use Config;
use Cwd;
# List explicitly here the variables you want Configure to
# generate. Metaconfig only looks for shell variables, so you
# have to mention them as if they were shell variables, not
# %Config entries. Thus you write
# $startperl
# to ensure Configure will look for $Config{startperl}.
# Wanted: $archlibexp
# This forces PL files to create target in same directory as PL file.
# This is so that make depend always knows where to find PL derivatives.
chdir dirname($0);
print "Extracting $file (with variable substitutions)\n";
# In this section, perl variables will be expanded during extraction.
# You can use $Config{...} to use Configure variables.
print OUT <<"!GROK!THIS!";
eval 'exec $Config{perlpath} -S \$0 \${1+"\$@"}'
if \$running_under_some_shell;
# In the following, perl variables are not expanded during extraction.
print OUT <<'!NO!SUBS!';
use strict;
use Config;
# Make sure read permissions for all are set:
if (defined umask && (umask() & 0444)) {
umask (umask() & ~0444);
}
getopts('Dd:rlhaQe');
my $Exit = 0;
die "Destination directory $Dest_dir doesn't exist or isn't a directory\n"
unless -d $Dest_dir;
my @isatype = split(' ',<<END);
END
my %isatype;
my $inif = 0;
my %Is_converted;
my %bad_file = ();
sub reindent($) {
my($text) = shift;
$text =~ s/\n/\n /g;
$text =~ s/ /\t/g;
$text;
}
next;
}
# Recover from header files with unbalanced cpp directives
$t = '';
$tab = 0;
# $eval_index goes into ``#line'' directives, to help locate syntax errors:
$eval_index = 1;
if ($file eq '-') {
open(IN, "-");
open(OUT, ">-");
} else {
print "$file -> $outfile\n" unless $opt_Q;
if ($file =~ m|^(.*)/|) {
$dir = $1;
mkpath "$Dest_dir/$dir";
}
if ($opt_a) { # automagic mode: locate header file in @inc_dirs
foreach (@inc_dirs) {
chdir $_;
last if -f $file;
}
}
}
print OUT
"require '_h2ph_pre.ph';\n\n",
"no warnings 'redefine';\n\n";
if (s/^\s*\#\s*//) {
if (s/^define\s+(\w+)//) {
$name = $1;
$new = '';
s/\s+$//;
s/\(\w+\s*\(\*\)\s*\(\w*\)\)\s*(-?\d+)/$1/; # (int (*)(foo_t))0
if (s/^\(([\w,\s]*)\)//) {
$args = $1;
my $proto = '() ';
if ($args ne '') {
$proto = '';
$arg =~ s/^\s*([^\s].*[^\s])\s*$/$1/;
}
$args =~ s/\b(\w)/\$$1/g;
$args = "local($args) = \@_;\n$t ";
}
s/^\s+//;
expr();
$new =~ s/(["\\])/\\$1/g; #"]);
if ($t ne '') {
$new =~ s/(['\\])/\\$1/g; #']);
if ($opt_h) {
print OUT $t,
"eval \"\\n#line $eval_index $outfile\\n\" . 'sub $name $proto\{\n$t ${args}eval q($new);\n$t}' unless defined(\&$name);\n";
$eval_index++;
} else {
print OUT $t,
"eval 'sub $name $proto\{\n$t ${args}eval q($new);\n$t}' unless defined(\&$name);\n";
}
} else {
print OUT "unless(defined(\&$name)) {\n sub $name $proto\{\n\t${args}eval q($new);\n }\n}\n";
