joachim99@14: #!/usr/bin/perl -w joachim99@14: joachim99@14: # Check namespace cleanness of a library. joachim99@14: # Allowed symbols are passed as arguments. joachim99@14: # They may have trailing * = wildcard. joachim99@14: # Wildcards may be also specified as *::* (e.g. K*::* for all KDE classes) joachim99@14: # Symbols are listed as full function unmangled names without arguments, joachim99@14: # e.g. 'foo bar* nspace::*' allows foo(), foo(int), bar(), barbar() joachim99@14: # and all symbols in namespace/class nspace. joachim99@14: # If an argument has comma in it, it's a filename of a file containing joachim99@14: # allowed symbols, one per line. joachim99@14: joachim99@14: joachim99@14: $thisProg = "$0"; # This programs name joachim99@14: joachim99@14: $library = ""; joachim99@14: $allowed_symbols = ""; joachim99@14: $debug = 0; joachim99@14: $allowed_weak = ""; joachim99@14: $weak_specified = 0; joachim99@14: joachim99@14: while( defined( $ARGV[ 0 ] )) joachim99@14: { joachim99@14: $_ = shift; joachim99@14: if( /^--verbose$|^-v$/ ) joachim99@14: { joachim99@14: $debug = 1; joachim99@14: } joachim99@14: elsif( /^--help$|^-h$/ ) joachim99@14: { joachim99@14: print STDOUT "Usage $thisProg [OPTION] ... library [allowed symbols] ...\n", joachim99@14: "\n", joachim99@14: "Check if the given library has only allowed public symbols.\n", joachim99@14: "\n", joachim99@14: " --allowweak=[symbol] allow only these weak symbols\n", joachim99@14: " -v, --verbose verbosely list files processed\n", joachim99@14: " -h, --help print this help, then exit\n"; joachim99@14: exit 0; joachim99@14: } joachim99@14: elsif( /^--allowweak=(.*)$/ ) joachim99@14: { joachim99@14: $allowed_weak .= " " . $1; joachim99@14: $weak_specified = 1; joachim99@14: } joachim99@14: elsif( /^--allowweak$/ ) # simply list all weak joachim99@14: { joachim99@14: $allowed_weak .= " "; joachim99@14: $weak_specified = 1; joachim99@14: } joachim99@14: elsif( /^--*/ ) joachim99@14: { joachim99@14: die "Invalid argument!\n"; joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: if( ! $library ) joachim99@14: { joachim99@14: $library = $_; joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: $allowed_symbols .= " " . $_; joachim99@14: } joachim99@14: } joachim99@14: } joachim99@14: joachim99@14: if( ! $weak_specified ) joachim99@14: { joachim99@14: $allowed_weak = "*"; joachim99@14: # allow all weak symbols by default joachim99@14: # instances of templates and similar stuff - unfortunately includes also things from other libraries, joachim99@14: # so it cannot be on by default joachim99@14: } joachim99@14: joachim99@14: print STDERR "library:" . $library . "\n" if $debug; joachim99@14: print STDERR "allowed_symbols:" . $allowed_symbols . "\n" if $debug; joachim99@14: print STDERR "allowed_weak:" . $allowed_weak . "\n" if $debug; joachim99@14: joachim99@14: $default_symbols = "_fini _init"; # system symbols joachim99@14: # on my system, every .so has : joachim99@14: # A _DYNAMIC joachim99@14: # A _GLOBAL_OFFSET_TABLE_ joachim99@14: # A __bss_start joachim99@14: # A _edata joachim99@14: # A _end joachim99@14: # T _fini joachim99@14: # T _init joachim99@14: # no need to list A symbols in $default_symbols joachim99@14: joachim99@14: print STDERR "default_symbols: " . $default_symbols . "\n" if $debug; joachim99@14: joachim99@14: print STDOUT "Namespace cleanness check for " . $library . " :\n"; joachim99@14: joachim99@14: $lib_file = ""; joachim99@14: if( $library =~ /\.la$/ ) joachim99@14: { joachim99@14: # get the real library file from .la joachim99@14: open( FILEIN, $library ) || die "Couldn't open $! !\n"; joachim99@14: while( $line = ) joachim99@14: { joachim99@14: if( $line =~ /library_names=\'([^ ]*).*/o ) joachim99@14: { joachim99@14: $lib_file = $1; joachim99@14: } joachim99@14: } joachim99@14: close( FILEIN ); joachim99@14: if( ! $lib_file ) joachim99@14: { joachim99@14: print STDERR "Library file not found in .la file!\n"; joachim99@14: exit 1; joachim99@14: } joachim99@14: my $libpath = $library; joachim99@14: $libpath =~ s%[^/]*$%%; joachim99@14: if( -e $libpath . ".libs/" . $lib_file ) joachim99@14: { joachim99@14: $lib_file = $libpath . ".libs/" . $lib_file; joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: $lib_file = $libpath . $lib_file; joachim99@14: } joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: $lib_file = $library; joachim99@14: } joachim99@14: joachim99@14: print STDERR "libfile: ". $lib_file . "\n" if $debug; joachim99@14: joachim99@14: $allowed_symbols .