. Third, the VMS port carefully sets up $^X and $Config{perlpath}\nso that no further processing is required. This is just as well, because the matching\nregular expression used below would then have to deal with a possible trailing version number\nin the VMS file name.\n\nTo convert $^X to a file pathname, taking account of the requirements of the various\noperating system possibilities, say:\n\nuse Config;\nmy $thisperl = $^X;\nif ($^O ne 'VMS') {\n$thisperl .= $Config{exe}\nunless $thisperl =~ m/\\Q$Config{exe}\\E$/i;\n}\n\nTo convert $Config{perlpath} to a file pathname, say:\n\nuse Config;\nmy $thisperl = $Config{perlpath};\nif ($^O ne 'VMS') {\n$thisperl .= $Config{exe}\nunless $thisperl =~ m/\\Q$Config{exe}\\E$/i;\n}\n" }, { "name": "Networking", "content": "Don't assume that you can reach the public Internet.\n\nDon't assume that there is only one way to get through firewalls to the public Internet.\n\nDon't assume that you can reach outside world through any other port than 80, or some web\nproxy. ftp is blocked by many firewalls.\n\nDon't assume that you can send email by connecting to the local SMTP port.\n\nDon't assume that you can reach yourself or any node by the name 'localhost'. The same goes\nfor '127.0.0.1'. You will have to try both.\n\nDon't assume that the host has only one network card, or that it can't bind to many virtual\nIP addresses.\n\nDon't assume a particular network device name.\n\nDon't assume a particular set of \"ioctl\"s will work.\n\nDon't assume that you can ping hosts and get replies.\n\nDon't assume that any particular port (service) will respond.\n\nDon't assume that \"Sys::Hostname\" (or any other API or command) returns either a fully\nqualified hostname or a non-qualified hostname: it all depends on how the system had been\nconfigured. Also remember that for things such as DHCP and NAT, the hostname you get back\nmight not be very useful.\n\nAll the above don'ts may look daunting, and they are, but the key is to degrade gracefully if\none cannot reach the particular network service one wants. Croaking or hanging do not look\nvery professional.\n" }, { "name": "Interprocess Communication (IPC)", "content": "In general, don't directly access the system in code meant to be portable. That means, no\n\"system\", \"exec\", \"fork\", \"pipe\", `` or \"qx//\", \"open\" with a \"|\", nor any of the other\nthings that makes being a Perl hacker worth being.\n\nCommands that launch external processes are generally supported on most platforms (though\nmany of them do not support any type of forking). The problem with using them arises from\nwhat you invoke them on. External tools are often named differently on different platforms,\nmay not be available in the same location, might accept different arguments, can behave\ndifferently, and often present their results in a platform-dependent way. Thus, you should\nseldom depend on them to produce consistent results. (Then again, if you're calling \"netstat\n-a\", you probably don't expect it to run on both Unix and CP/M.)\n\nOne especially common bit of Perl code is opening a pipe to sendmail:\n\nopen(my $mail, '|-', '/usr/lib/sendmail -t')\nor die \"cannot fork sendmail: $!\";\n\nThis is fine for systems programming when sendmail is known to be available. But it is not\nfine for many non-Unix systems, and even some Unix systems that may not have sendmail\ninstalled. If a portable solution is needed, see the various distributions on CPAN that deal\nwith it. \"Mail::Mailer\" and \"Mail::Send\" in the \"MailTools\" distribution are commonly used,\nand provide several mailing methods, including \"mail\", \"sendmail\", and direct SMTP (via\n\"Net::SMTP\") if a mail transfer agent is not available. \"Mail::Sendmail\" is a standalone\nmodule that provides simple, platform-independent mailing.\n\nThe Unix System V IPC (\"msg*(), sem*(), shm*()\") is not available even on all Unix platforms.\n\nDo not use either the bare result of \"pack(\"N\", 10, 20, 30, 40)\" or bare v-strings (such as\n\"v10.20.30.40\") to represent IPv4 addresses: both forms just pack the four bytes into network\norder. That this would be equal to the C language \"inaddr\" struct (which is what the socket\ncode internally uses) is not guaranteed. To be portable use the routines of the \"Socket\"\nmodule, such as \"inetaton\", \"inetntoa\", and \"sockaddrin\".\n\nThe rule of thumb for portable code is: Do it all in portable Perl, or use a module (that may\ninternally implement it with platform-specific code, but exposes a common interface).\n" }, { "name": "External Subroutines (XS)", "content": "XS code can usually be made to work with any platform, but dependent libraries, header files,\netc., might not be readily available or portable, or the XS code itself might be platform-\nspecific, just as Perl code might be. If the libraries and headers are portable, then it is\nnormally reasonable to make sure the XS code is portable, too.\n\nA different type of portability issue arises when writing XS code: availability of a C\ncompiler on the end-user's system. C brings with it its own portability issues, and writing\nXS code will expose you to some of those. Writing purely in Perl is an easier way to achieve\nportability.\n" }, { "name": "Standard Modules", "content": "In general, the standard modules work across platforms. Notable exceptions are the \"CPAN\"\nmodule (which currently makes connections to external programs that may not be available),\nplatform-specific modules (like \"ExtUtils::MMVMS\"), and DBM modules.\n\nThere is no one DBM module available on all platforms. \"SDBMFile\" and the others are\ngenerally available on all Unix and DOSish ports, but not in MacPerl, where only \"NDBMFile\"\nand \"DBFile\" are available.\n\nThe good news is that at least some DBM module should be available, and \"AnyDBMFile\" will\nuse whichever module it can find. Of course, then the code needs to be fairly strict,\ndropping to the greatest common factor (e.g., not exceeding 1K for each record), so that it\nwill work with any DBM module. See AnyDBMFile for more details.\n" }, { "name": "Time and Date", "content": "The system's notion of time of day and calendar date is controlled in widely different ways.\nDon't assume the timezone is stored in $ENV{TZ}, and even if it is, don't assume that you can\ncontrol the timezone through that variable. Don't assume anything about the three-letter\ntimezone abbreviations (for example that MST would be the Mountain Standard Time, it's been\nknown to stand for Moscow Standard Time). If you need to use timezones, express them in some\nunambiguous format like the exact number of minutes offset from UTC, or the POSIX timezone\nformat.\n\nDon't assume that the epoch starts at 00:00:00, January 1, 1970, because that is OS- and\nimplementation-specific. It is better to store a date in an unambiguous representation. The\nISO 8601 standard defines YYYY-MM-DD as the date format, or YYYY-MM-DDTHH:MM:SS (that's a\nliteral \"T\" separating the date from the time). Please do use the ISO 8601 instead of making\nus guess what date 02/03/04 might be. ISO 8601 even sorts nicely as-is. A text\nrepresentation (like \"1987-12-18\") can be easily converted into an OS-specific value using a\nmodule like \"Time::Piece\" (see \"Date Parsing\" in Time::Piece) or \"Date::Parse\". An array of\nvalues, such as those returned by \"localtime\", can be converted to an OS-specific\nrepresentation using \"Time::Local\".