NAME

h2xs - convert .h C header files to Perl extensions

SYNOPSIS

h2xs [-ACOPXcdf] [-v version] [-n module_name] [-p prefix] [-s sub] [headerfile ... [extra_libraries]]

h2xs -h

DESCRIPTION

h2xs builds a Perl extension from C header files. The extension will include functions which can be used to retrieve the value of any #define statement which was in the C header files.

The module_name will be used for the name of the extension. If module_name is not supplied then the name of the first header file will be used, with the first character capitalized.

If the extension might need extra libraries, they should be included here. The extension Makefile.PL will take care of checking whether the libraries actually exist and how they should be loaded. The extra libraries should be specified in the form -lm -lposix, etc, just as on the cc command line. By default, the Makefile.PL will search through the library path determined by Configure. That path can be augmented by including arguments of the form -L/another/library/path in the extra-libraries argument.

OPTIONS

-A

Omit all autoload facilities. This is the same as -c but also removes the use AutoLoader statement from the .pm file.

-C

Omits creation of the Changes file, and adds a HISTORY section to the POD template.

-F

Additional flags to specify to C preprocessor when scanning header for function declarations. Should not be used without -x.

-M regular expression

selects functions/macros to process.

-O

Allows a pre-existing extension directory to be overwritten.

-P

Omit the autogenerated stub POD section.

-X

Omit the XS portion. Used to generate templates for a module which is not XS-based. -c and -f are implicitly enabled.

-c

Omit constant() from the .xs file and corresponding specialised AUTOLOAD from the .pm file.

-d

Turn on debugging messages.

-f

Allows an extension to be created for a header even if that header is not found in standard include directories.

-h

Print the usage, help and version for this h2xs and exit.

-n module_name

Specifies a name to be used for the extension, e.g., -n RPC::DCE

-o regular expression

Use "opaque" data type for the C types matched by the regular expression, even if these types are typedef-equivalent to types from typemaps. Should not be used without -x.

This may be useful since, say, types which are typedef-equivalent to integers may represent OS-related handles, and one may want to work with these handles in OO-way, as in $handle->do_something(). Use -o . if you want to handle all the typedefed types as opaque types.

The type-to-match is whitewashed (except for commas, which have no whitespace before them, and multiple * which have no whitespace between them).

-p prefix

Specify a prefix which should be removed from the Perl function names, e.g., -p sec_rgy_ This sets up the XS PREFIX keyword and removes the prefix from functions that are autoloaded via the constant() mechanism.

-s sub1,sub2

Create a perl subroutine for the specified macros rather than autoload with the constant() subroutine. These macros are assumed to have a return type of char *, e.g., -s sec_rgy_wildcard_name,sec_rgy_wildcard_sid.

-v version

Specify a version number for this extension. This version number is added to the templates. The default is 0.01.

-x

Automatically generate XSUBs basing on function declarations in the header file. The package C::Scan should be installed. If this option is specified, the name of the header file may look like NAME1,NAME2. In this case NAME1 is used instead of the specified string, but XSUBs are emitted only for the declarations included from file NAME2.

Note that some types of arguments/return-values for functions may result in XSUB-declarations/typemap-entries which need hand-editing. Such may be objects which cannot be converted from/to a pointer (like long long), pointers to functions, or arrays. See also the section on "LIMITATIONS of -x".

EXAMPLES

	# Default behavior, extension is Rusers
	h2xs rpcsvc/rusers

	# Same, but extension is RUSERS
	h2xs -n RUSERS rpcsvc/rusers

	# Extension is rpcsvc::rusers. Still finds <rpcsvc/rusers.h>
	h2xs rpcsvc::rusers

	# Extension is ONC::RPC.  Still finds <rpcsvc/rusers.h>
	h2xs -n ONC::RPC rpcsvc/rusers

	# Without constant() or AUTOLOAD
	h2xs -c rpcsvc/rusers

	# Creates templates for an extension named RPC
	h2xs -cfn RPC

	# Extension is ONC::RPC.
	h2xs -cfn ONC::RPC

	# Makefile.PL will look for library -lrpc in 
	# additional directory /opt/net/lib
	h2xs rpcsvc/rusers -L/opt/net/lib -lrpc

