X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=utils%2Fh2ph.PL;h=6a5710b80abca88a0a0c4dd36cd15daa6087335e;hb=fe0496b9253e88edaeae0264e2f55168b1a33de1;hp=c7cff6706416932b202615327ecae86b2a919481;hpb=8a5546a1774bc5278c3b557b5a3cac8da26aa636;p=p5sagit%2Fp5-mst-13.2.git diff --git a/utils/h2ph.PL b/utils/h2ph.PL index c7cff67..6a5710b 100644 --- a/utils/h2ph.PL +++ b/utils/h2ph.PL @@ -36,13 +36,21 @@ $Config{startperl} print OUT <<'!NO!SUBS!'; +use strict; + use Config; use File::Path qw(mkpath); use Getopt::Std; -getopts('Dd:rlha'); +# Make sure read permissions for all are set: +if (defined umask && (umask() & 0444)) { + umask (umask() & ~0444); +} + +getopts('Dd:rlhaQe'); +use vars qw($opt_D $opt_d $opt_r $opt_l $opt_h $opt_a $opt_Q $opt_e); die "-r and -a options are mutually exclusive\n" if ($opt_r and $opt_a); -@inc_dirs = inc_dirs() if $opt_a; +my @inc_dirs = inc_dirs() if $opt_a; my $Exit = 0; @@ -50,20 +58,35 @@ my $Dest_dir = $opt_d || $Config{installsitearch}; die "Destination directory $Dest_dir doesn't exist or isn't a directory\n" unless -d $Dest_dir; -@isatype = split(' ',<-"); } else { ($outfile = $file) =~ s/\.h$/.ph/ || next; - print "$file -> $outfile\n"; + print "$file -> $outfile\n" unless $opt_Q; if ($file =~ m|^(.*)/|) { $dir = $1; mkpath "$Dest_dir/$dir"; @@ -97,43 +120,34 @@ while (defined ($file = next_file())) { open(IN,"$file") || (($Exit = 1),(warn "Can't open $file: $!\n"),next); open(OUT,">$Dest_dir/$outfile") || die "Can't create $outfile: $!\n"; } - while () { - chop; - while (/\\$/) { - chop; - $_ .= ; - chop; - } - print OUT "# $_\n" if $opt_D; - if (s:/\*:\200:g) { - s:\*/:\201:g; - s/\200[^\201]*\201//g; # delete single line comments - if (s/\200.*//) { # begin multi-line comment? - $_ .= '/*'; - $_ .= ; - redo; - } - } + + print OUT + "require '_h2ph_pre.ph';\n\n", + "no warnings 'redefine';\n\n"; + + while (defined (local $_ = next_line($file))) { 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 = '() '; + my $proto = '() '; if ($args ne '') { - $proto = ''; - foreach $arg (split(/,\s*/,$args)) { + $proto = ''; + foreach my $arg (split(/,\s*/,$args)) { $arg =~ s/^\s*([^\s].*[^\s])\s*$/$1/; $curargs{$arg} = 1; } $args =~ s/\b(\w)/\$$1/g; - $args = "local($args) = \@_;\n$t "; + $args = "my($args) = \@_;\n$t "; } s/^\s+//; expr(); $new =~ s/(["\\])/\\$1/g; #"]); + EMIT: $new = reindent($new); $args = reindent($args); if ($t ne '') { @@ -158,6 +172,7 @@ while (defined ($file = next_file())) { $args = reindent($args); 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"; $eval_index++; @@ -165,25 +180,37 @@ while (defined ($file = next_file())) { print OUT $t,"eval 'sub $name () {",$new,";}' unless defined(\&$name);\n"; } } else { + # Shunt around such directives as `#define FOO FOO': + next if " \&$name" eq $new; + print OUT $t,"unless(defined(\&$name)) {\n sub $name () {\t",$new,";}\n}\n"; } } - } elsif (/^(include|import)\s*[<"](.