X-Git-Url: https://git.xonotic.org/?p=xonotic%2Fxonotic.git;a=blobdiff_plain;f=misc%2Ftools%2Fprogs-analyzer.pl;h=af97908d935299d04f3013be0cee8262b0a59b75;hp=4dbd7d50d006fe634e2f21fe9899f8f82a22dcf2;hb=14c7702062d1a41a661d5b10f710eed167e00250;hpb=ef11020481270e2caa513958be81483a7f5cee6e diff --git a/misc/tools/progs-analyzer.pl b/misc/tools/progs-analyzer.pl old mode 100644 new mode 100755 index 4dbd7d50..af97908d --- a/misc/tools/progs-analyzer.pl +++ b/misc/tools/progs-analyzer.pl @@ -1,6 +1,9 @@ +#!/usr/bin/perl + use strict; use warnings; use Digest::SHA; +use Carp; sub id() { @@ -120,6 +123,14 @@ sub checkop($) { return { a => 'inglobalvec', isreturn => 1 }; } + if($op eq 'STATE') + { + return { a => 'inglobal', b => 'inglobalfunc' }; + } + if($op =~ /^INVALID#/) + { + return { isinvalid => 1 }; + } return { a => 'inglobal', b => 'inglobal', c => 'outglobal' }; } @@ -180,13 +191,26 @@ use constant DFUNCTION_T => [ [uchar8 => 'parm_size'], ]; +use constant LNOHEADER_T => [ + [int => 'lnotype'], + [int => 'version'], + [int => 'numglobaldefs'], + [int => 'numglobals'], + [int => 'numfielddefs'], + [int => 'numstatements'], +]; + +use constant LNO_T => [ + [int => 'v'], +]; + sub get_section($$$) { my ($fh, $start, $len) = @_; seek $fh, $start, 0 or die "seek: $!"; $len == read $fh, my $buf, $len - or die "short read"; + or die "short read from $start length $len (malformed progs header)"; return $buf; } @@ -214,7 +238,7 @@ sub parse_section($$$$$) my @out = map { $itemlen == read $fh, my $buf, $itemlen - or die "short read"; + or die "short read from $start length $cnt * $itemlen $(malformed progs header)"; my %h = (); @h{@packnames} = unpack $packspec, $buf; $h{$_->[1]} = TYPES->{$_->[0]}->[2]->($h{$_->[1]}) @@ -273,13 +297,8 @@ sub run_nfa($$$$$$) { my $func = $s->{a}; my $funcid = $progs->{globals}[$func]{v}{int}; - my $funcobj = $progs->{functions}[$funcid]; - if($funcobj && $funcobj->{first_statement} < 0) # builtin - { - my $def = $progs->{globaldef_byoffset}->($func); - last - if $def->{debugname} eq '_error'; - } + last + if $progs->{builtins}{error}{$funcid}; $ip += 1; } elsif($c->{isjump}) @@ -317,28 +336,32 @@ sub run_nfa($$$$$$) $nfa->($ip, $copy_handler->($state)); } -sub get_constant($$) +sub get_constant($$$) { - my ($progs, $g) = @_; - if($g->{int} == 0) - { - return undef; - } - elsif($g->{int} > 0 && $g->{int} < 16777216) - { - if($g->{int} < length $progs->{strings} && $g->{int} > 0) - { - return str($progs->{getstring}->($g->{int})); - } - else - { - return $g->{int} . "i"; - } - } - else - { - return $g->{float}; + my ($progs, $g, $type) = @_; + + if (!defined $type) { + $type = 'float'; + $type = 'int' + if $g->{int} > 0 && $g->{int} < 8388608; + $type = 'string' + if $g->{int} > 0 && $g->{int} < length $progs->{strings}; } + + return str($progs->{getstring}->($g->{int})) + if $type eq 'string'; + return $g->{float} + if $type eq 'float'; + return "'$g->{float} _ _'" + if $type eq 'vector'; + return "entity $g->{int}" + if $type eq 'entity'; + return ".