= " " . $default_symbols; joachim99@14: joachim99@14: sub process_symbols($\@\%\@); joachim99@14: joachim99@14: @wildcards = (); joachim99@14: %exacts = (); joachim99@14: @regwildcards = (); joachim99@14: process_symbols( $allowed_symbols, @wildcards, %exacts, @regwildcards ); joachim99@14: @weak_wildcards = (); joachim99@14: %weak_exacts = (); joachim99@14: @weak_regwildcards = (); joachim99@14: process_symbols( $allowed_weak, @weak_wildcards, %weak_exacts, @weak_regwildcards ); joachim99@14: joachim99@14: # grep is for stripping not exported symbols, which don't have address (=first column) joachim99@14: $nm_command = "nm -BDCg " . $lib_file . " | grep -v '^ ' |"; joachim99@14: joachim99@14: # TODO how portable is this nmcheck stuff? joachim99@14: joachim99@14: print STDERR "nm command:" . $nm_command . "\n" if $debug; joachim99@14: joachim99@14: open( FILEIN, $nm_command ) || die "nm command failed\n"; joachim99@14: joachim99@14: my $exit_code = 0; joachim99@14: joachim99@14: while( $line = ) joachim99@14: { joachim99@14: my $type; joachim99@14: my $symbol; joachim99@14: if( $line =~ /^[^ ]* (.) (.*)$/o ) joachim99@14: { joachim99@14: $type = $1; joachim99@14: $symbol = $2; joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: die "Invalid line: " . $line . "\n"; joachim99@14: } joachim99@14: joachim99@14: print STDERR "Type: " . $type . " , symbol: " . $symbol . "\n" if $debug; joachim99@14: if( $type eq "A" ) joachim99@14: { # these should be system symbols, so ignore them joachim99@14: next; joachim99@14: } joachim99@14: joachim99@14: my $orig_symbol = $symbol; joachim99@14: joachim99@14: if( $symbol =~ /\(anonymous namespace\)/o ) joachim99@14: { # TODO tell to prefer named namespaces? (shorter symbols) joachim99@14: next; joachim99@14: } joachim99@14: joachim99@14: # strip prefixes joachim99@14: # the :: appending is to make "CLASS::*" work also for "vtable for CLASS" joachim99@14: $symbol =~ s/^typeinfo for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^typeinfo fn for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^typeinfo name for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^vtable for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^guard variable for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^reference temporary for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^VTT for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^virtual thunk \[[^\]]*\] to (.*)$/$1::/o; joachim99@14: $symbol =~ s/^non-virtual thunk \[[^\]]*\] to (.*)$/$1::/o; joachim99@14: $symbol =~ s/^covariant return thunk \[[^\]]*\] to (.*)$/$1::/o; joachim99@14: $symbol =~ s/^construction vtable thunk for (.*)$/$1::/o; joachim99@14: $symbol =~ s/^construction vtable for .*-in-(.*) [0-9]*$/$1::/o; joachim99@14: joachim99@14: # templates seem to have also return types mangled in their name, and nm prints it too joachim99@14: # they have also template arguments in the symbol joachim99@14: # get rid of both of those joachim99@14: while( $symbol =~ /<.*>/o ) joachim99@14: { joachim99@14: $symbol =~ s/<[^<>]*>//o; # strip innermost <> joachim99@14: } joachim99@14: if( $symbol !~ /operator\(\)/o ) joachim99@14: { joachim99@14: $symbol =~ s/ ?\(.*\).*$//o; # strip () and all after it joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: $symbol =~ s/(^|:| )operator\(\) ?\(.*\).*$//o; # strip () and all after it joachim99@14: } joachim99@14: $symbol =~ s/\[.*\] *$//o; # strip [in-charge] etc. joachim99@14: if( $symbol =~ /(^|:| )operator /o ) joachim99@14: { joachim99@14: $symbol =~ s/.* ([^\s]*)operator /$1/o; # strip everything before 'X::operator blah' joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: $symbol =~ s/.