}
%curargs = ();
} else {
s/^\s+//;
expr();
if ($t ne '') {
$new =~ s/(['\\])/\\$1/g; #']);
if ($opt_h) {
print OUT $t,"eval \"\\n#line $eval_index $outfile\\n\" . 'sub $name () {",$new,";}' unless defined(\&$name);\n";
$eval_index++;
} else {
}
} else {
# Shunt around such directives as `#define FOO FOO':
next if " \&$name" eq $new;
}
}
$incl_type = $1;
$incl = $2;
if (($incl_type eq 'include_next') ||
print OUT ($t,
"eval {\n");
$tab += 4;
print OUT ($t, "my(\@REM);\n");
if ($incl_type eq 'include_next') {
print OUT ($t,
"my(\%INCD) = map { \$INC{\$_} => 1 } ",
"(grep { \$_ eq \"$incl\" } ",
"keys(\%INC));\n");
print OUT ($t,
"\@REM = map { \"\$_/$incl\" } ",
"(grep { not exists(\$INCD{\"\$_/$incl\"})",
" and -f \"\$_/$incl\" } \@INC);\n");
} else {
print OUT ($t,
"\@REM = map { \"\$_/$incl\" } ",
"(grep {-r \"\$_/$incl\" } \@INC);\n");
}
print OUT ($t,
"require \"\$REM[0]\" if \@REM;\n");
$tab -= 4;
print OUT ($t,
"};\n");
print OUT ($t,
"warn(\$\@) if \$\@;\n");
} else {
print OUT $t,"require '$incl';\n";
}
} elsif (/^ifdef\s+(\w+)/) {
print OUT $t,"if(defined(&$1)) {\n";
$tab += 4;
} elsif (/^ifndef\s+(\w+)/) {
print OUT $t,"unless(defined(&$1)) {\n";
$tab += 4;
} elsif (s/^if\s+//) {
$new = '';
$inif = 1;
expr();
$inif = 0;
print OUT $t,"if($new) {\n";
$tab += 4;
} elsif (s/^elif\s+//) {
$new = '';
$inif = 1;
expr();
$inif = 0;
$tab -= 4;
print OUT $t,"}\n elsif($new) {\n";
$tab += 4;
} elsif (/^else/) {
$tab -= 4;
print OUT $t,"} else {\n";
$tab += 4;
} elsif (/^endif/) {
$tab -= 4;
print OUT $t,"}\n";
} elsif(/^undef\s+(\w+)/) {
print OUT $t, "undef(&$1) if defined(&$1);\n";
} elsif(/^error\s+(".*")/) {
print OUT $t, "die($1);\n";
} elsif(/^error\s+(.*)/) {
} elsif(/^warning\s+(.*)/) {
} elsif(/^ident\s+(.*)/) {
print OUT $t, "# $1\n";
}
until(/\{[^}]*\}.*;/ || /;/) {
chomp $next;
# drop "#define FOO FOO" in enums
$next =~ s/^\s*#\s*define\s+(\w+)\s+\1\s*$//;
$_ .= $next;
}
s/#\s*if.*?#\s*endif//g; # drop #ifdefs
s@/\*.*?\*/@@g;
s/\s+/ /g;
(my $enum_subs = $3) =~ s/\s//g;
my $enum_val = -1;
$enum_value =~ s/^=//;
if ($opt_h) {
print OUT ($t,
"eval(\"\\n#line $eval_index $outfile\\n",
"sub $enum_name () \{ $enum_val; \}\") ",
"unless defined(\&$enum_name);\n");
++ $eval_index;
} else {
print OUT ($t,
"eval(\"sub $enum_name () \{ $enum_val; \}\") ",
"unless defined(\&$enum_name);\n");
}
}
}
}
$next = '';
} else {
print OUT "1;\n";
}
}
warn "Was unable to convert the following files:\n";
}
exit $Exit;
sub expr {
my $joined_args;
if(keys(%curargs)) {
}
while ($_ ne '') {
s/^\&([\(a-z\)]+)/$1/i; # hack for things that take the address of
s/^(\s+)// && do {$new .= ' '; next;};
&& do {my $hex = $1;
$hex =~ s/^0+//;