*)[>"]/) { - ($incl = $2) =~ s/\.h$/.ph/; - print OUT $t,"require '$incl';\n"; - } elsif(/^include_next\s*[<"](.*)[>"]/) { - ($incl = $1) =~ s/\.h$/.ph/; + } elsif (/^(include|import|include_next)\s*[<\"](.*)[>\"]/) { + $incl_type = $1; + $incl = $2; + if (($incl_type eq 'include_next') || + ($opt_e && exists($bad_file{$incl}))) { + $incl =~ s/\.h$/.ph/; print OUT ($t, "eval {\n"); $tab += 4; $t = "\t" x ($tab / 8) . ' ' x ($tab % 8); + print OUT ($t, "my(\@REM);\n"); + if ($incl_type eq 'include_next') { print OUT ($t, "my(\%INCD) = map { \$INC{\$_} => 1 } ", - "(grep { \$_ eq \"$incl\" } keys(\%INC));\n"); + "(grep { \$_ eq \"$incl\" } ", + "keys(\%INC));\n"); print OUT ($t, - "my(\@REM) = map { \"\$_/$incl\" } ", + "\@REM = map { \"\$_/$incl\" } ", "(grep { not exists(\$INCD{\"\$_/$incl\"})", - "and -f \"\$_/$incl\" } \@INC);\n"); + " and -f \"\$_/$incl\" } \@INC);\n"); + } else { + print OUT ($t, + "\@REM = map { \"\$_/$incl\" } ", + "(grep {-r \"\$_/$incl\" } \@INC);\n"); + } print OUT ($t, "require \"\$REM[0]\" if \@REM;\n"); $tab -= 4; @@ -192,6 +219,10 @@ while (defined ($file = next_file())) { "};\n"); print OUT ($t, "warn(\$\@) if \$\@;\n"); + } else { + $incl =~ s/\.h$/.ph/; + print OUT $t,"require '$incl';\n"; + } } elsif (/^ifdef\s+(\w+)/) { print OUT $t,"if(defined(&$1)) {\n"; $tab += 4; @@ -230,27 +261,36 @@ while (defined ($file = next_file())) { 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+(.*)/) { - print OUT $t, "die(\"$1\");\n"; + print OUT $t, "die(\"", quotemeta($1), "\");\n"; } elsif(/^warning\s+(.*)/) { - print OUT $t, "warn(\"$1\");\n"; + print OUT $t, "warn(\"", quotemeta($1), "\");\n"; } elsif(/^ident\s+(.*)/) { print OUT $t, "# $1\n"; } - } elsif(/^\s*(typedef\s*)?enum\b/) { - until(/\}.*?;/) { - chomp($next = ); + } elsif (/^\s*(typedef\s*)?enum\s*(\s+[a-zA-Z_]\w*\s*)?/) { # { for vi + until(/\{[^}]*\}.*;/ || /;/) { + last unless defined ($next = next_line($file)); + chomp $next; + # drop "#define FOO FOO" in enums + $next =~ s/^\s*#\s*define\s+(\w+)\s+\1\s*$//; + # #defines in enums (aliases) + $next =~ s/^\s*#\s*define\s+(\w+)\s+(\w+)\s*$/$1 = $2,/; $_ .= $next; print OUT "# $next\n" if $opt_D; } + s/#\s*if.*?#\s*endif//g; # drop #ifdefs s@/\*.*?\*/@@g; s/\s+/ /g; - /^\s?(typedef\s?)?enum\s?([a-zA-Z_]\w*)?\s?\{(.*)\}\s?([a-zA-Z_]\w*)?\s?;/; - ($enum_subs = $3) =~ s/\s//g; - @enum_subs = split(/,/, $enum_subs); - $enum_val = -1; - for $enum (@enum_subs) { - ($enum_name, $enum_value) = $enum =~ /^([a-zA-Z_]\w*)(=.+)?$/; + next unless /^\s?(typedef\s?)?enum\s?([a-zA-Z_]\w*)?\s?\{(.*)\}\s?([a-zA-Z_]\w*)?\s?;/; + (my $enum_subs = $3) =~ s/\s//g; + my @enum_subs = split(/,/, $enum_subs); + my $enum_val = -1; + foreach my $enum (@enum_subs) { + my ($enum_name, $enum_value) = $enum =~ /^([a-zA-Z_]\w*)(=.+)?