        # Extension is DCE::rgynbase
        # prefix "sec_rgy_" is dropped from perl function names
        h2xs -n DCE::rgynbase -p sec_rgy_ dce/rgynbase

        # Extension is DCE::rgynbase
        # prefix "sec_rgy_" is dropped from perl function names
        # subroutines are created for sec_rgy_wildcard_name and sec_rgy_wildcard_sid
        h2xs -n DCE::rgynbase -p sec_rgy_ \
        -s sec_rgy_wildcard_name,sec_rgy_wildcard_sid dce/rgynbase

	# Make XS without defines in perl.h, but with function declarations
	# visible from perl.h. Name of the extension is perl1.
	# When scanning perl.h, define -DEXT=extern -DdEXT= -DINIT(x)=
	# Extra backslashes below because the string is passed to shell.
	# Note that a directory with perl header files would 
	#  be added automatically to include path.
	h2xs -xAn perl1 -F "-DEXT=extern -DdEXT= -DINIT\(x\)=" perl.h

	# Same with function declaration in proto.h as visible from perl.h.
	h2xs -xAn perl2 perl.h,proto.h

	# Same but select only functions which match /^av_/
	h2xs -M '^av_' -xAn perl2 perl.h,proto.h

	# Same but treat SV* etc as "opaque" types
	h2xs -o '^[S]V \*$' -M '^av_' -xAn perl2 perl.h,proto.h

ENVIRONMENT

No environment variables are used.

AUTHOR

Larry Wall and others

SEE ALSO

perl, perlxstut, ExtUtils::MakeMaker, and AutoLoader.

DIAGNOSTICS

The usual warnings if it cannot read or write the files involved.

LIMITATIONS of -x

h2xs would not distinguish whether an argument to a C function which is of the form, say, int *, is an input, output, or input/output parameter. In particular, argument declarations of the form

    int
    foo(n)
	int *n

should be better rewritten as

    int
    foo(n)
	int &n

if n is an input parameter.

Additionally, h2xs has no facilities to intuit that a function

   int
   foo(addr,l)
	char *addr
	int   l

takes a pair of address and length of data at this address, so it is better to rewrite this function as

    int
    foo(sv)
	    SV *addr
	PREINIT:
	    STRLEN len;
	    char *s;
	CODE:
	    s = SvPV(sv,len);
	    RETVAL = foo(s, len);
	OUTPUT:
	    RETVAL

or alternately

    static int
    my_foo(SV *sv)
    {
	STRLEN len;
	char *s = SvPV(sv,len);

	return foo(s, len);
    }

    MODULE = foo	PACKAGE = foo	PREFIX = my_

    int
    foo(sv)
	SV *sv

See perlxs and perlxstut for additional details.

HISTORY

$TEMPLATE_VERSION

Original version; created by h2xs $H2XS_VERSION with options

@ARGS

EOT

my $exp_doc = <<EOD;

EXPORT

None by default.

EOD if (@const_names and not $opt_P) { $exp_doc .= <<EOD; =head2 Exportable constants

@{[join "\n  ", @const_names]}

EOD } if (defined $fdecls and @$fdecls and not $opt_P) { $exp_doc .= <<EOD; =head2 Exportable functions

EOD $exp_doc .= <<EOD if $opt_p; When accessing these functions from Perl, prefix $opt_p should be removed.