# Croak if nv_preserves_uv_bits < 64 ?
# The above will produce "errorneus" code
# if the hex constant was e.g. inside UINT64_C
# macro, but then again, h2ph is an approximation.
} else {
$new .= lc("0x$hex");
}
next;};
s/^'((\\"|[^"])*)'// && do {
if ($curargs{$1}) {
$new .= "ord('\$$1')";
} else {
$new .= "ord('$1')";
}
next;
};
# replace "sizeof(foo)" with "{foo}"
# also, remove * (C dereference operator) to avoid perl syntax
# problems. Where the %sizeof array comes from is anyone's
# guess (c2ph?), but this at least avoids fatal syntax errors.
# Behavior is undefined if sizeof() delimiters are unbalanced.
# This code was modified to able to handle constructs like this:
# sizeof(*(p)), which appear in the HP-UX 10.01 header files.
s/^sizeof\s*\(// && do {
$new .= '$sizeof';
# tack { on the front, and skip it in the loop
$_ = "{" . "$_";
my $index = 1;
# find balanced closing paren
$index++;
}
# tack } on the end, replacing )
# remove pesky * operators within the sizeof argument
next;
};
# Eliminate typedefs
/\(([\w\s]+)[\*\s]*\)\s*[\w\(]/ && do {
foreach (split /\s+/, $1) { # Make sure all the words are types,
}
s/\([\w\s]+[\*\s]*\)// && next; # then eliminate them.
};
my $id = $1;
$id =~ s/(\.|(->))([^\.\-]*)/->\{$3\}/g;
while($id =~ /\[\s*([^\$\&\d\]]+)\]/) {
my($index) = $1;
$index =~ s/\s//g;
$index = "\$$index";
} else {
$index = "&$index";
}
}
$new .= " (\$$id)";
};
my $id = $1;
s/^\s+(\w+)//;
}
$new .= "\$$id";
$new .= '->' if /^[\[\{]/;
} elsif ($id eq 'defined') {
$new .= 'defined';
} elsif (/^\s*\(/) {
$new .= " &$id";
if ($new =~ /{\s*$/) {
$new .= "'$id'";
} elsif ($new =~ /\(\s*$/ && /^[\s*]*\)/) {
$new =~ s/\(\s*$//;
s/^[\s*]*\)//;
} else {
$new .= q(').$id.q(');
}
} else {
} elsif (/^\[/) {
$new .= " \$$id";
} else {
}
}
next;
};
}
}
sub next_line
{
my $file = shift;
my $pre_sub_tri_graphs = 1;
chomp $in;
next unless length $in;
while (length $in) {
if ($pre_sub_tri_graphs) {
# Preprocess all tri-graphs
# including things stuck in quoted string constants.
$in =~ s/\?\?=/#/g; # | ??=| #|
$in =~ s/\?\?\!/|/g; # | ??!| ||
$in =~ s/\?\?'/^/g; # | ??'| ^|
$in =~ s/\?\?\(/[/g; # | ??(| [|
$in =~ s/\?\?\)/]/g; # | ??)| ]|
$in =~ s/\?\?\-/~/g; # | ??-| ~|
$in =~ s/\?\?\//\\/g; # | ??/| \|
$in =~ s/\?\?</{/g; # | ??<| {|
$in =~ s/\?\?>/}/g; # | ??>| }|
}
if ($in =~ /^\#ifdef __LANGUAGE_PASCAL__/) {
# Tru64 disassembler.h evilness: mixed C and Pascal.
while (<IN>) {
last if /^\#endif/;
}
next READ;
}
while (<IN>) {
last if /^}/;
}
next READ;
}
if ($in =~ s/\\$//) { # \-newline
$out .= ' ';
next READ;
} elsif ($in =~ s/^([^"'\\\/]+)//) { # Passthrough
$out .= $1;
} elsif ($in =~ s/^(\\.)//) { # \...
$out .= $1;
} elsif ($in =~ /^'/) { # '...
$out .= $1;
} else {
next READ;
}
} elsif ($in =~ /^"/) { # "...