\n\nWhen calculating specific times, such as for tests in time or date modules, it may be\nappropriate to calculate an offset for the epoch.\n\nuse Time::Local qw(timegm);\nmy $offset = timegm(0, 0, 0, 1, 0, 1970);\n\nThe value for $offset in Unix will be 0, but in Mac OS Classic will be some large number.\n$offset can then be added to a Unix time value to get what should be the proper value on any\nsystem.\n" }, { "name": "Character sets and character encoding", "content": "Assume very little about character sets.\n\nAssume nothing about numerical values (\"ord\", \"chr\") of characters. Do not use explicit code\npoint ranges (like \"\\xHH-\\xHH)\". However, starting in Perl v5.22, regular expression pattern\nbracketed character class ranges specified like \"qr/[\\N{U+HH}-\\N{U+HH}]/\" are portable, and\nstarting in Perl v5.24, the same ranges are portable in \"tr///\". You can portably use\nsymbolic character classes like \"[:print:]\".\n\nDo not assume that the alphabetic characters are encoded contiguously (in the numeric sense).\nThere may be gaps. Special coding in Perl, however, guarantees that all subsets of\n\"qr/[A-Z]/\", \"qr/[a-z]/\", and \"qr/[0-9]/\" behave as expected. \"tr///\" behaves the same for\nthese ranges. In patterns, any ranges specified with end points using the \"\\N{...}\"\nnotations ensures character set portability, but it is a bug in Perl v5.22 that this isn't\ntrue of \"tr///\", fixed in v5.24.\n\nDo not assume anything about the ordering of the characters. The lowercase letters may come\nbefore or after the uppercase letters; the lowercase and uppercase may be interlaced so that\nboth \"a\" and \"A\" come before \"b\"; the accented and other international characters may be\ninterlaced so that ä comes before \"b\". Unicode::Collate can be used to sort this all out.\n" }, { "name": "Internationalisation", "content": "If you may assume POSIX (a rather large assumption), you may read more about the POSIX locale\nsystem from perllocale. The locale system at least attempts to make things a little bit more\nportable, or at least more convenient and native-friendly for non-English users. The system\naffects character sets and encoding, and date and time formatting--amongst other things.\n\nIf you really want to be international, you should consider Unicode. See perluniintro and\nperlunicode for more information.\n\nBy default Perl assumes your source code is written in an 8-bit ASCII superset. To embed\nUnicode characters in your strings and regexes, you can use the \"\\x{HH}\" or (more portably)\n\"\\N{U+HH}\" notations. You can also use the \"utf8\" pragma and write your code in UTF-8, which\nlets you use Unicode characters directly (not just in quoted constructs but also in\nidentifiers).\n" }, { "name": "System Resources", "content": "If your code is destined for systems with severely constrained (or missing!) virtual memory\nsystems then you want to be especially mindful of avoiding wasteful constructs such as:\n\nmy @lines = <$verylargefile>; # bad\n\nwhile (<$fh>) {$file .= $} # sometimes bad\nmy $file = join('', <$fh>); # better\n\nThe last two constructs may appear unintuitive to most people. The first repeatedly grows a\nstring, whereas the second allocates a large chunk of memory in one go. On some systems, the\nsecond is more efficient than the first.\n" }, { "name": "Security", "content": "Most multi-user platforms provide basic levels of security, usually implemented at the\nfilesystem level. Some, however, unfortunately do not. Thus the notion of user id, or\n\"home\" directory, or even the state of being logged-in, may be unrecognizable on many\nplatforms. If you write programs that are security-conscious, it is usually best to know\nwhat type of system you will be running under so that you can write code explicitly for that\nplatform (or class of platforms).\n\nDon't assume the Unix filesystem access semantics: the operating system or the filesystem may\nbe using some ACL systems, which are richer languages than the usual \"rwx\". Even if the\n\"rwx\" exist, their semantics might be different.\n\n(From the security viewpoint, testing for permissions before attempting to do something is\nsilly anyway: if one tries this, there is potential for race conditions. Someone or something\nmight change the permissions between the permissions check and the actual operation. Just\ntry the operation.)\n\nDon't assume the Unix user and group semantics: especially, don't expect $< and $> (or $( and\n$)) to work for switching identities (or memberships).\n\nDon't assume set-uid and set-gid semantics. (And even if you do, think twice: set-uid and\nset-gid are a known can of security worms.)\n" }, { "name": "Style", "content": "For those times when it is necessary to have platform-specific code, consider keeping the\nplatform-specific code in one place, making porting to other platforms easier. Use the\n\"Config\" module and the special variable $^O to differentiate platforms, as described in\n\"PLATFORMS\".\n\nBeware of the \"else syndrome\":\n\nif ($^O eq 'MSWin32') {\n# code that assumes Windows\n} else {\n# code that assumes Linux\n}\n\nThe \"else\" branch should be used for the really ultimate fallback, not for code specific to\nsome platform.\n\nBe careful in the tests you supply with your module or programs. Module code may be fully\nportable, but its tests might not be. This often happens when tests spawn off other\nprocesses or call external programs to aid in the testing, or when (as noted above) the tests\nassume certain things about the filesystem and paths. Be careful not to depend on a specific\noutput style for errors, such as when checking $! after a failed system call. Using $! for\nanything else than displaying it as output is doubtful (though see the \"Errno\" module for\ntesting reasonably portably for error value). Some platforms expect a certain output format,\nand Perl on those platforms may have been adjusted accordingly. Most specifically, don't\nanchor a regex when testing an error value.\n\nCPAN Testers\nModules uploaded to CPAN are tested by a variety of volunteers on different platforms. These\nCPAN testers are notified by mail of each new upload, and reply to the list with PASS, FAIL,\nNA (not applicable to this platform), or UNKNOWN (unknown), along with any relevant\nnotations.\n\nThe purpose of the testing is twofold: one, to help developers fix any problems in their code\nthat crop up because of lack of testing on other platforms; two, to provide users with\ninformation about whether a given module works on a given platform.\n\nAlso see:\n\n• Mailing list: cpan-testers-discuss@perl.org\n\n• Testing results: \n" } ] }, "PLATFORMS": { "content": "Perl is built with a $^O variable that indicates the operating system it was built on. This\nwas implemented to help speed up code that would otherwise have to \"use Config\" and use the\nvalue of $Config{osname}. Of course, to get more detailed information about the system,\nlooking into %Config is certainly recommended.\n\n%Config cannot always be trusted, however, because it was built at compile time. If perl was\nbuilt in one place, then transferred elsewhere, some values may be wrong. The values may\neven have been edited after the fact.