$progs->{entityfieldnames}[$g->{int}][0]" + if $type eq 'field' and defined $progs->{entityfieldnames}[$g->{int}][0]; + return "$g->{int}i" + if $type eq 'int'; + + return "$type($g->{int})"; } use constant PRE_MARK_STATEMENT => ""; @@ -358,16 +381,39 @@ sub str($) return "\"$str\""; } +sub debugpos($$$) { + my ($progs, $func, $ip) = @_; + my $s = $func->{debugname}; + if ($progs->{cno}) { + my $column = $progs->{cno}[$ip]{v}; + $s =~ s/:/:$column:/; + } + if ($progs->{lno}) { + my $line = $progs->{lno}[$ip]{v}; + $s =~ s/:/:$line:/; + } + return $s; +} + sub disassemble_function($$;$) { my ($progs, $func, $highlight) = @_; print "$func->{debugname}:\n"; + if($func->{first_statement} < 0) # builtin + { + printf INSTRUCTION_FORMAT, '', '', '.BUILTIN'; + printf OPERAND_FORMAT, -$func->{first_statement}; + print INSTRUCTION_SEPARATOR; + return; + } + my $initializer = sub { my ($ofs) = @_; - my $g = get_constant($progs, $progs->{globals}[$ofs]{v}); + # TODO: Can we know its type? + my $g = get_constant($progs, $progs->{globals}[$ofs]{v}, undef); print " = $g" if defined $g; }; @@ -411,7 +457,6 @@ sub disassemble_function($$;$) my $getname = sub { my ($ofs) = @_; - $ofs &= 0xFFFF; return $override_locals{$ofs} if exists $override_locals{$ofs}; my $def = $progs->{globaldef_byoffset}->($ofs); @@ -456,25 +501,11 @@ sub disassemble_function($$;$) } }; - my %statements = (); - my %come_from = (); - run_nfa $progs, $func->{first_statement}, "", id, nfa_default_state_checker, - sub - { - my ($ip, $state, $s, $c) = @_; - ++$statements{$ip}; - - if(my $j = $c->{isjump}) - { - my $t = $ip + $s->{$j}; - $come_from{$t}{$ip} = $c->{isconditional}; - } - - return 0; - }; + my $statements = $func->{statements}; + my $come_from = $func->{come_from}; my $ipprev = undef; - for my $ip(sort { $a <=> $b } keys %statements) + for my $ip(sort { $a <=> $b } keys %$statements) { if($ip == $func->{first_statement}) { @@ -487,7 +518,7 @@ sub disassemble_function($$;$) printf OPERAND_FORMAT, $ip - $ipprev - 1; print INSTRUCTION_SEPARATOR; } - if(my $cf = $come_from{$ip}) + if(my $cf = $come_from->{$ip}) { printf INSTRUCTION_FORMAT, $ip, '', '.XREF'; my $cnt = 0; @@ -504,11 +535,26 @@ sub disassemble_function($$;$) my $ipt = $progs->{statements}[$ip]; my $opprop = checkop $op; + if($highlight and $highlight->{$ip}) + { + for(values %{$highlight->{$ip}}) + { + for(sort keys %$_) + { + print PRE_MARK_STATEMENT; + printf INSTRUCTION_FORMAT, '', '', '.WARN'; + my $pos = debugpos $progs, $func, $ip; + printf OPERAND_FORMAT, "$_ (in $pos)"; + print INSTRUCTION_SEPARATOR; + } + } + } + print PRE_MARK_STATEMENT if $highlight and $highlight->{$ip}; my $showip = $opprop->{isjump}; - printf INSTRUCTION_FORMAT, $showip ? $ip : '', $highlight->{$ip} ? "" : "", $op; + printf INSTRUCTION_FORMAT, $showip ? $ip : '', $highlight->{$ip} ? '' : '', $op; my $cnt = 0; for my $o(qw(a b c)) @@ -535,7 +581,6 @@ sub find_uninitialized_locals($$) { my ($progs, $func) = @_; - return if $func->{first_statement} < 0; # builtin @@ -555,51 +600,38 @@ sub find_uninitialized_locals($$) for(keys %{$progs->{temps}}) { - $watchme{$_} = WATCHME_T | WATCHME_X - if not exists $watchme{$_}; + next + if exists $watchme{$_}; + if($progs->{temps}{$_}) + { + # shared temp + $watchme{$_} = WATCHME_T | WATCHME_X + } + else + { + # unique temp + $watchme{$_} = WATCHME_X + } } + $watchme{$_} |= WATCHME_R + for keys %{$func->{globals_read}}; + $watchme{$_} |= WATCHME_W + for keys %{$func->{globals_written}}; + my %write_places = (); - run_nfa $progs, $func->{first_statement}, "", id, nfa_default_state_checker, - sub + for my $ofs(keys %{$func->{globals_written}}) + { + next + unless exists $watchme{$ofs} and $watchme{$ofs} & WATCHME_X; + for my $ip(keys %{$func->{globals_written}{$ofs}}) { - my ($ip, $state, $s, $c) = @_; - for(qw(a b c)) + for my $op(keys %{$func->{globals_written}{$ofs}{$ip}}) { - my $type = $c->{$_}; - next - unless defined $type; - - my $ofs = $s->{$_}; - if($type eq 'inglobal' || $type eq 'inglobalfunc') - { - $watchme{$ofs} |= WATCHME_R; - } - elsif($type eq 'inglobalvec') - { - $watchme{$ofs} |= WATCHME_R; - $watchme{$ofs+1} |= WATCHME_R; - $watchme{$ofs+2} |= WATCHME_R; - } - elsif($type eq 'outglobal') - { - $watchme{$ofs} |= WATCHME_W; - $write_places{$ip}{$_} = [$ofs] - if $watchme{$ofs} & WATCHME_X; - } - elsif($type eq 'outglobalvec') - { - $watchme{$ofs} |= WATCHME_W; - $watchme{$ofs+1} |= WATCHME_W; - $watchme{$ofs+2} |= WATCHME_W; - my @l = grep { $watchme{$_} & WATCHME_X } $ofs .. ($ofs+2); - $write_places{$ip}{$_} = \@l - if @l; - } + push @{$write_places{$ip}{$op}}, $ofs; } - - return 0; - }; + } + } for(keys %watchme) { @@ -624,32 +656,6 @@ sub find_uninitialized_locals($$) $watchme{$_}{valid} = [1, undef, undef] if defined $watchme{$_}; } - # an initial run of STORE instruction is for receiving extra parameters - # (beyond 8). Only possible if the function is declared as having 8 params. - # Extra parameters behave otherwise like temps, but are initialized at - # startup. - for($func->{first_statement} .. (@{$progs->{statements}}-1)) - { - my $s = $progs->{statements}[$_]; - if($s->{op} eq 'STORE_V') - { - $watchme{$s->{a}}{valid} = [1, undef, undef] - if defined $watchme{$s->{a}}; - $watchme{$s->{a}+1}{valid} = [1, undef, undef] - if defined $watchme{$s->{a}+1}; - $watchme{$s->{a}+2}{valid} = [1, undef, undef] - if defined $watchme{$s->{a}+2}; - } - elsif($s->{op} =~ /^STORE_/) - { - $watchme{$s->{a}}{valid} = [1, undef, undef] - if defined $watchme{$s->{a}}; - } - else - { - last; - } - } my %warned = (); my %ip_seen = (); @@ -694,8 +700,22 @@ sub find_uninitialized_locals($$) my ($ip, $state, $s, $c) = @_; my $op = $s->{op}; + # QCVM BUG: RETURN always takes vector, there is no float equivalent my $return_hack = $c->{isreturn} // 0; + if($op eq 'STORE_V') + { + # COMPILER BUG of QCC: params are always copied using STORE_V + if($s->{b} >= 4 && $s->{b} < 28) # parameter range + { + $return_hack = 1; + } + } + + if($c->{isinvalid}) + { + ++$warned{$ip}{''}{"Invalid opcode"}; + } for(qw(a b c)) { my $type = $c->{$_}; @@ -714,18 +734,18 @@ sub find_uninitialized_locals($$) my $valid = $state->{$ofs}{valid}; if($valid->[0] == 0) { - if($return_hack <= 2 and ($op ne 'OR' && $op ne 'AND' || $_ ne 'b')) # fteqcc logicops cause this + # COMPILER BUG of FTEQCC: AND and OR may take uninitialized as second argument (logicops) + if($return_hack <= 2 and ($op ne 'OR' && $op ne 'AND' || $_ ne 'b')) { - print "; Use of uninitialized value $ofs in $func->{debugname} at $ip.$_\n"; - ++$warned{$ip}{$_}; + ++$warned{$ip}{$_}{"Use of uninitialized value"}; } } elsif($valid->[0] < 0) { - if($return_hack <= 2 and ($op ne 'OR' && $op ne 'AND' || $_ ne 'b')) # fteqcc logicops cause this + # COMPILER BUG of FTEQCC: AND and OR may take uninitialized as second argument (logicops) + if($return_hack <= 2 and ($op ne 'OR' && $op ne 'AND' || $_ ne 'b')) { - print "; Use of temporary $ofs across CALL in $func->{debugname} at $ip.