$out .= $1;
} else {
next READ;
}
} elsif ($in =~ s/^\/\/.*//) { # //...
# fall through
} elsif ($in =~ m/^\/\*/) { # /*...
# C comment removal adapted from perlfaq6:
if ($in =~ s/^\/\*[^*]*\*+([^\/*][^*]*\*+)*\///) {
$out .= ' ';
} else { # Incomplete /* */
next READ;
}
} elsif ($in =~ s/^(\/)//) { # /...
$out .= $1;
} elsif ($in =~ s/^([^\'\"\\\/]+)//) {
$out .= $1;
} elsif ($^O eq 'linux' &&
$in =~ s!\'T KNOW!!) {
} else {
if ($opt_e) {
warn "Cannot parse $file:\n$in\n";
$in = '';
$out = undef;
last READ;
} else {
die "Cannot parse:\n$in\n";
}
}
}
}
return $out;
}
# Handle recursive subdirectories without getting a grotesquely big stack.
# Could this be implemented using File::Find?
sub next_file
{
my $file;
while (@ARGV) {
return $file;
} elsif (-d $file) {
if ($opt_r) {
expand_glob($file);
} else {
print STDERR "Skipping directory `$file'\n";
}
} elsif ($opt_a) {
return $file;
} else {
print STDERR "Skipping `$file': not a file or directory\n";
}
}
return undef;
}
# Put all the files in $directory into @ARGV for processing.
sub expand_glob
{
my ($directory) = @_;
$directory =~ s:/$::;
foreach (readdir DIR) {
next if ($_ eq '.' or $_ eq '..');
# expand_glob() is going to be called until $ARGV[0] isn't a
# directory; so push directories, and unshift everything else.
else { unshift @ARGV, "$directory/$_" }
}
closedir DIR;
}
# Given $file, a symbolic link to a directory in the C include directory,
# make an equivalent symbolic link in $Dest_dir, if we can figure out how.
# Otherwise, just duplicate the file or directory.
sub link_if_possible
{
my ($dirlink) = @_;
my $target = eval 'readlink($dirlink)';
# The target of a parent or absolute link could leave the $Dest_dir
# hierarchy, so let's put all of the contents of $dirlink (actually,
# the contents of $target) into @ARGV; as a side effect down the
# line, $dirlink will get created as an _actual_ directory.
} else {
if (-l "$Dest_dir/$dirlink") {
unlink "$Dest_dir/$dirlink" or
print STDERR "Could not remove link $Dest_dir/$dirlink: $!\n";
}
if (eval 'symlink($target, "$Dest_dir/$dirlink")') {
print "Linking $target -> $Dest_dir/$dirlink\n";
# Make sure that the link _links_ to something:
if (! -e "$Dest_dir/$target") {
print STDERR "Could not create $Dest_dir/$target/\n";
}
} else {
print STDERR "Could not symlink $target -> $Dest_dir/$dirlink: $!\n";
}
}
}
# Push all #included files in $file onto our stack, except for STDIN
# and files we've already processed.
{
my ($file) = @_;
my $line;
return if ($file eq "-");
while (/\\$/) { # Handle continuation lines
chop $line;
}
if ($line =~ /^#\s*include\s+<(.*?)>/) {
}
}
close HEADER;