EOD $exp_doc .= <<EOD; @{[join "\n ", @known_fnames{@fnames}]}

EOD }

my $pod = <<"END" unless $opt_P; ## Below is stub documentation for your module. You better edit it! # #=head1 NAME # #$module - Perl extension for blah blah blah # #=head1 SYNOPSIS # # use $module; # blah blah blah # #=head1 DESCRIPTION # #Stub documentation for $module, created by h2xs. It looks like the #author of the extension was negligent enough to leave the stub #unedited. # #Blah blah blah. #$exp_doc$revhist #=head1 AUTHOR # #$author, $email # #=head1 SEE ALSO # #perl(1). # #=cut END

$pod =~ s/^\#//gm unless $opt_P; print PM $pod unless $opt_P;

close PM;

if( ! $opt_X ){ # print XS, unless it is disabled warn "Writing $ext$modpname/$modfname.xs\n";

print XS <<"END"; #include "EXTERN.h" #include "perl.h" #include "XSUB.h"

END if( @path_h ){ foreach my $path_h (@path_h_ini) { my($h) = $path_h; $h =~ s#^/usr/include/##; if ($^O eq 'VMS') { $h =~ s#.*vms\]#sys/# or $h =~ s#.*[:>\]]##; } print XS qq{#include <$h>\n}; } print XS "\n"; }

my %pointer_typedefs; my %struct_typedefs;

sub td_is_pointer { my $type = shift; my $out = $pointer_typedefs{$type}; return $out if defined $out; my $otype = $type; $out = ($type =~ /\*$/); # This converts only the guys which do not have trailing part in the typedef if (not $out and $typedef_rex and $type =~ s/($typedef_rex)/$typedefs_pre{$1}/go) { $type = normalize_type($type); print "Is-Pointer: Type mutation via typedefs: $otype ==> $type\n" if $opt_d; $out = td_is_pointer($type); } return ($pointer_typedefs{$otype} = $out); }

sub td_is_struct { my $type = shift; my $out = $struct_typedefs{$type}; return $out if defined $out; my $otype = $type; $out = ($type =~ /^struct\b/) && !td_is_pointer($type); # This converts only the guys which do not have trailing part in the typedef if (not $out and $typedef_rex and $type =~ s/($typedef_rex)/$typedefs_pre{$1}/go) { $type = normalize_type($type); print "Is-Struct: Type mutation via typedefs: $otype ==> $type\n" if $opt_d; $out = td_is_struct($type); } return ($struct_typedefs{$otype} = $out); }

# Some macros will bomb if you try to return them from a double-returning func. # Say, ((char *)0), or strlen (if somebody #define STRLEN strlen). # Fortunately, we can detect both these cases... sub protect_convert_to_double { my $in = shift; my $val; return '' unless defined ($val = $seen_define{$in}); return '(IV)' if $known_fnames{$val}; # OUT_t of ((OUT_t)-1): return '' unless $val =~ /^\s*(\(\s*)?\(\s*([^()]*?)\s*\)/; td_is_pointer($2) ? '(IV)' : ''; }

# For each of the generated functions, length($pref) leading # letters are already checked. Moreover, it is recommended that # the generated functions uses switch on letter at offset at least # $off + length($pref). # # The given list has length($pref) chars removed at front, it is # guarantied that $off leading chars in the rest are the same for all # elts of the list. # # Returns: how at which offset it was decided to make a switch, or -1 if none.

sub write_const;

sub write_const { my ($fh, $pref, $off, $list) = (shift,shift,shift,shift); my %leading; my $offarg = length $pref;

  if (@$list == 0) {		# Can happen on the initial iteration only
    print $fh <<"END";
static double
constant(char *name, int len, int arg)
{
    errno = EINVAL;
    return 0;
}
END
    return -1;
  }

  if (@$list == 1) {		# Can happen on the initial iteration only
    my $protect = protect_convert_to_double("$pref$list->[0]");

    print $fh <<"END";
static double
constant(char *name, int len, int arg)
{
    if (strEQ(name + $offarg, "$list->[0]")) {	/* $pref removed */
#ifdef $pref$list->[0]
	return $protect$pref$list->[0];
#else
	errno = ENOENT;
	return 0;
#endif
    }
    errno = EINVAL;
    return 0;
}
END
    return -1;
  }

  for my $n (@$list) {
    my $c = substr $n, $off, 1;
    $leading{$c} = [] unless exists $leading{$c};
    push @{$leading{$c}}, substr $n, $off + 1;
  }

  if (keys(%leading) == 1) {
    return 1 + write_const $fh, $pref, $off + 1, $list;
  }