$_\n"; - ++$warned{$ip}{$_}; + ++$warned{$ip}{$_}{"Use of temporary across CALL"}; } } else @@ -764,6 +784,13 @@ sub find_uninitialized_locals($$) $write->($ofs+1); $write->($ofs+2); } + elsif($type eq 'ipoffset') + { + ++$warned{$ip}{$_}{"Endless loop"} + if $ofs == 0; + ++$warned{$ip}{$_}{"No-operation jump"} + if $ofs == 1; + } } if($c->{iscall}) { @@ -852,12 +879,63 @@ sub find_uninitialized_locals($$) if(!$isread) { - print "; Value is never used in $func->{debugname} at $ip.$operand\n"; - ++$warned{$ip}{$operand}; + ++$warned{$ip}{$operand}{"Value is never used"}; } } } - + + my %solid_seen = (); + run_nfa $progs, $func->{first_statement}, do { my $state = -1; \$state; }, + sub + { + my $state = ${$_[0]}; + return \$state; + }, + sub + { + my ($ip, $state) = @_; + return $solid_seen{"$ip $$state"}++; + }, + sub + { + my ($ip, $state, $s, $c) = @_; + + if($s->{op} eq 'ADDRESS') + { + my $field_ptr_ofs = $s->{b}; + my $def = $progs->{globaldef_byoffset}->($field_ptr_ofs); + use Data::Dumper; + if (($def->{globaltype} eq 'read_only' || $def->{globaltype} eq 'const') && + grep { $_ eq 'solid' } @{$progs->{entityfieldnames}[$progs->{globals}[$field_ptr_ofs]{v}{int}]}) + { + # Taking address of 'solid' for subsequent write! + # TODO check if this address is then actually used in STOREP. + $$state = $ip; + } + } + + if($c->{iscall}) + { + # TODO check if the entity passed is actually the one on which solid was set. + my $func = $s->{a}; + my $funcid = $progs->{globals}[$func]{v}{int}; + if ($progs->{builtins}{setmodel}{$funcid} || $progs->{builtins}{setsize}{$funcid}) + { + # All is clean. + $$state = -1; + } + } + + if($c->{isreturn}) + { + if ($$state >= 0) { + ++$warned{$$state}{''}{"Changing .solid without setmodel/setsize breaks area grid linking in Quake"}; + } + } + + return 0; + }; + disassemble_function($progs, $func, \%warned) if keys %warned; } @@ -903,142 +981,457 @@ sub defaultglobal($) return { ofs => $ofs, s_name => undef, debugname => "\@$ofs", type => undef }; } -sub parse_progs($) +sub detect_constants($) { - my ($fh) = @_; - - my %p = (); - - print STDERR "Parsing header...\n"; - $p{header} = parse_section $fh, DPROGRAMS_T, 0, undef, 1; - - print STDERR "Parsing strings...\n"; - $p{strings} = get_section $fh, $p{header}{ofs_strings}, $p{header}{numstrings}; - $p{getstring} = sub + my ($progs) = @_; + use constant GLOBALFLAG_R => 1; # read + use constant GLOBALFLAG_W => 2; # written + use constant GLOBALFLAG_S => 4; # saved + use constant GLOBALFLAG_I => 8; # initialized + use constant GLOBALFLAG_N => 16; # named + use constant GLOBALFLAG_Q => 32; # unique to function + use constant GLOBALFLAG_U => 64; # unused + use constant GLOBALFLAG_P => 128; # possibly parameter passing + use constant GLOBALFLAG_D => 256; # has a def + my @globalflags = (GLOBALFLAG_Q | GLOBALFLAG_U) x (@{$progs->{globals}} + 2); + + for(@{$progs->{functions}}) { - my ($startpos) = @_; - my $endpos = index $p{strings}, "\0", $startpos; - return substr $p{strings}, $startpos, $endpos - $startpos; - }; - - print STDERR "Parsing statements...