\n", "subsections": [ { "name": "Unix", "content": "Perl works on a bewildering variety of Unix and Unix-like platforms (see e.g. most of the\nfiles in the hints/ directory in the source code kit). On most of these systems, the value\nof $^O (hence $Config{osname}, too) is determined either by lowercasing and stripping\npunctuation from the first field of the string returned by typing \"uname -a\" (or a similar\ncommand) at the shell prompt or by testing the file system for the presence of uniquely named\nfiles such as a kernel or header file. Here, for example, are a few of the more popular Unix\nflavors:\n\nuname $^O $Config{archname}\n--------------------------------------------\nAIX aix aix\nBSD/OS bsdos i386-bsdos\nDarwin darwin darwin\nDYNIX/ptx dynixptx i386-dynixptx\nFreeBSD freebsd freebsd-i386\nHaiku haiku BePC-haiku\nLinux linux arm-linux\nLinux linux armv5tel-linux\nLinux linux i386-linux\nLinux linux i586-linux\nLinux linux ppc-linux\nHP-UX hpux PA-RISC1.1\nIRIX irix irix\nMac OS X darwin darwin\nNeXT 3 next next-fat\nNeXT 4 next OPENSTEP-Mach\nopenbsd openbsd i386-openbsd\nOSF1 decosf alpha-decosf\nreliantunix-n svr4 RM400-svr4\nSCOSV scosv i386-scosv\nSINIX-N svr4 RM400-svr4\nsn4609 unicos CRAYC90-unicos\nsn6521 unicosmk t3e-unicosmk\nsn9617 unicos CRAYJ90-unicos\nSunOS solaris sun4-solaris\nSunOS solaris i86pc-solaris\nSunOS4 sunos sun4-sunos\n\nBecause the value of $Config{archname} may depend on the hardware architecture, it can vary\nmore than the value of $^O.\n" }, { "name": "DOS and Derivatives", "content": "Perl has long been ported to Intel-style microcomputers running under systems like PC-DOS,\nMS-DOS, OS/2, and most Windows platforms you can bring yourself to mention (except for\nWindows CE, if you count that). Users familiar with COMMAND.COM or CMD.EXE style shells\nshould be aware that each of these file specifications may have subtle differences:\n\nmy $filespec0 = \"c:/foo/bar/file.txt\";\nmy $filespec1 = \"c:\\\\foo\\\\bar\\\\file.txt\";\nmy $filespec2 = 'c:\\foo\\bar\\file.txt';\nmy $filespec3 = 'c:\\\\foo\\\\bar\\\\file.txt';\n\nSystem calls accept either \"/\" or \"\\\" as the path separator. However, many command-line\nutilities of DOS vintage treat \"/\" as the option prefix, so may get confused by filenames\ncontaining \"/\". Aside from calling any external programs, \"/\" will work just fine, and\nprobably better, as it is more consistent with popular usage, and avoids the problem of\nremembering what to backwhack and what not to.\n\nThe DOS FAT filesystem can accommodate only \"8.3\" style filenames. Under the \"case-\ninsensitive, but case-preserving\" HPFS (OS/2) and NTFS (NT) filesystems you may have to be\ncareful about case returned with functions like \"readdir\" or used with functions like \"open\"\nor \"opendir\".\n\nDOS also treats several filenames as special, such as AUX, PRN, NUL, CON, COM1, LPT1, LPT2,\netc. Unfortunately, sometimes these filenames won't even work if you include an explicit\ndirectory prefix. It is best to avoid such filenames, if you want your code to be portable\nto DOS and its derivatives. It's hard to know what these all are, unfortunately.\n\nUsers of these operating systems may also wish to make use of scripts such as pl2bat.bat to\nput wrappers around your scripts.\n\nNewline (\"\\n\") is translated as \"\\015\\012\" by the I/O system when reading from and writing to\nfiles (see \"Newlines\"). binmode($filehandle) will keep \"\\n\" translated as \"\\012\" for that\nfilehandle. \"binmode\" should always be used for code that deals with binary data. That's\nassuming you realize in advance that your data is in binary. General-purpose programs should\noften assume nothing about their data.\n\nThe $^O variable and the $Config{archname} values for various DOSish perls are as follows:\n\nOS $^O $Config{archname} ID Version\n---------------------------------------------------------\nMS-DOS dos ?\nPC-DOS dos ?\nOS/2 os2 ?\nWindows 3.1 ? ? 0 3 01\nWindows 95 MSWin32 MSWin32-x86 1 4 00\nWindows 98 MSWin32 MSWin32-x86 1 4 10\nWindows ME MSWin32 MSWin32-x86 1 ?\nWindows NT MSWin32 MSWin32-x86 2 4 xx\nWindows NT MSWin32 MSWin32-ALPHA 2 4 xx\nWindows NT MSWin32 MSWin32-ppc 2 4 xx\nWindows 2000 MSWin32 MSWin32-x86 2 5 00\nWindows XP MSWin32 MSWin32-x86 2 5 01\nWindows 2003 MSWin32 MSWin32-x86 2 5 02\nWindows Vista MSWin32 MSWin32-x86 2 6 00\nWindows 7 MSWin32 MSWin32-x86 2 6 01\nWindows 7 MSWin32 MSWin32-x64 2 6 01\nWindows 2008 MSWin32 MSWin32-x86 2 6 01\nWindows 2008 MSWin32 MSWin32-x64 2 6 01\nWindows CE MSWin32 ? 3\nCygwin cygwin cygwin\n\nThe various MSWin32 Perl's can distinguish the OS they are running on via the value of the\nfifth element of the list returned from Win32::GetOSVersion(). For example:\n\nif ($^O eq 'MSWin32') {\nmy @osversioninfo = Win32::GetOSVersion();\nprint +('3.1','95','NT')[$osversioninfo[4]],\"\\n\";\n}\n\nThere are also \"Win32::IsWinNT()|Win32/Win32::IsWinNT()\",\n\"Win32::IsWin95()|Win32/Win32::IsWin95()\", and Win32::GetOSName(); try \"perldoc Win32\". The\nvery portable POSIX::uname() will work too:\n\nc:\\> perl -MPOSIX -we \"print join '|', uname\"\nWindows NT|moonru|5.0|Build 2195 (Service Pack 2)|x86\n\nErrors set by Winsock functions are now put directly into $^E, and the relevant \"WSAE*\" error\ncodes are now exported from the Errno and POSIX modules for testing this against.\n\nThe previous behavior of putting the errors (converted to POSIX-style \"E*\" error codes since\nPerl 5.20.0) into $! was buggy due to the non-equivalence of like-named Winsock and POSIX\nerror constants, a relationship between which has unfortunately been established in one way\nor another since Perl 5.8.0.\n\nThe new behavior provides a much more robust solution for checking Winsock errors in portable\nsoftware without accidentally matching POSIX tests that were intended for other OSes and may\nhave different meanings for Winsock.\n\nThe old behavior is currently retained, warts and all, for backwards compatibility, but users\nare encouraged to change any code that tests $! against \"E*\" constants for Winsock errors to\ninstead test $^E against \"WSAE*\" constants. After a suitable deprecation period, which\nstarted with Perl 5.24, the old behavior may be removed, leaving $! unchanged after Winsock\nfunction calls, to avoid any possible confusion over which error variable to check.\n\nAlso see:\n\n• The EMX environment for DOS, OS/2, etc. emx@iaehv.nl,\n Also perlos2.\n\n• Build instructions for Win32 in perlwin32, or under the Cygnus environment in perlcygwin.\n\n• The \"Win32::*\" modules in Win32.\n\n• The ActiveState Pages, \n\n• The Cygwin environment for Win32; README.cygwin (installed as perlcygwin),\n\n\n• Build instructions for OS/2, perlos2\n\nVMS\nPerl on VMS is discussed in perlvms in the Perl distribution.\n\nThe official name of VMS as of this writing is OpenVMS.\n\nInteracting with Perl from the Digital Command Language (DCL) shell often requires a\ndifferent set of quotation marks than Unix shells do. For example:\n\n$ perl -e \"print \"\"Hello, world.\\n\"\"\"\nHello, world.\n\nThere are several ways to wrap your Perl scripts in DCL .COM files, if you are so inclined.\nFor example:\n\n$ write sys$output \"Hello from DCL!\"\n$ if p1 .eqs. \"\"\n$ then perl -x 'f$environment(\"PROCEDURE\")\n$ else perl -x - 'p1 'p2 'p3 'p4 'p5 'p6 'p7 'p8\n$ deck/dollars=\"END\"\n#!/usr/bin/perl\n\nprint \"Hello from Perl!\\n\";\n\nEND\n$ endif\n\nDo take care with \"$ ASSIGN/nolog/user SYS$COMMAND: SYS$INPUT\" if your Perl-in-DCL script\nexpects to do things like \"$read = ;\".\n\nThe VMS operating system has two filesystems, designated by their on-disk structure (ODS)\nlevel: ODS-2 and its successor ODS-5. The initial port of Perl to VMS pre-dates ODS-5, but\nall current testing and development assumes ODS-5 and its capabilities, including case\npreservation, extended characters in filespecs, and names up to 8192 bytes long.