$/; + $enum_name or next; $enum_value =~ s/^=//; $enum_val = (length($enum_value) ? $enum_value : $enum_val + 1); if ($opt_h) { @@ -265,33 +305,119 @@ while (defined ($file = next_file())) { "unless defined(\&$enum_name);\n"); } } + } elsif (/^(?:__extension__\s+)?(?:extern|static)\s+(?:__)?inline(?:__)?\s+/ + and !/;\s*$/ and !/{\s*}\s*$/) + { # { for vi + # This is a hack to parse the inline functions in the glibc headers. + # Warning: massive kludge ahead. We suppose inline functions + # are mainly constructed like macros. + while (1) { + last unless defined ($next = next_line($file)); + chomp $next; + undef $_, last if $next =~ /__THROW\s*;/ + or $next =~ /^(__extension__|extern|static)\b/; + $_ .= " $next"; + print OUT "# $next\n" if $opt_D; + last if $next =~ /^}|^{.*}\s*$/; + } + next if not defined; # because it's only a prototype + s/\b(__extension__|extern|static|(?:__)?inline(?:__)?)\b//g; + # violently drop #ifdefs + s/#\s*if.*?#\s*endif//g + and print OUT "# some #ifdef were dropped here -- fill in the blanks\n"; + if (s/^(?:\w|\s|\*)*\s(\w+)\s*//) { + $name = $1; + } else { + warn "name not found"; next; # shouldn't occur... + } + my @args; + if (s/^\(([^()]*)\)\s*(\w+\s*)*//) { + for my $arg (split /,/, $1) { + if ($arg =~ /(\w+)\s*$/) { + $curargs{$1} = 1; + push @args, $1; + } + } + } + $args = ( + @args + ? "my(" . (join ',', map "\$$_", @args) . ") = \@_;\n$t " + : "" + ); + my $proto = @args ? '' : '() '; + $new = ''; + s/\breturn\b//g; # "return" doesn't occur in macros usually... + expr(); + # try to find and perlify local C variables + our @local_variables = (); # needs to be a our(): (?{...}) bug workaround + { + use re "eval"; + my $typelist = join '|', keys %isatype; + $new =~ s[' + (?:(?:un)?signed\s+)? + (?:long\s+)? + (?:$typelist)\s+ + (\w+) + (?{ push @local_variables, $1 }) + '] + [my \$$1]gx; + $new =~ s[' + (?:(?:un)?signed\s+)? + (?:long\s+)? + (?:$typelist)\s+ + ' \s+ &(\w+) \s* ; + (?{ push @local_variables, $1 }) + ] + [my \$$1;]gx; + } + $new =~ s/&$_\b/\$$_/g for @local_variables; + $new =~ s/(["\\])/\\$1/g; #"]); + # now that's almost like a macro (we hope) + goto EMIT; } } - print OUT "1;\n"; + $Is_converted{$file} = 1; + if ($opt_e && exists($bad_file{$file})) { + unlink($Dest_dir . '/' . $outfile); + $next = ''; + } else { + print OUT "1;\n"; + queue_includes_from($file) if $opt_a; + } +} - $is_converted{$file} = 1; - queue_includes_from($file) if ($opt_a); +if ($opt_e && (scalar(keys %bad_file) > 0)) { + warn "Was unable to convert the following files:\n"; + warn "\t" . join("\n\t",sort(keys %bad_file)) . "\n"; } exit $Exit; -sub reindent($) { - my($text) = shift; - $text =~ s/\n/\n /g; - $text =~ s/ /\t/g; - $text; -} - sub expr { + $new = '"(assembly code)"' and return if /\b__asm__\b/; # freak out. + my $joined_args; if(keys(%curargs)) { - my($joined_args) = join('|', keys(%curargs)); + $joined_args = join('|', keys(%curargs)); } while ($_ ne '') { s/^\&\&// && do { $new .