* ([^\s]+) *$/$1/o; # get last word (strip return type) joachim99@14: } joachim99@14: joachim99@14: # print STDERR "Processed symbol: " . $symbol . "\n" if $debug; joachim99@14: joachim99@14: my $found = 0; joachim99@14: if( $exacts{ $symbol } ) joachim99@14: { joachim99@14: $found = 1; joachim99@14: } joachim99@14: if( ! $found ) joachim99@14: { joachim99@14: for my $wild ( @wildcards ) joachim99@14: { joachim99@14: if( index( $symbol, $wild ) == 0 ) joachim99@14: { joachim99@14: $found = 1; joachim99@14: last; joachim99@14: } joachim99@14: } joachim99@14: } joachim99@14: if( ! $found ) joachim99@14: { joachim99@14: for my $wild ( @regwildcards ) joachim99@14: { joachim99@14: if( $symbol =~ /^$wild$/ ) joachim99@14: { joachim99@14: $found = 1; joachim99@14: last; joachim99@14: } joachim99@14: } joachim99@14: } joachim99@14: if( ( ! $found ) && ( $type eq "W" || $type eq "V" )) joachim99@14: { joachim99@14: if( $weak_exacts{ $symbol } ) joachim99@14: { joachim99@14: $found = 1; joachim99@14: } joachim99@14: if( ! $found ) joachim99@14: { joachim99@14: for my $wild ( @weak_wildcards ) joachim99@14: { joachim99@14: if( index( $symbol, $wild ) == 0 ) joachim99@14: { joachim99@14: $found = 1; joachim99@14: last; joachim99@14: } joachim99@14: } joachim99@14: } joachim99@14: if( ! $found ) joachim99@14: { joachim99@14: for my $wild ( @weak_regwildcards ) joachim99@14: { joachim99@14: if( $symbol =~ /^$wild$/ ) joachim99@14: { joachim99@14: $found = 1; joachim99@14: last; joachim99@14: } joachim99@14: } joachim99@14: } joachim99@14: } joachim99@14: joachim99@14: if( ! $found ) joachim99@14: { joachim99@14: print STDERR "Public symbol " . $orig_symbol . " is not allowed!\n"; joachim99@14: $exit_code = 1; joachim99@14: } joachim99@14: } joachim99@14: joachim99@14: close( FILEIN ); joachim99@14: joachim99@14: print STDOUT $exit_code == 0 ? "OK\n" : "FAILED\n"; joachim99@14: joachim99@14: exit $exit_code; joachim99@14: joachim99@14: sub process_symbols($\@\%\@) joachim99@14: { joachim99@14: my $allowed_symbols = $_[ 0 ]; joachim99@14: my $wildcards_ref = $_[ 1 ]; joachim99@14: my $exacts_ref = $_[ 2 ]; joachim99@14: my $regwildcards_ref = $_[ 3 ]; joachim99@14: joachim99@14: $allowed_symbols =~ s/^ *//o; # strip whitespace joachim99@14: $allowed_symbols =~ s/ *$//o; joachim99@14: joachim99@14: if( $allowed_symbols eq "NONE" ) joachim99@14: { joachim99@14: $allowed_symbols = ""; joachim99@14: } joachim99@14: joachim99@14: my @symbols1 = split( ' ', $allowed_symbols ); joachim99@14: my $i = 0; joachim99@14: my @symbols2 = (); joachim99@14: while( defined( $symbols1[ $i ] )) joachim99@14: { joachim99@14: my $symbol = $symbols1[ $i ]; joachim99@14: if( $symbol =~ /\./ ) # dot in name -> file joachim99@14: { joachim99@14: open( SYMIN, $symbol ) || die ( "Cannot open file " . $symbol . "!" ); joachim99@14: while( $line = ) joachim99@14: { joachim99@14: $line =~ s/^\s*//o; # strip whitespace joachim99@14: $line =~ s/\s*$//o; joachim99@14: if( $line !~ /^$/o # empty line joachim99@14: && $line !~ /^\s*#/ ) # comment line starting with # joachim99@14: { joachim99@14: $symbols2[ $#symbols2 + 1 ] = $line; joachim99@14: } joachim99@14: } joachim99@14: close( SYMIN ); joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: $symbols2[ $#symbols2 + 1 ] = $symbol; joachim99@14: } joachim99@14: $i++; joachim99@14: } joachim99@14: $i = 0; joachim99@14: while( defined( $symbols2[ $i ] )) joachim99@14: { joachim99@14: my $symbol = $symbols2[ $i ]; joachim99@14: if( $symbol =~ /__/ joachim99@14: || $symbol =~ /^_[A-Z]/ ) joachim99@14: { # ISO C++ 2.10.2 joachim99@14: die "Symbols containing a double underscore or beginning with an underscore and an upper-case letter are reserved!\n"; joachim99@14: } joachim99@14: elsif( $symbol eq "main" joachim99@14: || $symbol eq "main*" ) joachim99@14: { joachim99@14: die "Symbol main is not allowed!\n"; joachim99@14: } joachim99@14: if( $symbol =~ /^([^\*]*)\*$/o # trailing * without any * before it joachim99@14: && $symbol !~ /operator\*$/o ) joachim99@14: { joachim99@14: print STDERR "wildcard:" . $symbol . "\n" if $debug; joachim99@14: $wildcards_ref->[ $#{$wildcards_ref} + 1 ] = $1; joachim99@14: } joachim99@14: elsif( $symbol =~ /\*$/o joachim99@14: && ( $symbol =~ /\*::/o || $symbol =~ /::\*/o ) joachim99@14: && $symbol !~ /^\*/o joachim99@14: && $symbol !~ /operator\*$/o ) joachim99@14: { joachim99@14: print STDERR "regwildcard:" . $symbol . "\n" if $debug; joachim99@14: $symbol =~ s/\*/\.\*/go; # change * to .* (regexp) joachim99@14: $regwildcards_ref->[ $#{$regwildcards_ref} + 1 ] = $symbol; joachim99@14: } joachim99@14: else joachim99@14: { joachim99@14: print STDERR "exact:" . $symbol . "\n" if $debug; joachim99@14: $exacts_ref->{ $symbol } = 1; joachim99@14: } joachim99@14: $i++; joachim99@14: } joachim99@14: }