  my $leader = substr $list->[0], 0, $off;
  foreach my $letter (keys %leading) {
    write_const $fh, "$pref$leader$letter", 0, $leading{$letter}
      if @{$leading{$letter}} > 1;
  }

  my $npref = "_$pref";
  $npref = '' if $pref eq '';

  print $fh <<"END";
static double
constant$npref(char *name, int len, int arg)
{
    errno = 0;
END

  print $fh <<"END" if $off;
    if ($offarg + $off >= len ) {
	errno = EINVAL;
	return 0;
    }
END

  print $fh <<"END";
    switch (name[$offarg + $off]) {
END

  foreach my $letter (sort keys %leading) {
    my $let = $letter;
    $let = '\0' if $letter eq '';

    print $fh <<EOP;
    case '$let':
EOP
    if (@{$leading{$letter}} > 1) {
      # It makes sense to call a function
      if ($off) {
	print $fh <<EOP;
	if (!strnEQ(name + $offarg,"$leader", $off))
	    break;
EOP
      }
      print $fh <<EOP;
	return constant_$pref$leader$letter(name, len, arg);
EOP
    }
    else {
      # Do it ourselves
      my $protect
	= protect_convert_to_double("$pref$leader$letter$leading{$letter}[0]");

      print $fh <<EOP;
	if (strEQ(name + $offarg, "$leader$letter$leading{$letter}[0]")) {	/* $pref removed */
#ifdef $pref$leader$letter$leading{$letter}[0]
	    return $protect$pref$leader$letter$leading{$letter}[0];
#else
	    goto not_there;
#endif
	}
EOP
    }
  }
  print $fh <<"END";
    }
    errno = EINVAL;
    return 0;

not_there: errno = ENOENT; return 0; }

END

}

if( ! $opt_c ) { print XS <<"END"; static int not_here(char *s) { croak("$module::%s not implemented on this architecture", s); return -1; }

END

write_const(\*XS, '', 0, \@const_names);
}

my $prefix; $prefix = "PREFIX = $opt_p" if defined $opt_p;

# Now switch from C to XS by issuing the first MODULE declaration: print XS <<"END";

MODULE = $module PACKAGE = $module $prefix

END

foreach (sort keys %const_xsub) { print XS <<"END"; char * $_()

    CODE:
#ifdef $_
	RETVAL = $_;
#else
	croak("Your vendor has not defined the $module macro $_");
#endif

    OUTPUT:
	RETVAL

END }

# If a constant() function was written then output a corresponding # XS declaration: print XS <<"END" unless $opt_c;

double constant(sv,arg) PREINIT: STRLEN len; INPUT: SV * sv char * s = SvPV(sv, len); int arg CODE: RETVAL = constant(s,len,arg); OUTPUT: RETVAL

END

my %seen_decl; my %typemap;

sub print_decl { my $fh = shift; my $decl = shift; my ($type, $name, $args) = @$decl; return if $seen_decl{$name}++; # Need to do the same for docs as well?

my @argnames = map {$_->[1]} @$args;
my @argtypes = map { normalize_type( $_->[0], 1 ) } @$args;
my @argarrays = map { $_->[4] || '' } @$args;
my $numargs = @$args;
if ($numargs and $argtypes[-1] eq '...') {
  $numargs--;
  $argnames[-1] = '...';
}
local $" = ', ';
$type = normalize_type($type, 1);

print $fh <<"EOP";

$type $name(@argnames) EOP

  for my $arg (0 .. $numargs - 1) {
    print $fh <<"EOP";
	$argtypes[$arg]	$argnames[$arg]$argarrays[$arg]
EOP
  }
}

# Should be called before any actual call to normalize_type(). sub get_typemap { # We do not want to read ./typemap by obvios reasons. my @tm = qw(../../../typemap ../../typemap ../typemap); my $stdtypemap = "$Config::Config{privlib}/ExtUtils/typemap"; unshift @tm, $stdtypemap; my $proto_re = "[" . quotemeta('\$%&*@;') . "]" ;