\n"; - $p{statements} = [parse_section $fh, DSTATEMENT_T, $p{header}{ofs_statements}, undef, $p{header}{numstatements}]; + for(keys %{$_->{globals_used}}) + { + if($globalflags[$_] & GLOBALFLAG_U) + { + $globalflags[$_] &= ~GLOBALFLAG_U; + } + elsif($globalflags[$_] & GLOBALFLAG_Q) + { + $globalflags[$_] &= ~GLOBALFLAG_Q; + } + } + $globalflags[$_] |= GLOBALFLAG_R + for keys %{$_->{globals_read}}; + $globalflags[$_] |= GLOBALFLAG_W + for keys %{$_->{globals_written}}; + next + if $_->{first_statement} < 0; + for my $ip($_->{first_statement} .. (@{$progs->{statements}}-1)) + { + my $s = $progs->{statements}[$ip]; + if($s->{op} eq 'STORE_V') + { + $globalflags[$s->{a}] |= GLOBALFLAG_P + if $s->{b} >= $_->{parm_start} and $s->{b} < $_->{parm_start} + $_->{locals}; + $globalflags[$s->{a}+1] |= GLOBALFLAG_P + if $s->{b}+1 >= $_->{parm_start} and $s->{b}+1 < $_->{parm_start} + $_->{locals}; + $globalflags[$s->{a}+2] |= GLOBALFLAG_P + if $s->{b}+2 >= $_->{parm_start} and $s->{b}+2 < $_->{parm_start} + $_->{locals}; + } + elsif($s->{op} =~ /^STORE_/) + { + $globalflags[$s->{a}] |= GLOBALFLAG_P + if $s->{b} >= $_->{parm_start} and $s->{b} < $_->{parm_start} + $_->{locals}; + } + else + { + last; + } + } + } - print STDERR "Parsing globaldefs...\n"; - $p{globaldefs} = [parse_section $fh, DDEF_T, $p{header}{ofs_globaldefs}, undef, $p{header}{numglobaldefs}]; + # parameter passing globals are only ever used in STORE_ instructions + for my $s(@{$progs->{statements}}) + { + next + if $s->{op} =~ /^STORE_/; - print STDERR "Parsing fielddefs...\n"; - $p{fielddefs} = [parse_section $fh, DDEF_T, $p{header}{ofs_fielddefs}, undef, $p{header}{numfielddefs}]; + my $c = checkop $s->{op}; - print STDERR "Parsing globals...\n"; - $p{globals} = [parse_section $fh, DGLOBAL_T, $p{header}{ofs_globals}, undef, $p{header}{numglobals}]; + for(qw(a b c)) + { + my $type = $c->{$_}; + next + unless defined $type; - print STDERR "Parsing functions...\n"; - $p{functions} = [parse_section $fh, DFUNCTION_T, $p{header}{ofs_functions}, undef, $p{header}{numfunctions}]; + my $ofs = $s->{$_}; + if($type eq 'inglobal' || $type eq 'inglobalfunc' || $type eq 'outglobal') + { + $globalflags[$ofs] &= ~GLOBALFLAG_P; + } + if($type eq 'inglobalvec' || $type eq 'outglobalvec') + { + $globalflags[$ofs] &= ~GLOBALFLAG_P; + $globalflags[$ofs+1] &= ~GLOBALFLAG_P; + $globalflags[$ofs+2] &= ~GLOBALFLAG_P; + } + } + } - print STDERR "Detecting temps...\n"; my %offsets_saved = (); - for(@{$p{globaldefs}}) + for(@{$progs->{globaldefs}}) { my $type = $_->{type}; - my $name = $p{getstring}->($_->{s_name}); - next - unless $type->{save} or $name ne ""; - for my $i(0..(typesize($_->{type}{type})-1)) + my $name = $progs->{getstring}->($_->{s_name}); + $name = '' + if $name eq 'IMMEDIATE'; # for fteqcc I had: or $name =~ /^\./; + $_->{debugname} = $name + if $name ne ''; + $globalflags[$_->{ofs}] |= GLOBALFLAG_D; + if($type->{save}) + { + $globalflags[$_->{ofs}] |= GLOBALFLAG_S; + } + if(defined $_->{debugname}) + { + $globalflags[$_->{ofs}] |= GLOBALFLAG_N; + } + } + # fix up vectors + my @extradefs = (); + for(@{$progs->{globaldefs}}) + { + my $type = $_->{type}; + for my $i(1..(typesize($type->{type})-1)) { - ++$offsets_saved{$_->{ofs}+$i}; + # add missing def + if(!