\n\nPerl on VMS can accept either VMS- or Unix-style file specifications as in either of the\nfollowing:\n\n$ perl -ne \"print if /perlsetup/i\" SYS$LOGIN:LOGIN.COM\n$ perl -ne \"print if /perlsetup/i\" /sys$login/login.com\n\nbut not a mixture of both as in:\n\n$ perl -ne \"print if /perlsetup/i\" sys$login:/login.com\nCan't open sys$login:/login.com: file specification syntax error\n\nIn general, the easiest path to portability is always to specify filenames in Unix format\nunless they will need to be processed by native commands or utilities. Because of this\nlatter consideration, the File::Spec module by default returns native format specifications\nregardless of input format. This default may be reversed so that filenames are always\nreported in Unix format by specifying the \"DECC$FILENAMEUNIXREPORT\" feature logical in the\nenvironment.\n\nThe file type, or extension, is always present in a VMS-format file specification even if\nit's zero-length. This means that, by default, \"readdir\" will return a trailing dot on a\nfile with no extension, so where you would see \"a\" on Unix you'll see \"a.\" on VMS. However,\nthe trailing dot may be suppressed by enabling the \"DECC$READDIRDROPDOTNOTYPE\" feature in\nthe environment (see the CRTL documentation on feature logical names).\n\nWhat \"\\n\" represents depends on the type of file opened. It usually represents \"\\012\" but it\ncould also be \"\\015\", \"\\012\", \"\\015\\012\", \"\\000\", \"\\040\", or nothing depending on the file\norganization and record format. The \"VMS::Stdio\" module provides access to the special\nfopen() requirements of files with unusual attributes on VMS.\n\nThe value of $^O on OpenVMS is \"VMS\". To determine the architecture that you are running on\nrefer to $Config{archname}.\n\nOn VMS, perl determines the UTC offset from the \"SYS$TIMEZONEDIFFERENTIAL\" logical name.\nAlthough the VMS epoch began at 17-NOV-1858 00:00:00.00, calls to \"localtime\" are adjusted to\ncount offsets from 01-JAN-1970 00:00:00.00, just like Unix.\n\nAlso see:\n\n• README.vms (installed as READMEvms), perlvms\n\n• vmsperl list, vmsperl-subscribe@perl.org\n\n• vmsperl on the web, \n\n• VMS Software Inc. web site, \n\nVOS\nPerl on VOS (also known as OpenVOS) is discussed in README.vos in the Perl distribution\n(installed as perlvos). Perl on VOS can accept either VOS- or Unix-style file specifications\nas in either of the following:\n\n$ perl -ne \"print if /perlsetup/i\" >system>notices\n$ perl -ne \"print if /perlsetup/i\" /system/notices\n\nor even a mixture of both as in:\n\n$ perl -ne \"print if /perlsetup/i\" >system/notices\n\nEven though VOS allows the slash character to appear in object names, because the VOS port of\nPerl interprets it as a pathname delimiting character, VOS files, directories, or links whose\nnames contain a slash character cannot be processed. Such files must be renamed before they\ncan be processed by Perl.\n\nOlder releases of VOS (prior to OpenVOS Release 17.0) limit file names to 32 or fewer\ncharacters, prohibit file names from starting with a \"-\" character, and prohibit file names\nfrom containing \" \" (space) or any character from the set \"!#%&'()*;<=>?\".\n\nNewer releases of VOS (OpenVOS Release 17.0 or later) support a feature known as extended\nnames. On these releases, file names can contain up to 255 characters, are prohibited from\nstarting with a \"-\" character, and the set of prohibited characters is reduced to \"#%*<>?\".\nThere are restrictions involving spaces and apostrophes: these characters must not begin or\nend a name, nor can they immediately precede or follow a period. Additionally, a space must\nnot immediately precede another space or hyphen. Specifically, the following character\ncombinations are prohibited: space-space, space-hyphen, period-space, space-period, period-\napostrophe, apostrophe-period, leading or trailing space, and leading or trailing apostrophe.\nAlthough an extended file name is limited to 255 characters, a path name is still limited to\n256 characters.\n\nThe value of $^O on VOS is \"vos\". To determine the architecture that you are running on\nrefer to $Config{archname}.\n\nAlso see:\n\n• README.vos (installed as perlvos)\n\n• The VOS mailing list.\n\nThere is no specific mailing list for Perl on VOS. You can contact the Stratus\nTechnologies Customer Assistance Center (CAC) for your region, or you can use the contact\ninformation located in the distribution files on the Stratus Anonymous FTP site.\n\n• Stratus Technologies on the web at \n\n• VOS Open-Source Software on the web at \n" }, { "name": "EBCDIC Platforms", "content": "v5.22 core Perl runs on z/OS (formerly OS/390). Theoretically it could run on the successors\nof OS/400 on AS/400 minicomputers as well as VM/ESA, and BS2000 for S/390 Mainframes. Such\ncomputers use EBCDIC character sets internally (usually Character Code Set ID 0037 for OS/400\nand either 1047 or POSIX-BC for S/390 systems).\n\nThe rest of this section may need updating, but we don't know what it should say. Please\nsubmit comments to .\n\nOn the mainframe Perl currently works under the \"Unix system services for OS/390\" (formerly\nknown as OpenEdition), VM/ESA OpenEdition, or the BS200 POSIX-BC system (BS2000 is supported\nin Perl 5.6 and greater). See perlos390 for details. Note that for OS/400 there is also a\nport of Perl 5.8.1/5.10.0 or later to the PASE which is ASCII-based (as opposed to ILE which\nis EBCDIC-based), see perlos400.\n\nAs of R2.5 of USS for OS/390 and Version 2.3 of VM/ESA these Unix sub-systems do not support\nthe \"#!\" shebang trick for script invocation. Hence, on OS/390 and VM/ESA Perl scripts can\nbe executed with a header similar to the following simple script:\n\n: # use perl\neval 'exec /usr/local/bin/perl -S $0 ${1+\"$@\"}'\nif 0;\n#!/usr/local/bin/perl # just a comment really\n\nprint \"Hello from perl!\\n\";\n\nOS/390 will support the \"#!\" shebang trick in release 2.8 and beyond. Calls to \"system\" and\nbackticks can use POSIX shell syntax on all S/390 systems.\n\nOn the AS/400, if PERL5 is in your library list, you may need to wrap your Perl scripts in a\nCL procedure to invoke them like so:\n\nBEGIN\nCALL PGM(PERL5/PERL) PARM('/QOpenSys/hello.pl')\nENDPGM\n\nThis will invoke the Perl script hello.pl in the root of the QOpenSys file system. On the\nAS/400 calls to \"system\" or backticks must use CL syntax.\n\nOn these platforms, bear in mind that the EBCDIC character set may have an effect on what\nhappens with some Perl functions (such as \"chr\", \"pack\", \"print\", \"printf\", \"ord\", \"sort\",\n\"sprintf\", \"unpack\"), as well as bit-fiddling with ASCII constants using operators like \"^\",\n\"&\" and \"|\", not to mention dealing with socket interfaces to ASCII computers (see\n\"Newlines\").\n\nFortunately, most web servers for the mainframe will correctly translate the \"\\n\" in the\nfollowing statement to its ASCII equivalent (\"\\r\" is the same under both Unix and z/OS):\n\nprint \"Content-type: text/html\\r\\n\\r\\n\";\n\nThe values of $^O on some of these platforms include:\n\nuname $^O $Config{archname}\n--------------------------------------------\nOS/390 os390 os390\nOS400 os400 os400\nPOSIX-BC posix-bc BS2000-posix-bc\n\nSome simple tricks for determining if you are running on an EBCDIC platform could include any\nof the following (perhaps all):\n\nif (\"\\t\" eq \"\\005\") { print \"EBCDIC may be spoken here!