  # Start with useful default values
  $typemap{float} = 'T_DOUBLE';

  foreach my $typemap (@tm) {
    next unless -e $typemap ;
    # skip directories, binary files etc.
    warn " Scanning $typemap\n";
    warn("Warning: ignoring non-text typemap file '$typemap'\n"), next 
      unless -T $typemap ;
    open(TYPEMAP, $typemap) 
      or warn ("Warning: could not open typemap file '$typemap': $!\n"), next;
    my $mode = 'Typemap';
    while (<TYPEMAP>) {
      next if /^\s*\#/;
      if (/^INPUT\s*$/)   { $mode = 'Input'; next; }
      elsif (/^OUTPUT\s*$/)  { $mode = 'Output'; next; }
      elsif (/^TYPEMAP\s*$/) { $mode = 'Typemap'; next; }
      elsif ($mode eq 'Typemap') {
	next if /^\s*($|\#)/ ;
	my ($type, $image);
	if ( ($type, $image) =
	     /^\s*(.*?\S)\s+(\S+)\s*($proto_re*)\s*$/o
	     # This may reference undefined functions:
	     and not ($image eq 'T_PACKED' and $typemap eq $stdtypemap)) {
	  $typemap{normalize_type($type)} = $image;
	}
      }
    }
    close(TYPEMAP) or die "Cannot close $typemap: $!";
  }
  %std_types = %types_seen;
  %types_seen = ();
}

sub normalize_type { # Second arg: do not strip const's before \* my $type = shift; my $do_keep_deep_const = shift; # If $do_keep_deep_const this is heuristical only my $keep_deep_const = ($do_keep_deep_const ? '\b(?![^(,)]*\*)' : ''); my $ignore_mods = "(?:\\b(?:(?:__const__|const)$keep_deep_const|static|inline|__inline__)\\b\\s*)*"; if ($do_keep_deep_const) { # Keep different compiled /RExen/o separately! $type =~ s/$ignore_mods//go; } else { $type =~ s/$ignore_mods//go; } $type =~ s/([^\s\w])/ \1 /g; $type =~ s/\s+$//; $type =~ s/^\s+//; $type =~ s/\s+/ /g; $type =~ s/\* (?=\*)/*/g; $type =~ s/\. \. \./.../g; $type =~ s/ ,/,/g; $types_seen{$type}++ unless $type eq '...' or $type eq 'void' or $std_types{$type}; $type; }

my $need_opaque;

sub assign_typemap_entry { my $type = shift; my $otype = $type; my $entry; if ($tmask and $type =~ /$tmask/) { print "Type $type matches -o mask\n" if $opt_d; $entry = (td_is_struct($type) ? "T_OPAQUE_STRUCT" : "T_PTROBJ"); } elsif ($typedef_rex and $type =~ s/($typedef_rex)/$typedefs_pre{$1}/go) { $type = normalize_type $type; print "Type mutation via typedefs: $otype ==> $type\n" if $opt_d; $entry = assign_typemap_entry($type); } $entry ||= $typemap{$otype} || (td_is_struct($type) ? "T_OPAQUE_STRUCT" : "T_PTROBJ"); $typemap{$otype} = $entry; $need_opaque = 1 if $entry eq "T_OPAQUE_STRUCT"; return $entry; }

if ($opt_x) { for my $decl (@$fdecls_parsed) { print_decl(\*XS, $decl) } }

close XS;

if (%types_seen) { my $type; warn "Writing $ext$modpname/typemap\n"; open TM, ">typemap" or die "Cannot open typemap file for write: $!";

  for $type (sort keys %types_seen) {
    my $entry = assign_typemap_entry $type;
    print TM $type, "\t" x (5 - int((length $type)/8)), "\t$entry\n"
  }

  print TM <<'EOP' if $need_opaque; # Older Perls do not have correct entry
#############################################################################
INPUT
T_OPAQUE_STRUCT
	if (sv_derived_from($arg, \"${ntype}\")) {
	    STRLEN len;
	    char  *s = SvPV((SV*)SvRV($arg), len);