= " &&"; next;}; # handle && operator s/^\&([\(a-z\)]+)/$1/i; # hack for things that take the address of s/^(\s+)// && do {$new .= ' '; next;}; - s/^(0X[0-9A-F]+)[UL]*//i && do {$new .= lc($1); next;}; - s/^(-?\d+\.\d+E[-+]\d+)F?//i && do {$new .= $1; next;}; + s/^0X([0-9A-F]+)[UL]*//i + && do {my $hex = $1; + $hex =~ s/^0+//; + if (length $hex > 8 && !$Config{use64bitint}) { + # Croak if nv_preserves_uv_bits < 64 ? + $new .= hex(substr($hex, -8)) + + 2**32 * hex(substr($hex, 0, -8)); + # 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/^(-?\d+\.\d+E[-+]?\d+)[FL]?//i && do {$new .= $1; next;}; s/^(\d+)\s*[LU]*//i && do {$new .= $1; next;}; s/^("(\\"|[^"])*")// && do {$new .= $1; next;}; s/^'((\\"|[^"])*)'// && do { @@ -329,14 +455,20 @@ sub expr { }; # Eliminate typedefs /\(([\w\s]+)[\*\s]*\)\s*[\w\(]/ && do { + my $doit = 1; foreach (split /\s+/, $1) { # Make sure all the words are types, - last unless ($isatype{$_} or $_ eq 'struct'); + unless($isatype{$_} or $_ eq 'struct' or $_ eq 'union'){ + $doit = 0; + last; + } + } + if( $doit ){ + s/\([\w\s]+[\*\s]*\)// && next; # then eliminate them. } - s/\([\w\s]+[\*\s]*\)// && next; # then eliminate them. }; # struct/union member, including arrays: s/^([_A-Z]\w*(\[[^\]]+\])?((\.|->)[_A-Z]\w*(\[[^\]]+\])?)+)//i && do { - $id = $1; + my $id = $1; $id =~ s/(\.|(->))([^\.\-]*)/->\{$3\}/g; $id =~ s/\b([^\$])($joined_args)/$1\$$2/g if length($joined_args); while($id =~ /\[\s*([^\$\&\d\]]+)\]/) { @@ -352,8 +484,8 @@ sub expr { $new .= " (\$$id)"; }; s/^([_a-zA-Z]\w*)// && do { - $id = $1; - if ($id eq 'struct') { + my $id = $1; + if ($id eq 'struct' || $id eq 'union') { s/^\s+(\w+)//; $id .= ' ' . $1; $isatype{$id} = 1; @@ -366,8 +498,8 @@ sub expr { $new .= '->' if /^[\[\{]/; } elsif ($id eq 'defined') { $new .= 'defined'; - } elsif (/^\(/) { - s/^\((\w),/("$1",/ if $id =~ /^_IO[WR]*$/i; # cheat + } elsif (/^\s*\(/) { + s/^\s*\((\w),/("$1",/ if $id =~ /^_IO[WR]*$/i; # cheat $new .= " &$id"; } elsif ($isatype{$id}) { if ($new =~ /{\s*$/) { @@ -394,6 +526,103 @@ sub expr { } +sub next_line +{ + my $file = shift; + my ($in, $out); + my $pre_sub_tri_graphs = 1; + + READ: while (not eof IN) { + $in .= ; + 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; # | ??/| \| + $in =~ s/\?\?/}/g; # | ??>| }| + } + if ($in =~ /^\#ifdef __LANGUAGE_PASCAL__/) { + # Tru64 disassembler.h evilness: mixed C and Pascal. + while () { + last if /^\#endif/; + } + $in = ""; + next READ; + } + if ($in =~ /^extern inline / && # Inlined assembler. + $^O eq 'linux' && $file =~ m!(?:^|/)asm/[^/]+\.h$!) { + while () { + last if /^}/; + } + $in = ""; + next READ; + } + if ($in =~ s/\\$//) { # \-newline + $out .= ' '; + next READ; + } elsif ($in =~ s/^([^"'\\\/]+)//) { # Passthrough + $out .= $1; + } elsif ($in =~ s/^(\\.)