\\n\"; }\n\nif (ord('A') == 193) { print \"EBCDIC may be spoken here!\\n\"; }\n\nif (chr(169) eq 'z') { print \"EBCDIC may be spoken here!\\n\"; }\n\nOne thing you may not want to rely on is the EBCDIC encoding of punctuation characters since\nthese may differ from code page to code page (and once your module or script is rumoured to\nwork with EBCDIC, folks will want it to work with all EBCDIC character sets).\n\nAlso see:\n\n• perlos390, perlos400, perlbs2000, perlebcdic.\n\n• The perl-mvs@perl.org list is for discussion of porting issues as well as general usage\nissues for all EBCDIC Perls. Send a message body of \"subscribe perl-mvs\" to\nmajordomo@perl.org.\n\n• AS/400 Perl information at as well as on CPAN in the\nports/ directory.\n" }, { "name": "Acorn RISC OS", "content": "Because Acorns use ASCII with newlines (\"\\n\") in text files as \"\\012\" like Unix, and because\nUnix filename emulation is turned on by default, most simple scripts will probably work \"out\nof the box\". The native filesystem is modular, and individual filesystems are free to be\ncase-sensitive or insensitive, and are usually case-preserving. Some native filesystems have\nname length limits, which file and directory names are silently truncated to fit. Scripts\nshould be aware that the standard filesystem currently has a name length limit of 10\ncharacters, with up to 77 items in a directory, but other filesystems may not impose such\nlimitations.\n\nNative filenames are of the form\n\nFilesystem#SpecialField::DiskName.$.Directory.Directory.File\n\nwhere\n\nSpecialField is not usually present, but may contain . and $ .\nFilesystem =~ m|[A-Za-z0-9]|\nDsicName =~ m|[A-Za-z0-9/]|\n$ represents the root directory\n. is the path separator\n@ is the current directory (per filesystem but machine global)\n^ is the parent directory\nDirectory and File =~ m|[^\\0- \"\\.\\$\\%\\&:\\@\\\\^\\|\\177]+|\n\nThe default filename translation is roughly \"tr|/.|./|\", swapping dots and slashes.\n\nNote that \"\"ADFS::HardDisk.$.File\" ne 'ADFS::HardDisk.$.File'\" and that the second stage of\n\"$\" interpolation in regular expressions will fall foul of the $. variable if scripts are not\ncareful.\n\nLogical paths specified by system variables containing comma-separated search lists are also\nallowed; hence \"System:Modules\" is a valid filename, and the filesystem will prefix \"Modules\"\nwith each section of \"System$Path\" until a name is made that points to an object on disk.\nWriting to a new file \"System:Modules\" would be allowed only if \"System$Path\" contains a\nsingle item list. The filesystem will also expand system variables in filenames if enclosed\nin angle brackets, so \".Modules\" would look for the file\n\"$ENV{'System$Dir'} . 'Modules'\". The obvious implication of this is that fully qualified\nfilenames can start with \"<>\" and the three-argument form of \"open\" should always be used.\n\nBecause \".\" was in use as a directory separator and filenames could not be assumed to be\nunique after 10 characters, Acorn implemented the C compiler to strip the trailing \".c\" \".h\"\n\".s\" and \".o\" suffix from filenames specified in source code and store the respective files\nin subdirectories named after the suffix. Hence files are translated:\n\nfoo.h h.foo\nC:foo.h C:h.foo (logical path variable)\nsys/os.h sys.h.os (C compiler groks Unix-speak)\n10charname.c c.10charname\n10charname.o o.10charname\n11charname.c c.11charname (assuming filesystem truncates at 10)\n\nThe Unix emulation library's translation of filenames to native assumes that this sort of\ntranslation is required, and it allows a user-defined list of known suffixes that it will\ntranspose in this fashion. This may seem transparent, but consider that with these rules\nfoo/bar/baz.h and foo/bar/h/baz both map to foo.bar.h.baz, and that \"readdir\" and \"glob\"\ncannot and do not attempt to emulate the reverse mapping. Other \".\"'s in filenames are\ntranslated to \"/\".\n\nAs implied above, the environment accessed through %ENV is global, and the convention is that\nprogram specific environment variables are of the form \"Program$Name\". Each filesystem\nmaintains a current directory, and the current filesystem's current directory is the global\ncurrent directory. Consequently, sociable programs don't change the current directory but\nrely on full pathnames, and programs (and Makefiles) cannot assume that they can spawn a\nchild process which can change the current directory without affecting its parent (and\neveryone else for that matter).\n\nBecause native operating system filehandles are global and are currently allocated down from\n255, with 0 being a reserved value, the Unix emulation library emulates Unix filehandles.\nConsequently, you can't rely on passing \"STDIN\", \"STDOUT\", or \"STDERR\" to your children.\n\nThe desire of users to express filenames of the form \".Bar\" on the command line\nunquoted causes problems, too: `` command output capture has to perform a guessing game. It\nassumes that a string \"<[^<>]+\\$[^<>]>\" is a reference to an environment variable, whereas\nanything else involving \"<\" or \">\" is redirection, and generally manages to be 99% right. Of\ncourse, the problem remains that scripts cannot rely on any Unix tools being available, or\nthat any tools found have Unix-like command line arguments.\n\nExtensions and XS are, in theory, buildable by anyone using free tools. In practice, many\ndon't, as users of the Acorn platform are used to binary distributions. MakeMaker does run,\nbut no available make currently copes with MakeMaker's makefiles; even if and when this\nshould be fixed, the lack of a Unix-like shell will cause problems with makefile rules,\nespecially lines of the form \"cd sdbm && make all\", and anything using quoting.\n\n\"RISC OS\" is the proper name for the operating system, but the value in $^O is \"riscos\"\n(because we don't like shouting).\n" }, { "name": "Other perls", "content": "Perl has been ported to many platforms that do not fit into any of the categories listed\nabove. Some, such as AmigaOS, QNX, Plan 9, and VOS, have been well-integrated into the\nstandard Perl source code kit. You may need to see the ports/ directory on CPAN for\ninformation, and possibly binaries, for the likes of: aos, Atari ST, lynxos, riscos, Novell\nNetware, Tandem Guardian, etc. (Yes, we know that some of these OSes may fall under the Unix\ncategory, but we are not a standards body.)\n\nSome approximate operating system names and their $^O values in the \"OTHER\" category include:\n\nOS $^O $Config{archname}\n------------------------------------------\nAmiga DOS amigaos m68k-amigos\n\nSee also:\n\n• Amiga, README.amiga (installed as perlamiga).\n\n• Plan 9, README.plan9\n" } ] }, "FUNCTION IMPLEMENTATIONS": { "content": "Listed below are functions that are either completely unimplemented or else have been\nimplemented differently on various platforms. Preceding each description will be, in\nparentheses, a list of platforms that the description applies to.\n\nThe list may well be incomplete, or even wrong in some places. When in doubt, consult the\nplatform-specific README files in the Perl source distribution, and any other documentation\nresources accompanying a given port.\n\nBe aware, moreover, that even among Unix-ish systems there are variations.\n\nFor many functions, you can also query %Config, exported by default from the \"Config\" module.