($globalflags[$_->{ofs}+$i] & GLOBALFLAG_D)) + { + print "Missing globaldef for a component@{[defined $_->{debugname} ? ' of ' . $_->{debugname} : '']} at $_->{ofs}+$i\n"; + push @extradefs, { + type => { + saved => 0, + type => 'float' + }, + ofs => $_->{ofs} + $i, + debugname => defined $_->{debugname} ? $_->{debugname} . "[$i]" : undef + }; + } + # "saved" and "named" states hit adjacent globals too + $globalflags[$_->{ofs}+$i] |= $globalflags[$_->{ofs}] & (GLOBALFLAG_S | GLOBALFLAG_N | GLOBALFLAG_D); } } + push @{$progs->{globaldefs}}, @extradefs; + my %offsets_initialized = (); - for(0..(@{$p{globals}}-1)) + for(0..(@{$progs->{globals}}-1)) { - if($p{globals}[$_]{v}{int}) + if($progs->{globals}[$_]{v}{int}) { - ++$offsets_initialized{$_}; + $globalflags[$_] |= GLOBALFLAG_I; } } + + my @globaltypes = (undef) x @{$progs->{globals}}; + my %istemp = (); - my %isconst = (); - for(0..(@{$p{globals}}-1)) + for(0..(@{$progs->{globals}}-1)) { next if $_ < @{(DEFAULTGLOBALS)}; - ++$isconst{$_} - if !$offsets_saved{$_} and $offsets_initialized{$_}; - ++$istemp{$_} - if !$offsets_saved{$_} and !$offsets_initialized{$_}; + if(($globalflags[$_] & (GLOBALFLAG_R | GLOBALFLAG_W)) == 0) + { + $globaltypes[$_] = "unused"; + } + elsif(($globalflags[$_] & (GLOBALFLAG_R | GLOBALFLAG_W)) == GLOBALFLAG_R) + { + # so it is ro + if(($globalflags[$_] & GLOBALFLAG_N) == GLOBALFLAG_N) + { + $globaltypes[$_] = "read_only"; + } + elsif(($globalflags[$_] & GLOBALFLAG_S) == 0) + { + $globaltypes[$_] = "const"; + } + else + { + $globaltypes[$_] = "read_only"; + } + } + elsif(($globalflags[$_] & (GLOBALFLAG_R | GLOBALFLAG_W)) == GLOBALFLAG_W) + { + $globaltypes[$_] = "write_only"; + } + else + { + # now we know it is rw + if(($globalflags[$_] & GLOBALFLAG_N) == GLOBALFLAG_N) + { + $globaltypes[$_] = "global"; + } + elsif(($globalflags[$_] & (GLOBALFLAG_S | GLOBALFLAG_I)) == 0) + { + if($globalflags[$_] & GLOBALFLAG_P) + { + $globaltypes[$_] = "OFS_PARM"; + } + elsif($globalflags[$_] & GLOBALFLAG_Q) + { + $globaltypes[$_] = "uniquetemp"; + $istemp{$_} = 0; + } + else + { + $globaltypes[$_] = "temp"; + $istemp{$_} = 1; + } + } + elsif(($globalflags[$_] & (GLOBALFLAG_S | GLOBALFLAG_I)) == GLOBALFLAG_I) + { + $globaltypes[$_] = "not_saved"; + } + else + { + $globaltypes[$_] = "global"; + } + } } - $p{temps} = \%istemp; - $p{consts} = \%isconst; - - print STDERR "Naming...\n"; + $progs->{temps} = \%istemp; # globaldefs - my @globaldefs = (); - for(@{$p{globaldefs}}) - { - my $s = $p{getstring}->($_->{s_name}); - $_->{debugname} //= "_$s" - if length $s; - } - for(@{$p{globaldefs}}) + my @globaldefs = (undef) x @{$progs->{globals}}; + for(@{$progs->{globaldefs}}) { $globaldefs[$_->{ofs}] //= $_ if defined $_->{debugname}; } - for(@{$p{globaldefs}}) + for(@{$progs->{globaldefs}}) { $globaldefs[$_->{ofs}] //= $_; } - for(0..(@{$p{globals}}-1)) + for(0..(@{$progs->{globals}}-1)) { $globaldefs[$_] //= { ofs => $_, s_name => undef, - debugname => undef + debugname => undef, + type => undef }; } - my %globaldefs = (); + for(0..