//) { # \... + $out .= $1; + } elsif ($in =~ /^'/) { # '... + if ($in =~ s/^('(\\.|[^'\\])*')//) { + $out .= $1; + } else { + next READ; + } + } elsif ($in =~ /^"/) { # "... + if ($in =~ s/^("(\\.|[^"\\])*")//) { + $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' && + $file =~ m!(?:^|/)linux/byteorder/pdp_endian\.h$! && + $in =~ s!\'T KNOW!!) { + $out =~ s!I DON$!I_DO_NOT_KNOW!; + } else { + if ($opt_e) { + warn "Cannot parse $file:\n$in\n"; + $bad_file{$file} = 1; + $in = ''; + $out = undef; + last READ; + } else { + die "Cannot parse:\n$in\n"; + } + } + } + + last READ if $out =~ /\S/; + } + + return $out; +} + + # Handle recursive subdirectories without getting a grotesquely big stack. # Could this be implemented using File::Find? sub next_file @@ -494,7 +723,7 @@ sub queue_includes_from } if ($line =~ /^#\s*include\s+<(.*?)>/) { - push(@ARGV, $1) unless $is_converted{$1}; + push(@ARGV, $1) unless $Is_converted{$1}; } } close HEADER; @@ -512,6 +741,74 @@ sub inc_dirs } +# Create "_h2ph_pre.ph", if it doesn't exist or was built by a different +# version of h2ph. +sub build_preamble_if_necessary +{ + # 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: + open PREAMBLE, $preamble or die "Cannot open $preamble: $!"; + my $line = ; + $line =~ /(\b\d+\b)/; + close PREAMBLE or die "Cannot close $preamble: $!"; + + # Don't build preamble if a compatible preamble exists: + return if $1 == $VERSION; + } + + my (%define) = _extract_cc_defines(); + + open PREAMBLE, ">$preamble" or die "Cannot open $preamble: $!"; + print PREAMBLE "# This file was created by h2ph version $VERSION\n"; + + foreach (sort keys %define) { + if ($opt_D) { + print PREAMBLE "# $_=$define{$_}\n"; + } + + if ($define{$_} =~ /^(\d+)U?L{0,2}$/i) { + 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. +sub _extract_cc_defines +{ + my %define; + my $allsymbols = join " ", + @Config{'ccsymbols', 'cppsymbols', 'cppccsymbols'}; + + # Split compiler pre-definitions into `key=value' pairs: + foreach (split /\s+/, $allsymbols) { + /(.+?)=(.+)/ and $define{$1} = $2; + + if ($opt_D) { + print STDERR "$_: $1 -> $2\n"; + } + } + + return %define; +} + + 1; ############################################################################## @@ -536,6 +833,10 @@ It is most easily run while in /usr/include: or + cd /usr/include; h2ph * sys/* arpa/* netinet/* + +or + cd /usr/include; h2ph -r -l . The output files are placed in the hierarchy rooted at Perl's @@ -590,6 +891,10 @@ However, the B<.ph> files almost double in size when built using B<-h>. Include the code from the B<.h> file as a comment in the B<.ph> file. This is primarily used for debugging I. +=item -Q + +``Quiet'' mode; don't print out the names of the files being converted. + =back =head1 ENVIRONMENT @@ -626,6 +931,22 @@ 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 !NO!SUBS!