\nFor example, to check whether the platform has the \"lstat\" call, check $Config{dlstat}. See\nConfig for a full description of available variables.\n", "subsections": [ { "name": "Alphabetical Listing of Perl Functions", "content": "-X (Win32) \"-w\" only inspects the read-only file attribute (FILEATTRIBUTEREADONLY),\nwhich determines whether the directory can be deleted, not whether it can be written\nto. Directories always have read and write access unless denied by discretionary\naccess control lists (DACLs).\n\n(VMS) \"-r\", \"-w\", \"-x\", and \"-o\" tell whether the file is accessible, which may not\nreflect UIC-based file protections.\n\n(RISC OS) \"-s\" by name on an open file will return the space reserved on disk, rather\nthan the current extent. \"-s\" on an open filehandle returns the current size.\n\n(Win32, VMS, RISC OS) \"-R\", \"-W\", \"-X\", \"-O\" are indistinguishable from \"-r\", \"-w\",\n\"-x\", \"-o\".\n\n(Win32, VMS, RISC OS) \"-g\", \"-k\", \"-l\", \"-u\", \"-A\" are not particularly meaningful.\n\n(Win32) \"-l\" returns true for both symlinks and directory junctions.\n\n(VMS, RISC OS) \"-p\" is not particularly meaningful.\n\n(VMS) \"-d\" is true if passed a device spec without an explicit directory.\n\n(Win32) \"-x\" (or \"-X\") determine if a file ends in one of the executable suffixes.\n\"-S\" is meaningless.\n\n(RISC OS) \"-x\" (or \"-X\") determine if a file has an executable file type.\n\nalarm (Win32) Emulated using timers that must be explicitly polled whenever Perl wants to\ndispatch \"safe signals\" and therefore cannot interrupt blocking system calls.\n\natan2 (Tru64, HP-UX 10.20) Due to issues with various CPUs, math libraries, compilers, and\nstandards, results for \"atan2\" may vary depending on any combination of the above.\nPerl attempts to conform to the Open Group/IEEE standards for the results returned\nfrom \"atan2\", but cannot force the issue if the system Perl is run on does not allow\nit.\n\nThe current version of the standards for \"atan2\" is available at\n.\n\nbinmode (RISC OS) Meaningless.\n\n(VMS) Reopens file and restores pointer; if function fails, underlying filehandle may\nbe closed, or pointer may be in a different position.\n\n(Win32) The value returned by \"tell\" may be affected after the call, and the\nfilehandle may be flushed.\n\nchdir (Win32) The current directory reported by the system may include any symbolic links\nspecified to chdir().\n\nchmod (Win32) Only good for changing \"owner\" read-write access; \"group\" and \"other\" bits\nare meaningless.\n\n(RISC OS) Only good for changing \"owner\" and \"other\" read-write access.\n\n(VOS) Access permissions are mapped onto VOS access-control list changes.\n\n(Cygwin) The actual permissions set depend on the value of the \"CYGWIN\" variable in\nthe SYSTEM environment settings.\n\n(Android) Setting the exec bit on some locations (generally /sdcard) will return true\nbut not actually set the bit.\n\n(VMS) A mode argument of zero sets permissions to the user's default permission mask\nrather than disabling all permissions.\n\nchown (Plan 9, RISC OS) Not implemented.\n\n(Win32) Does nothing, but won't fail.\n\n(VOS) A little funky, because VOS's notion of ownership is a little funky.\n\nchroot (Win32, VMS, Plan 9, RISC OS, VOS) Not implemented.\n\ncrypt (Win32) May not be available if library or source was not provided when building\nperl.\n\n(Android) Not implemented.\n\ndbmclose\n(VMS, Plan 9, VOS) Not implemented.\n\ndbmopen (VMS, Plan 9, VOS) Not implemented.\n\ndump (RISC OS) Not useful.\n\n(Cygwin, Win32) Not supported.\n\n(VMS) Invokes VMS debugger.\n\nexec (Win32) \"exec LIST\" without the use of indirect object syntax (\"exec PROGRAM LIST\")\nmay fall back to trying the shell if the first spawn() fails.\n\nNote that the list form of exec() is emulated since the Win32 API CreateProcess()\naccepts a simple string rather than an array of command-line arguments. This may\nhave security implications for your code.\n\n(SunOS, Solaris, HP-UX) Does not automatically flush output handles on some\nplatforms.\n\nexit (VMS) Emulates Unix \"exit\" (which considers \"exit 1\" to indicate an error) by mapping\nthe 1 to \"SS$ABORT\" (44). This behavior may be overridden with the pragma \"use\nvmsish 'exit'\". As with the CRTL's exit() function, \"exit 0\" is also mapped to an\nexit status of \"SS$NORMAL\" (1); this mapping cannot be overridden. Any other\nargument to \"exit\" is used directly as Perl's exit status. On VMS, unless the future\nPOSIXEXIT mode is enabled, the exit code should always be a valid VMS exit code and\nnot a generic number. When the POSIXEXIT mode is enabled, a generic number will be\nencoded in a method compatible with the C library POSIXEXIT macro so that it can be\ndecoded by other programs, particularly ones written in C, like the GNV package.\n\n(Solaris) \"exit\" resets file pointers, which is a problem when called from a child\nprocess (created by \"fork\") in \"BEGIN\". A workaround is to use \"POSIX::exit\".\n\nexit unless $Config{archname} =~ /\\bsolaris\\b/;\nrequire POSIX;\nPOSIX::exit(0);\n\nfcntl (Win32) Not implemented.\n\n(VMS) Some functions available based on the version of VMS.\n\nflock (VMS, RISC OS, VOS) Not implemented.\n\nfork (AmigaOS, RISC OS, VMS) Not implemented.\n\n(Win32) Emulated using multiple interpreters. See perlfork.\n\n(SunOS, Solaris, HP-UX) Does not automatically flush output handles on some\nplatforms.\n\ngetlogin\n(RISC OS) Not implemented.\n\ngetpgrp (Win32, VMS, RISC OS) Not implemented.\n\ngetppid (Win32, RISC OS) Not implemented.\n\ngetpriority\n(Win32, VMS, RISC OS, VOS) Not implemented.\n\ngetpwnam\n(Win32) Not implemented.\n\n(RISC OS) Not useful.\n\ngetgrnam\n(Win32, VMS, RISC OS) Not implemented.\n\ngetnetbyname\n(Android, Win32, Plan 9) Not implemented.\n\ngetpwuid\n(Win32) Not implemented.\n\n(RISC OS) Not useful.\n\ngetgrgid\n(Win32, VMS, RISC OS) Not implemented.\n\ngetnetbyaddr\n(Android, Win32, Plan 9) Not implemented.\n\ngetprotobynumber\n(Android) Not implemented.\n\ngetpwent\n(Android, Win32) Not implemented.\n\ngetgrent\n(Android, Win32, VMS) Not implemented.\n\ngethostbyname\n(Irix 5) gethostbyname('localhost') does not work everywhere: you may have to use\ngethostbyname('127.0.0.1').\n\ngethostent\n(Win32) Not implemented.\n\ngetnetent\n(Android, Win32, Plan 9) Not implemented.\n\ngetprotoent\n(Android, Win32, Plan 9) Not implemented.\n\ngetservent\n(Win32, Plan 9) Not implemented.\n\nseekdir (Android) Not implemented.\n\nsethostent\n(Android, Win32, Plan 9, RISC OS) Not implemented.\n\nsetnetent\n(Win32, Plan 9, RISC OS) Not implemented.\n\nsetprotoent\n(Android, Win32, Plan 9, RISC OS) Not implemented.\n\nsetservent\n(Plan 9, Win32, RISC OS) Not implemented.\n\nendpwent\n(Win32) Not implemented.\n\n(Android) Either not implemented or a no-op.\n\nendgrent\n(Android, RISC OS, VMS, Win32) Not implemented.\n\nendhostent\n(Android, Win32) Not implemented.\n\nendnetent\n(Android, Win32, Plan 9) Not implemented.\n\nendprotoent\n(Android, Win32, Plan 9) Not implemented.\n\nendservent\n(Plan 9, Win32) Not implemented.\n\ngetsockopt\n(Plan 9) Not implemented.\n\nglob This operator is implemented via the \"File::Glob\" extension on most platforms. See\nFile::Glob for portability information.\n\ngmtime In theory, \"gmtime\" is reliable from -263 to 263-1. However, because work-\narounds in the implementation use floating point numbers, it will become inaccurate\nas the time gets larger. This is a bug and will be fixed in the future.