	    if (len != sizeof($var))
		croak(\"Size %d of packed data != expected %d\",
			len, sizeof($var));
	    $var = *($type *)s;
	}
	else
	    croak(\"$var is not of type ${ntype}\")
#############################################################################
OUTPUT
T_OPAQUE_STRUCT
	sv_setref_pvn($arg, \"${ntype}\", (char *)&$var, sizeof($var));
EOP

  close TM or die "Cannot close typemap file for write: $!";
}

} # if( ! $opt_X )

warn "Writing $ext$modpname/Makefile.PL\n"; open(PL, ">Makefile.PL") || die "Can't create $ext$modpname/Makefile.PL: $!\n";

print PL <<'END'; use ExtUtils::MakeMaker; # See lib/ExtUtils/MakeMaker.pm for details of how to influence # the contents of the Makefile that is written. END print PL "WriteMakefile(\n"; print PL " 'NAME' => '$module',\n"; print PL " 'VERSION_FROM' => '$modfname.pm', # finds \$VERSION\n"; if( ! $opt_X ){ # print C stuff, unless XS is disabled $opt_F = '' unless defined $opt_F; print PL " 'LIBS' => ['$extralibs'], # e.g., '-lm' \n"; print PL " 'DEFINE' => '$opt_F', # e.g., '-DHAVE_SOMETHING' \n"; print PL " 'INC' => '', # e.g., '-I/usr/include/other' \n"; } print PL ");\n"; close(PL) || die "Can't close $ext$modpname/Makefile.PL: $!\n";

warn "Writing $ext$modpname/test.pl\n"; open(EX, ">test.pl") || die "Can't create $ext$modpname/test.pl: $!\n"; print EX <<'_END_'; # Before `make install' is performed this script should be runnable with # `make test'. After `make install' it should work as `perl test.pl'

######################### We start with some black magic to print on failure.

# Change 1..1 below to 1..last_test_to_print . # (It may become useful if the test is moved to ./t subdirectory.)

BEGIN { $| = 1; print "1..1\n"; } END {print "not ok 1\n" unless $loaded;} _END_ print EX <<_END_; use $module; _END_ print EX <<'_END_'; $loaded = 1; print "ok 1\n";

######################### End of black magic.

# Insert your test code below (better if it prints "ok 13" # (correspondingly "not ok 13") depending on the success of chunk 13 # of the test code):

_END_ close(EX) || die "Can't close $ext$modpname/test.pl: $!\n";

unless ($opt_C) { warn "Writing $ext$modpname/Changes\n"; $" = ' '; open(EX, ">Changes") || die "Can't create $ext$modpname/Changes: $!\n"; @ARGS = map {/[\s\"\'\`\$*?^|&<>\[\]\{\}\(\)]/ ? "'$_'" : $_} @ARGS; print EX <<EOP; Revision history for Perl extension $module.

$TEMPLATE_VERSION @{[scalar localtime]} \t- original version; created by h2xs $H2XS_VERSION with options \t\t@ARGS

EOP close(EX) || die "Can't close $ext$modpname/Changes: $!\n"; }

warn "Writing $ext$modpname/MANIFEST\n"; open(MANI,'>MANIFEST') or die "Can't create MANIFEST: $!"; my @files = <*>; if (!@files) { eval {opendir(D,'.');}; unless ($@) { @files = readdir(D); closedir(D); } } if (!@files) { @files = map {chomp && $_} `ls`; } if ($^O eq 'VMS') { foreach (@files) { # Clip trailing '.' for portability -- non-VMS OSs don't expect it s%\.$%%; # Fix up for case-sensitive file systems s/$modfname/$modfname/i && next; $_ = "\U$_" if $_ eq 'manifest' or $_ eq 'changes'; $_ = 'Makefile.PL' if $_ eq 'makefile.pl'; } } print MANI join("\n",@files), "\n"; close MANI; !NO!SUBS!

close OUT or die "Can't close $file: $!"; chmod 0755, $file or die "Can't reset permissions for $file: $!\n"; exec("$Config{'eunicefix'} $file") if $Config{'eunicefix'} ne ':'; chdir $origdir;