(@{(DEFAULTGLOBALS)}-1)) + { + $globaldefs[$_] = { ofs => $_, s_name => undef, debugname => DEFAULTGLOBALS->[$_], type => undef }; + $globaltypes[$_] = 'defglobal'; + } + my %globaldefs_namecount = (); for(@globaldefs) { - if(!defined $_->{debugname}) + $_->{globaltype} = $globaltypes[$_->{ofs}]; + if(defined $_->{debugname}) { - if($istemp{$_->{ofs}}) - { - $_->{debugname} = "temp_$_->{ofs}"; - } - elsif($isconst{$_->{ofs}}) - { - $_->{debugname} = "(" . get_constant(\%p, $p{globals}[$_->{ofs}]{v}) . ")"; - } - else - { - $_->{debugname} = "global_$_->{ofs}"; - } + # already has debugname + } + elsif($_->{globaltype} eq 'const') + { + $_->{debugname} = get_constant($progs, $progs->{globals}[$_->{ofs}]{v}, $_->{type}{type}); + } + else + { + $_->{debugname} = "$_->{globaltype}_$_->{ofs}"; } - ++$globaldefs{$_->{debugname}}; + ++$globaldefs_namecount{$_->{debugname}}; } for(@globaldefs) { next - if $globaldefs{$_->{debugname}} <= 1; - print "Not unique: $_->{debugname} at $_->{ofs}\n"; + if $globaldefs_namecount{$_->{debugname}} <= 1 && !$ENV{FORCE_OFFSETS}; + #print "Not unique: $_->{debugname} at $_->{ofs}\n"; $_->{debugname} .= "\@$_->{ofs}"; } - $p{globaldef_byoffset} = sub + $progs->{globaldef_byoffset} = sub { my ($ofs) = @_; - $ofs &= 0xFFFF; - if($ofs >= 0 && $ofs < @{(DEFAULTGLOBALS)}) - { - return { ofs => $ofs, s_name => undef, debugname => DEFAULTGLOBALS->[$ofs], type => undef }; - } my $def = $globaldefs[$ofs]; return $def; }; +} + +sub parse_progs($$) +{ + my ($fh, $lnofh) = @_; + + my %p = (); + + print STDERR "Parsing header...\n"; + $p{header} = parse_section $fh, DPROGRAMS_T, 0, undef, 1; + + if (defined $lnofh) { + print STDERR "Parsing LNO...\n"; + my $lnoheader = parse_section $lnofh, LNOHEADER_T, 0, undef, 1; + eval { + die "Not a LNOF" + if $lnoheader->{lnotype} != unpack 'V', 'LNOF'; + die "Not version 1" + if $lnoheader->{version} != 1; + die "Not same count of globaldefs" + if $lnoheader->{numglobaldefs} != $p{header}{numglobaldefs}; + die "Not same count of globals" + if $lnoheader->{numglobals} != $p{header}{numglobals}; + die "Not same count of fielddefs" + if $lnoheader->{numfielddefs} != $p{header}{numfielddefs}; + die "Not same count of statements" + if $lnoheader->{numstatements} != $p{header}{numstatements}; + $p{lno} = [parse_section $lnofh, LNO_T, 24, undef, $lnoheader->{numstatements}]; + eval { + $p{lno} = [parse_section $lnofh, LNO_T, 24, undef, $lnoheader->{numstatements} * 2]; + $p{cno} = [splice $p{lno}, $lnoheader->{numstatements}]; + print STDERR "Cool, this LNO even has column number info!\n"; + }; + } or warn "Skipping LNO: $@"; + } + + print STDERR "Parsing strings...\n"; + $p{strings} = get_section $fh, $p{header}{ofs_strings}, $p{header}{numstrings}; + $p{getstring} = sub + { + my ($startpos) = @_; + my $endpos = index $p{strings}, "\0", $startpos; + return substr $p{strings}, $startpos, $endpos - $startpos; + }; + + print STDERR "Parsing globals...\n"; + $p{globals} = [parse_section $fh, DGLOBAL_T, $p{header}{ofs_globals}, undef, $p{header}{numglobals}]; + + print STDERR "Parsing globaldefs...\n"; + $p{globaldefs} = [parse_section $fh, DDEF_T, $p{header}{ofs_globaldefs}, undef, $p{header}{numglobaldefs}]; + + print STDERR "Range checking globaldefs...\n"; + for(0 .. (@{$p{globaldefs}}-1)) + { + my $g = $p{globaldefs}[$_]; + die "Out of range name in globaldef $_" + if $g->{s_name} < 0 || $g->{s_name} >= length $p{strings}; + my $name = $p{getstring}->($g->{s_name}); + die "Out of range ofs $g->{ofs} in globaldef $_ (name: \"$name\")" + if $g->{ofs} >= $p{globals}; + } - # functions - my %functions = (); + print STDERR "Parsing fielddefs...