\n\n(VOS) Time values are 32-bit quantities.\n\nioctl (VMS) Not implemented.\n\n(Win32) Available only for socket handles, and it does what the ioctlsocket() call in\nthe Winsock API does.\n\n(RISC OS) Available only for socket handles.\n\nkill (RISC OS) Not implemented, hence not useful for taint checking.\n\n(Win32) \"kill\" doesn't send a signal to the identified process like it does on Unix\nplatforms. Instead \"kill($sig, $pid)\" terminates the process identified by $pid, and\nmakes it exit immediately with exit status $sig. As in Unix, if $sig is 0 and the\nspecified process exists, it returns true without actually terminating it.\n\n(Win32) \"kill(-9, $pid)\" will terminate the process specified by $pid and recursively\nall child processes owned by it. This is different from the Unix semantics, where\nthe signal will be delivered to all processes in the same process group as the\nprocess specified by $pid.\n\n(VMS) A pid of -1 indicating all processes on the system is not currently supported.\n\nlink (RISC OS, VOS) Not implemented.\n\n(AmigaOS) Link count not updated because hard links are not quite that hard (They are\nsort of half-way between hard and soft links).\n\n(Win32) Hard links are implemented on Win32 under NTFS only. They are natively\nsupported on Windows 2000 and later. On Windows NT they are implemented using the\nWindows POSIX subsystem support and the Perl process will need Administrator or\nBackup Operator privileges to create hard links.\n\n(VMS) Available on 64 bit OpenVMS 8.2 and later.\n\nlocaltime\n\"localtime\" has the same range as \"gmtime\", but because time zone rules change, its\naccuracy for historical and future times may degrade but usually by no more than an\nhour.\n\nlstat (RISC OS) Not implemented.\n\n(Win32) Treats directory junctions as symlinks.\n\nmsgctl\nmsgget\nmsgsnd\nmsgrcv (Android, Win32, VMS, Plan 9, RISC OS, VOS) Not implemented.\n\nopen (RISC OS) Open modes \"|-\" and \"-|\" are unsupported.\n\n(SunOS, Solaris, HP-UX) Opening a process does not automatically flush output handles\non some platforms.\n\n(Win32) Both of modes \"|-\" and \"-|\" are supported, but the list form is emulated\nsince the Win32 API CreateProcess() accepts a simple string rather than an array of\narguments. This may have security implications for your code.\n\nreadlink\n(VMS, RISC OS) Not implemented.\n\n(Win32) readlink() on a directory junction returns the object name, not a simple\npath.\n\nrename (Win32) Can't move directories between directories on different logical volumes.\n\nrewinddir\n(Win32) Will not cause \"readdir\" to re-read the directory stream. The entries\nalready read before the \"rewinddir\" call will just be returned again from a cache\nbuffer.\n\nselect (Win32, VMS) Only implemented on sockets.\n\n(RISC OS) Only reliable on sockets.\n\nNote that the \"select FILEHANDLE\" form is generally portable.\n\nsemctl\nsemget\nsemop (Android, Win32, VMS, RISC OS) Not implemented.\n\nsetgrent\n(Android, VMS, Win32, RISC OS) Not implemented.\n\nsetpgrp (Win32, VMS, RISC OS, VOS) Not implemented.\n\nsetpriority\n(Win32, VMS, RISC OS, VOS) Not implemented.\n\nsetpwent\n(Android, Win32, RISC OS) Not implemented.\n\nsetsockopt\n(Plan 9) Not implemented.\n\nshmctl\nshmget\nshmread\nshmwrite\n(Android, Win32, VMS, RISC OS) Not implemented.\n\nsleep (Win32) Emulated using synchronization functions such that it can be interrupted by\n\"alarm\", and limited to a maximum of 4294967 seconds, approximately 49 days.\n\nsocketpair\n(RISC OS) Not implemented.\n\n(VMS) Available on 64 bit OpenVMS 8.2 and later.\n\nstat Platforms that do not have \"rdev\", \"blksize\", or \"blocks\" will return these as '', so\nnumeric comparison or manipulation of these fields may cause 'not numeric' warnings.\n\n(Mac OS X) \"ctime\" not supported on UFS.\n\n(Win32) \"ctime\" is creation time instead of inode change time.\n\n(VMS) \"dev\" and \"ino\" are not necessarily reliable.\n\n(RISC OS) \"mtime\", \"atime\" and \"ctime\" all return the last modification time. \"dev\"\nand \"ino\" are not necessarily reliable.\n\n(OS/2) \"dev\", \"rdev\", \"blksize\", and \"blocks\" are not available. \"ino\" is not\nmeaningful and will differ between stat calls on the same file.\n\n(Cygwin) Some versions of cygwin when doing a stat(\"foo\") and not finding it may then\nattempt to stat(\"foo.exe\").\n\nsymlink (RISC OS) Not implemented.\n\n(Win32) Requires either elevated permissions or developer mode and a sufficiently\nrecent version of Windows 10. You can check whether the current process has the\nrequired privileges using the Win32::IsSymlinkCreationAllowed() function.\n\nSince Windows needs to know whether the target is a directory or not when creating\nthe link the target Perl will only create the link as a directory link when the\ntarget exists and is a directory.\n\nWindows does not recognize forward slashes as path separators in symbolic links.\nHence on Windows, any \"/\" in the OLDFILE parameter to symlink() are converted to \"\\\".\nThis is reflected in the result returned by readlink(), the \"\\\" in the result are not\nconverted back to \"/\".\n\n(VMS) Implemented on 64 bit VMS 8.3. VMS requires the symbolic link to be in Unix\nsyntax if it is intended to resolve to a valid path.\n\nsyscall (Win32, VMS, RISC OS, VOS) Not implemented.\n\nsysopen (Mac OS, OS/390) The traditional 0, 1, and 2 MODEs are implemented with different\nnumeric values on some systems. The flags exported by \"Fcntl\" (\"ORDONLY\",\n\"OWRONLY\", \"ORDWR\") should work everywhere though.\n\nsystem (Win32) As an optimization, may not call the command shell specified in\n$ENV{PERL5SHELL}. \"system(1, @args)\" spawns an external process and immediately\nreturns its process designator, without waiting for it to terminate. Return value\nmay be used subsequently in \"wait\" or \"waitpid\". Failure to spawn() a subprocess is\nindicated by setting $? to \"255 << 8\". $? is set in a way compatible with Unix (i.e.\nthe exit status of the subprocess is obtained by \"$? >> 8\", as described in the\ndocumentation).\n\nNote that the list form of system() is emulated since the Win32 API CreateProcess()\naccepts a simple string rather than an array of command-line arguments. This may\nhave security implications for your code.\n\n(RISC OS) There is no shell to process metacharacters, and the native standard is to\npass a command line terminated by \"\\n\" \"\\r\" or \"\\0\" to the spawned program.\nRedirection such as \"> foo\" is performed (if at all) by the run time library of the\nspawned program. \"system LIST\" will call the Unix emulation library's \"exec\"\nemulation, which attempts to provide emulation of the stdin, stdout, stderr in force\nin the parent, provided the child program uses a compatible version of the emulation\nlibrary. \"system SCALAR\" will call the native command line directly and no such\nemulation of a child Unix program will occur. Mileage will vary.\n\n(Win32) \"system LIST\" without the use of indirect object syntax (\"system PROGRAM\nLIST\") may fall back to trying the shell if the first spawn() fails.\n\n(SunOS, Solaris, HP-UX) Does not automatically flush output handles on some\nplatforms.