}
# Determine include directories; $Config{usrinc} should be enough for (all
# non-GCC?) C compilers, but gcc uses an additional include directory.
sub inc_dirs
{
}
# Create "_h2ph_pre.ph", if it doesn't exist or was built by a different
# version of h2ph.
{
# Increment $VERSION every time this function is modified:
my $VERSION = 2;
my $preamble = "$Dest_dir/_h2ph_pre.ph";
# Can we skip building the preamble file?
if (-r $preamble) {
# Extract version number from first line of preamble:
$line =~ /(\b\d+\b)/;
close PREAMBLE or die "Cannot close $preamble: $!";
# Don't build preamble if a compatible preamble exists:
return if $1 == $VERSION;
}
my (%define) = _extract_cc_defines();
print PREAMBLE "# This file was created by h2ph version $VERSION\n";
foreach (sort keys %define) {
if ($opt_D) {
print PREAMBLE "# $_=$define{$_}\n";
}
print PREAMBLE
"unless (defined &$_) { sub $_() { $1 } }\n\n";
} elsif ($define{$_} =~ /^\w+$/) {
print PREAMBLE
"unless (defined &$_) { sub $_() { &$define{$_} } }\n\n";
} else {
print PREAMBLE
"unless (defined &$_) { sub $_() { \"",
quotemeta($define{$_}), "\" } }\n\n";
}
}
close PREAMBLE or die "Cannot close $preamble: $!";
}
# %Config contains information on macros that are pre-defined by the
# system's compiler. We need this information to make the .ph files
# function with perl as the .h files do with cc.
{
my %define;
my $allsymbols = join " ",
# Split compiler pre-definitions into `key=value' pairs:
foreach (split /\s+/, $allsymbols) {
if ($opt_D) {
print STDERR "$_: $1 -> $2\n";
}
}
return %define;
}
1;
##############################################################################
=head1 NAME
h2ph - convert .h C header files to .ph Perl header files
=head1 SYNOPSIS
B<h2ph [-d destination directory] [-r | -a] [-l] [headerfiles]>
=head1 DESCRIPTION
I<h2ph>
converts any C header files specified to the corresponding Perl header file
format.
or
or
The output files are placed in the hierarchy rooted at Perl's
architecture dependent library directory. You can specify a different
hierarchy with a B<-d> switch.
If run with no arguments, filters standard input to standard output.
=head1 OPTIONS
=over 4
=item -d destination_dir
Put the resulting B<.ph> files beneath B<destination_dir>, instead of
beneath the default Perl library location (C<$Config{'installsitsearch'}>).
=item -r
Run recursively; if any of B<headerfiles> are directories, then run I<h2ph>
on all files in those directories (and their subdirectories, etc.). B<-r>
and B<-a> are mutually exclusive.
=item -a
Run automagically; convert B<headerfiles>, as well as any B<.h> files
which they include. This option will search for B<.h> files in all
directories which your C compiler ordinarily uses. B<-a> and B<-r> are
mutually exclusive.
=item -l
Symbolic links will be replicated in the destination directory. If B<-l>
is not specified, then links are skipped over.
=item -h
Put ``hints'' in the .ph files which will help in locating problems with
I<h2ph>. In those cases when you B<require> a B<.ph> file containing syntax
errors, instead of the cryptic
[ some error condition ] at (eval mmm) line nnn
you will see the slightly more helpful
[ some error condition ] at filename.ph line nnn
However, the B<.ph> files almost double in size when built using B<-h>.
=item -D
Include the code from the B<.h> file as a comment in the B<.ph> file.
This is primarily used for debugging I<h2ph>.
=item -Q
``Quiet'' mode; don't print out the names of the files being converted.
=back
=head1 ENVIRONMENT
No environment variables are used.
=head1 FILES
etc.
=head1 AUTHOR
Larry Wall
=head1 SEE ALSO
perl(1)
=head1 DIAGNOSTICS
The usual warnings if it can't read or write the files involved.
=head1 BUGS
Doesn't construct the %sizeof array for you.
It doesn't handle all C constructs, but it does attempt to isolate
definitions inside evals so that you can get at the definitions
that it can translate.
It's only intended as a rough tool.
You may need to dicker with the files produced.
You have to run this program by hand; it's not run as part of the Perl
installation.
Doesn't handle complicated expressions built piecemeal, a la:
enum {
FIRST_VALUE,
SECOND_VALUE,
#ifdef ABC
THIRD_VALUE
#endif
};
Doesn't necessarily locate all of your C compiler's internally-defined
symbols.
=cut
close OUT or die "Can't close $file: $!";
chdir $origdir;