@rem = '--*-Perl-*-- @echo off if "%OS%" == "Windows_NT" goto WinNT perl -x -S "%0" %1 %2 %3 %4 %5 %6 %7 %8 %9 goto endofperl :WinNT perl -x -S %0 %* if NOT "%COMSPEC%" == "%SystemRoot%\system32\cmd.exe" goto endofperl if %errorlevel% == 9009 echo You do not have Perl in your PATH. if errorlevel 1 goto script_failed_so_exit_with_non_zero_val 2>nul goto endofperl @rem '; #!perl #line 15 eval 'exec C:\Perl\bin\perl.exe -S $0 ${1+"$@"}' if $running_under_some_shell; --$running_under_some_shell; # Version 2.0, Simon Cozens, Thu Mar 30 17:52:45 JST 2000 # Version 2.01, Tom Christiansen, Thu Mar 30 08:25:14 MST 2000 # Version 2.02, Simon Cozens, Sun Apr 16 01:53:36 JST 2000 # Version 2.03, Edward Peschko, Mon Feb 26 12:04:17 PST 2001 use strict; use warnings; use v5.6.0; use FileHandle; use Config; use Fcntl qw(:DEFAULT :flock); use File::Temp qw(tempfile); use Cwd; our $VERSION = 2.03; $| = 1; $SIG{INT} = sub { exit(); }; # exit gracefully and clean up after ourselves. use subs qw{ cc_harness check_read check_write checkopts_byte choose_backend compile_byte compile_cstyle compile_module generate_code grab_stash parse_argv sanity_check vprint yclept spawnit }; sub opt(*); # imal quoting our ($Options, $BinPerl, $Backend); our ($Input => $Output); our ($logfh); our ($cfile); # eval { main(); 1 } or die; main(); sub main { parse_argv(); check_write($Output); choose_backend(); generate_code(); run_code(); _die("XXX: Not reached?"); } ####################################################################### sub choose_backend { # Choose the backend. $Backend = 'C'; if (opt(B)) { checkopts_byte(); $Backend = 'Bytecode'; } if (opt(S) && opt(c)) { # die "$0: Do you want me to compile this or not?\n"; delete $Options->{S}; } $Backend = 'CC' if opt(O); } sub generate_code { vprint 0, "Compiling $Input"; $BinPerl = yclept(); # Calling convention for perl. if (opt(shared)) { compile_module(); } else { if ($Backend eq 'Bytecode') { compile_byte(); } else { compile_cstyle(); } } exit(0) if (!opt('r')); } sub run_code { vprint 0, "Running code"; run("$Output @ARGV"); exit(0); } # usage: vprint [level] msg args sub vprint { my $level; if (@_ == 1) { $level = 1; } elsif ($_[0] =~ /^\d$/) { $level = shift; } else { # well, they forgot to use a number; means >0 $level = 0; } my $msg = "@_"; $msg .= "\n" unless substr($msg, -1) eq "\n"; if (opt(v) > $level) { print "$0: $msg" if !opt('log'); print $logfh "$0: $msg" if opt('log'); } } sub parse_argv { use Getopt::Long; # Getopt::Long::Configure("bundling"); turned off. this is silly because # it doesn't allow for long switches. Getopt::Long::Configure("no_ignore_case"); # no difference in exists and defined for %ENV; also, a "0" # argument or a "" would not help cc, so skip unshift @ARGV, split ' ', $ENV{PERLCC_OPTS} if $ENV{PERLCC_OPTS}; $Options = {}; Getopt::Long::GetOptions( $Options, 'L:s', # lib directory 'I:s', # include directories (FOR C, NOT FOR PERL) 'o:s', # Output executable 'v:i', # Verbosity level 'e:s', # One-liner 'r', # run resulting executable 'B', # Byte compiler backend 'O', # Optimised C backend 'c', # Compile only 'h', # Help me 'S', # Dump C files 'r', # run the resulting executable 'static', # Dirty hack to enable -shared/-static 'shared', # Create a shared library (--shared for compat.) 'log:s' # where to log compilation process information ); # This is an attempt to make perlcc's arg. handling look like cc. # if ( opt('s') ) { # must quote: looks like s)foo)bar)! # if (opt('s') eq 'hared') { # $Options->{shared}++; # } elsif (opt('s') eq 'tatic') { # $Options->{static}++; # } else { # warn "$0: Unknown option -s", opt('s'); # } # } $Options->{v} += 0; helpme() if opt(h); # And exit $Output = opt(o) || 'a.out'; $Output = relativize($Output); $logfh = new FileHandle(">> " . opt('log')) if (opt('log')); if (opt(e)) { warn "$0: using -e 'code' as input file, ignoring @ARGV\n" if @ARGV; # We don't use a temporary file here; why bother? # XXX: this is not bullet proof -- spaces or quotes in name! $Input = "-e '".opt(e)."'"; # Quotes eaten by shell } else { $Input = shift @ARGV; # XXX: more files? _usage_and_die("$0: No input file specified\n") unless $Input; # DWIM modules. This is bad but necessary. $Options->{shared}++ if $Input =~ /\.pm\z/; warn "$0: using $Input as input file, ignoring @ARGV\n" if @ARGV; check_read($Input); check_perl($Input); sanity_check(); } } sub opt(*) { my $opt = shift; return exists($Options->{$opt}) && ($Options->{$opt} || 0); } sub compile_module { die "$0: Compiling to shared libraries is currently disabled\n"; } sub compile_byte { require ByteLoader; my $stash = grab_stash(); my $command = "$BinPerl -MO=Bytecode,$stash $Input"; # The -a option means we'd have to close the file and lose the # lock, which would create the tiniest of races. Instead, append # the output ourselves. vprint 1, "Writing on $Output"; my $openflags = O_WRONLY | O_CREAT; $openflags |= O_BINARY if eval { O_BINARY; 1 }; $openflags |= O_EXLOCK if eval { O_EXLOCK; 1 }; # these dies are not "$0: .... \n" because they "can't happen" sysopen(OUT, $Output, $openflags) or die "can't write to $Output: $!"; # this is blocking; hold on; why are we doing this?? # flock OUT, LOCK_EX or die "can't lock $Output: $!" # unless eval { O_EXLOCK; 1 }; truncate(OUT, 0) or die "couldn't trunc $Output: $!"; print OUT < ".c"); close $cfh; # See comment just below } vprint 1, "Writing C on $cfile"; my $max_line_len = ''; if ($^O eq 'MSWin32' && $Config{cc} =~ /^cl/i) { $max_line_len = '-l2000,'; } # This has to do the write itself, so we can't keep a lock. Life # sucks. my $command = "$BinPerl -MO=$Backend,$max_line_len$stash,-o$cfile $Input"; vprint 1, "Compiling..."; vprint 1, "Calling $command"; my ($output_r, $error_r) = spawnit($command); my @output = @$output_r; my @error = @$error_r; if (@error && $? != 0) { _die("$0: $Input did not compile, which can't happen:\n@error\n"); } cc_harness($cfile,$stash) unless opt(c); if ($lose) { vprint 2, "unlinking $cfile"; unlink $cfile or _die("can't unlink $cfile: $!"); } } sub cc_harness { my ($cfile,$stash)=@_; use ExtUtils::Embed (); my $command = ExtUtils::Embed::ccopts." -o $Output $cfile "; $command .= " -I".$_ for split /\s+/, opt(I); $command .= " -L".$_ for split /\s+/, opt(L); my @mods = split /-?u /, $stash; $command .= " ".ExtUtils::Embed::ldopts("-std", \@mods); vprint 3, "running $Config{cc} $command"; system("$Config{cc} $command"); } # Where Perl is, and which include path to give it. sub yclept { my $command = "$^X "; # DWIM the -I to be Perl, not C, include directories. if (opt(I) && $Backend eq "Bytecode") { for (split /\s+/, opt(I)) { if (-d $_) { push @INC, $_; } else { warn "$0: Include directory $_ not found, skipping\n"; } } } $command .= "-I$_ " for @INC; return $command; } # Use B::Stash to find additional modules and stuff. { my $_stash; sub grab_stash { warn "already called get_stash once" if $_stash; my $command = "$BinPerl -MB::Stash -c $Input"; # Filename here is perfectly sanitised. vprint 3, "Calling $command\n"; my ($stash_r, $error_r) = spawnit($command); my @stash = @$stash_r; my @error = @$error_r; if (@error && $? != 0) { _die("$0: $Input did not compile:\n@error\n"); } $stash[0] =~ s/,-u\//; vprint 2, "Stash: ", join " ", split /,?-u/, $stash[0]; chomp $stash[0]; return $_stash = $stash[0]; } } # Check the consistency of options if -B is selected. # To wit, (-B|-O) ==> no -shared, no -S, no -c sub checkopts_byte { _die("$0: Please choose one of either -B and -O.\n") if opt(O); if (opt(shared)) { warn "$0: Will not create a shared library for bytecode\n"; delete $Options->{shared}; } for my $o ( qw[c S] ) { if (opt($o)) { warn "$0: Compiling to bytecode is a one-pass process--", "-$o ignored\n"; delete $Options->{$o}; } } } # Check the input and output files make sense, are read/writeable. sub sanity_check { if ($Input eq $Output) { if ($Input eq 'a.out') { _die("$0: Compiling a.out is probably not what you want to do.\n"); # You fully deserve what you get now. No you *don't*. typos happen. } else { warn "$0: Will not write output on top of input file, ", "compiling to a.out instead\n"; $Output = "a.out"; } } } sub check_read { my $file = shift; unless (-r $file) { _die("$0: Input file $file is a directory, not a file\n") if -d _; unless (-e _) { _die("$0: Input file $file was not found\n"); } else { _die("$0: Cannot read input file $file: $!\n"); } } unless (-f _) { # XXX: die? don't try this on /dev/tty warn "$0: WARNING: input $file is not a plain file\n"; } } sub check_write { my $file = shift; if (-d $file) { _die("$0: Cannot write on $file, is a directory\n"); } if (-e _) { _die("$0: Cannot write on $file: $!\n") unless -w _; } unless (-w cwd()) { _die("$0: Cannot write in this directory: $!