\n\n(VMS) As with Win32, \"system(1, @args)\" spawns an external process and immediately\nreturns its process designator without waiting for the process to terminate. In this\ncase the return value may be used subsequently in \"wait\" or \"waitpid\". Otherwise the\nreturn value is POSIX-like (shifted up by 8 bits), which only allows room for a made-\nup value derived from the severity bits of the native 32-bit condition code (unless\noverridden by \"use vmsish 'status'\"). If the native condition code is one that has a\nPOSIX value encoded, the POSIX value will be decoded to extract the expected exit\nvalue. For more details see \"$?\" in perlvms.\n\ntelldir (Android) Not implemented.\n\ntimes (Win32) \"Cumulative\" times will be bogus. On anything other than Windows NT or\nWindows 2000, \"system\" time will be bogus, and \"user\" time is actually the time\nreturned by the clock() function in the C runtime library.\n\n(RISC OS) Not useful.\n\ntruncate\n(Older versions of VMS) Not implemented.\n\n(VOS) Truncation to same-or-shorter lengths only.\n\n(Win32) If a FILEHANDLE is supplied, it must be writable and opened in append mode\n(i.e., use \"open(my $fh, '>>', 'filename')\" or \"sysopen(my $fh, ...,\nOAPPEND|ORDWR)\". If a filename is supplied, it should not be held open elsewhere.\n\numask Returns \"undef\" where unavailable.\n\n(AmigaOS) \"umask\" works but the correct permissions are set only when the file is\nfinally closed.\n\nutime (VMS, RISC OS) Only the modification time is updated.\n\n(Win32) May not behave as expected. Behavior depends on the C runtime library's\nimplementation of utime(), and the filesystem being used. The FAT filesystem\ntypically does not support an \"access time\" field, and it may limit timestamps to a\ngranularity of two seconds.\n\nwait\nwaitpid (Win32) Can only be applied to process handles returned for processes spawned using\n\"system(1, ...)\" or pseudo processes created with \"fork\".\n\n(RISC OS) Not useful.\n" } ] }, "Supported Platforms": { "content": "The following platforms are known to build Perl 5.12 (as of April 2010, its release date)\nfrom the standard source code distribution available at \n\nLinux (x86, ARM, IA64)\nHP-UX\nAIX\nWin32\nWindows 2000\nWindows XP\nWindows Server 2003\nWindows Vista\nWindows Server 2008\nWindows 7\nCygwin\nSome tests are known to fail:\n\n• ext/XS-APItest/t/callchecker.t - see \n\n• dist/I18N-Collate/t/I18N-Collate.t\n\n• ext/Win32CORE/t/win32core.t - may fail on recent cygwin installs.\n\nSolaris (x86, SPARC)\nOpenVMS\nAlpha (7.2 and later)\nI64 (8.2 and later)\nNetBSD\nFreeBSD\nDebian GNU/kFreeBSD\nHaiku\nIrix (6.5. What else?)\nOpenBSD\nDragonfly BSD\nMidnight BSD\nQNX Neutrino RTOS (6.5.0)\nMirOS BSD\nStratus OpenVOS (17.0 or later)\nCaveats:\n\ntimet issues that may or may not be fixed\nStratus VOS / OpenVOS\nAIX\nAndroid\nFreeMINT\nPerl now builds with FreeMiNT/Atari. It fails a few tests, that needs some investigation.\n\nThe FreeMiNT port uses GNU dld for loadable module capabilities. So ensure you have that\nlibrary installed when building perl.\n\nEOL Platforms", "subsections": [ { "name": "(Perl 5.37.1)", "content": "The following platforms were supported by a previous version of Perl but have been officially\nremoved from Perl's source code as of 5.37.1:\n\nUltrix\n" }, { "name": "(Perl 5.36)", "content": "The following platforms were supported by a previous version of Perl but have been officially\nremoved from Perl's source code as of 5.36:\n\nNetWare\nDOS/DJGPP\nAT&T UWIN\n" }, { "name": "(Perl 5.20)", "content": "The following platforms were supported by a previous version of Perl but have been officially\nremoved from Perl's source code as of 5.20:\n\nAT&T 3b1\n" }, { "name": "(Perl 5.14)", "content": "The following platforms were supported up to 5.10. They may still have worked in 5.12, but\nsupporting code has been removed for 5.14:\n\nWindows 95\nWindows 98\nWindows ME\nWindows NT4\n" }, { "name": "(Perl 5.12)", "content": "The following platforms were supported by a previous version of Perl but have been officially\nremoved from Perl's source code as of 5.12:\n\nAtari MiNT\nApollo Domain/OS\nApple Mac OS 8/9\nTenon Machten\n" } ] }, "Supported Platforms (Perl 5.8)": { "content": "As of July 2002 (the Perl release 5.8.0), the following platforms were able to build Perl\nfrom the standard source code distribution available at \n\nAIX\nBeOS\nBSD/OS (BSDi)\nCygwin\nDG/UX\nDOS DJGPP 1)\nDYNIX/ptx\nEPOC R5\nFreeBSD\nHI-UXMPP (Hitachi) (5.8.0 worked but we didn't know it)\nHP-UX\nIRIX\nLinux\nMac OS Classic\nMac OS X (Darwin)\nMPE/iX\nNetBSD\nNetWare\nNonStop-UX\nReliantUNIX (formerly SINIX)\nOpenBSD\nOpenVMS (formerly VMS)\nOpen UNIX (Unixware) (since Perl 5.8.1/5.9.0)\nOS/2\nOS/400 (using the PASE) (since Perl 5.8.1/5.9.0)\nPOSIX-BC (formerly BS2000)\nQNX\nSolaris\nSunOS 4\nSUPER-UX (NEC)\nTru64 UNIX (formerly DEC OSF/1, Digital UNIX)\nUNICOS\nUNICOS/mk\nUTS\nVOS / OpenVOS\nWin95/98/ME/2K/XP 2)\nWinCE\nz/OS (formerly OS/390)\nVM/ESA\n\n1) in DOS mode either the DOS or OS/2 ports can be used\n2) compilers: Borland, MinGW (GCC), VC6\n\nThe following platforms worked with the previous releases (5.6 and 5.7), but we did not\nmanage either to fix or to test these in time for the 5.8.0 release. There is a very good\nchance that many of these will work fine with the 5.8.0.\n\nBSD/OS\nDomainOS\nHurd\nLynxOS\nMachTen\nPowerMAX\nSCO SV\nSVR4\nUnixware\nWindows 3.1\n\nKnown to be broken for 5.8.0 (but 5.6.1 and 5.7.2 can be used):\n\nAmigaOS 3\n\nThe following platforms have been known to build Perl from source in the past (5.00503 and\nearlier), but we haven't been able to verify their status for the current release, either\nbecause the hardware/software platforms are rare or because we don't have an active champion\non these platforms--or both. They used to work, though, so go ahead and try compiling them,\nand let know of any trouble.\n\n3b1\nA/UX\nConvexOS\nCX/UX\nDC/OSx\nDDE SMES\nDOS EMX\nDynix\nEP/IX\nESIX\nFPS\nGENIX\nGreenhills\nISC\nMachTen 68k\nMPC\nNEWS-OS\nNextSTEP\nOpenSTEP\nOpus\nPlan 9\nRISC/os\nSCO ODT/OSR\nStellar\nSVR2\nTI1500\nTitanOS\nUnisys Dynix\n\nThe following platforms have their own source code distributions and binaries available via\n\n\nPerl release\n\nOS/400 (ILE) 5.00502\nTandem Guardian 5.004\n\nThe following platforms have only binaries available via\n :\n\nPerl release\n\nAcorn RISCOS 5.00502\nAOS 5.002\nLynxOS 5.00402\n\nAlthough we do suggest that you always build your own Perl from the source code, both for\nmaximal configurability and for security, in case you are in a hurry you can check\n for binary distributions.\n", "subsections": [] }, "SEE ALSO": { "content": "perlaix, perlamiga, perlbs2000, perlcygwin, perlebcdic, perlfreebsd, perlhurd, perlhpux,\nperlirix, perlmacosx, perlos2, perlos390, perlos400, perlplan9, perlqnx, perlsolaris,\nperltru64, perlunicode, perlvms, perlvos, perlwin32, and Win32.\n", "subsections": [] }, "AUTHORS / CONTRIBUTORS": { "content": "Abigail , Charles Bailey , Graham Barr\n, Tom Christiansen , Nicholas Clark ,\nThomas Dorner , Andy Dougherty , Dominic\nDunlop , Neale Ferguson , David J. Fiander\n, Paul Green , M.J.T. Guy , Jarkko\nHietaniemi , Luther Huffman , Nick Ing-Simmons\n, Andreas J. König , Markus Laker\n, Andrew M. Langmead , Lukas Mai ,\nLarry Moore , Paul Moore , Chris Nandor\n, Matthias Neeracher , Philip Newton , Gary\nNg <71564.1743@CompuServe.COM>, Tom Phoenix , André Pirard\n, Peter Prymmer , Hugo van der Sanden\n, Gurusamy Sarathy , Paul J. Schinder\n, Michael G Schwern , Dan Sugalski ,\nNathan Torkington , John Malmberg \n\nperl v5.38.2 2026-08-18 PERLPORT(1)", "subsections": [] } }, "summary": "perlport - Writing portable Perl", "flags": [], "examples": [], "see_also": [] }