\n"); } } sub check_perl { my $file = shift; unless (-T $file) { warn "$0: Binary `$file' sure doesn't smell like perl source!\n"; print "Checking file type... "; system("file", $file); _die("Please try a perlier file!\n"); } open(my $handle, "<", $file) or _die("XXX: can't open $file: $!"); local $_ = <$handle>; if (/^#!/ && !/perl/) { _die("$0: $file is a ", /^#!\s*(\S+)/, " script, not perl\n"); } } # File spawning and error collecting sub spawnit { my ($command) = shift; my (@error,@output); my $errname; (undef, $errname) = tempfile("pccXXXXX"); { open (S_OUT, "$command 2>$errname |") or _die("$0: Couldn't spawn the compiler.\n"); @output = ; } open (S_ERROR, $errname) or _die("$0: Couldn't read the error file.\n"); @error = ; close S_ERROR; close S_OUT; unlink $errname or _die("$0: Can't unlink error file $errname"); return (\@output, \@error); } sub helpme { print "perlcc compiler frontend, version $VERSION\n\n"; { no warnings; exec "pod2usage $0"; exec "perldoc $0"; exec "pod2text $0"; } } sub relativize { my ($args) = @_; return() if ($args =~ m"^[/\\]"); return("./$args"); } sub _die { $logfh->print(@_) if opt('log'); print STDERR @_; exit(); # should die eventually. However, needed so that a 'make compile' # can compile all the way through to the end for standard dist. } sub _usage_and_die { _die(<print(interruptrun(@commands)) if (opt('log')); } sub interruptrun { my (@commands) = @_; my $command = join('', @commands); local(*FD); my $pid = open(FD, "$command |"); my $text; local($SIG{HUP}) = sub { kill 9, $pid; exit }; local($SIG{INT}) = sub { kill 9, $pid; exit }; my $needalarm = ($ENV{PERLCC_TIMEOUT} && $Config{'osname'} ne 'MSWin32' && $command =~ m"(^|\s)perlcc\s"); eval { local($SIG{ALRM}) = sub { die "INFINITE LOOP"; }; alarm($ENV{PERLCC_TIMEOUT}) if ($needalarm); $text = join('', ); alarm(0) if ($needalarm); }; if ($@) { eval { kill 'HUP', $pid }; vprint 0, "SYSTEM TIMEOUT (infinite loop?)\n"; } close(FD); return($text); } END { unlink $cfile if ($cfile && !opt(S) && !opt(c)); } __END__ =head1 NAME perlcc - generate executables from Perl programs =head1 SYNOPSIS $ perlcc hello # Compiles into executable 'a.out' $ perlcc -o hello hello.pl # Compiles into executable 'hello' $ perlcc -O file # Compiles using the optimised C backend $ perlcc -B file # Compiles using the bytecode backend $ perlcc -c file # Creates a C file, 'file.c' $ perlcc -S -o hello file # Creates a C file, 'file.c', # then compiles it to executable 'hello' $ perlcc -c out.c file # Creates a C file, 'out.c' from 'file' $ perlcc -e 'print q//' # Compiles a one-liner into 'a.out' $ perlcc -c -e 'print q//' # Creates a C file 'a.out.c' $ perlcc -r hello # compiles 'hello' into 'a.out', runs 'a.out'. $ perlcc -r hello a b c # compiles 'hello' into 'a.out', runs 'a.out'. # with arguments 'a b c' $ perlcc hello -log c # compiles 'hello' into 'a.out' logs compile # log into 'c'. =head1 DESCRIPTION F creates standalone executables from Perl programs, using the code generators provided by the L module. At present, you may either create executable Perl bytecode, using the C<-B> option, or generate and compile C files using the standard and 'optimised' C backends. The code generated in this way is not guaranteed to work. The whole codegen suite (C included) should be considered B experimental. Use for production purposes is strongly discouraged. =head1 OPTIONS =over 4 =item -LI Adds the given directories to the library search path when C code is passed to your C compiler. =item -II Adds the given directories to the include file search path when C code is passed to your C compiler; when using the Perl bytecode option, adds the given directories to Perl's include path. =item -o I Specifies the file name for the final compiled executable. =item -c I Create C code only; do not compile to a standalone binary. =item -e I Compile a one-liner, much the same as C =item -S Do not delete generated C code after compilation. =item -B Use the Perl bytecode code generator. =item -O Use the 'optimised' C code generator. This is more experimental than everything else put together, and the code created is not guaranteed to compile in finite time and memory, or indeed, at all. =item -v Increase verbosity of output; can be repeated for more verbose output. =item -r Run the resulting compiled script after compiling it. =item -log Log the output of compiling to a file rather than to stdout. =back =cut __END__ :endofperl