\n"; + $p{fielddefs} = [parse_section $fh, DDEF_T, $p{header}{ofs_fielddefs}, undef, $p{header}{numfielddefs}]; + + print STDERR "Range checking fielddefs...\n"; + for(0 .. (@{$p{fielddefs}}-1)) + { + my $g = $p{fielddefs}[$_]; + die "Out of range name in fielddef $_" + if $g->{s_name} < 0 || $g->{s_name} >= length $p{strings}; + my $name = $p{getstring}->($g->{s_name}); + die "Out of range ofs $g->{ofs} in fielddef $_ (name: \"$name\")" + if $g->{ofs} >= $p{header}{entityfields}; + push @{$p{entityfieldnames}[$g->{ofs}]}, $name; + } + + print STDERR "Parsing statements...\n"; + $p{statements} = [parse_section $fh, DSTATEMENT_T, $p{header}{ofs_statements}, undef, $p{header}{numstatements}]; + + print STDERR "Parsing functions...\n"; + $p{functions} = [parse_section $fh, DFUNCTION_T, $p{header}{ofs_functions}, undef, $p{header}{numfunctions}]; + + print STDERR "Range checking functions...\n"; + for(0 .. (@{$p{functions}} - 1)) + { + my $f = $p{functions}[$_]; + die "Out of range name in function $_" + if $f->{s_name} < 0 || $f->{s_name} >= length $p{strings}; + my $name = $p{getstring}->($f->{s_name}); + die "Out of range file in function $_" + if $f->{s_file} < 0 || $f->{s_file} >= length $p{strings}; + my $file = $p{getstring}->($f->{s_file}); + die "Out of range first_statement in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + if $f->{first_statement} >= @{$p{statements}}; + if($f->{first_statement} >= 0) + { + die "Out of range parm_start in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + if $f->{parm_start} < 0 || $f->{parm_start} >= @{$p{globals}}; + die "Out of range locals in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + if $f->{locals} < 0 || $f->{parm_start} + $f->{locals} > @{$p{globals}}; + die "Out of range numparms $f->{numparms} in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + if $f->{numparms} < 0 || $f->{numparms} > 8; + my $totalparms = 0; + for(0..($f->{numparms}-1)) + { + die "Out of range parm_size[$_] in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + unless { 0 => 1, 1 => 1, 3 => 1 }->{$f->{parm_size}[$_]}; + $totalparms += $f->{parm_size}[$_]; + } + die "Out of range parms in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + if $f->{parm_start} + $totalparms > @{$p{globals}}; + die "More parms than locals in function $_ (name: \"$name\", file: \"$file\", first statement: $f->{first_statement})" + if $totalparms > $f->{locals}; + } + } + + print STDERR "Range checking statements...\n"; + for my $ip(0 .. (@{$p{statements}}-1)) + { + my $s = $p{statements}[$ip]; + my $c = checkop $s->{op}; + + for(qw(a b c)) + { + my $type = $c->{$_}; + next + unless defined $type; + + if($type eq 'inglobal' || $type eq 'inglobalfunc') + { + $s->{$_} &= 0xFFFF; + die "Out of range global offset in statement $ip - cannot continue" + if $s->{$_} >= @{$p{globals}}; + } + elsif($type eq 'inglobalvec') + { + $s->{$_} &= 0xFFFF; + if($c->{isreturn}) + { + die "Out of range global offset in statement $ip - cannot continue" + if $s->{$_} >= @{$p{globals}}; + print "Potentially out of range global offset in statement $ip - may crash engines" + if $s->{$_} >= @{$p{globals}}-2; + } + else + { + die "Out of range global offset in statement $ip - cannot continue" + if $s->{$_} >= @{$p{globals}}-2; + } + } + elsif($type eq 'outglobal') + { + $s->{$_} &= 0xFFFF; + die "Out of range global offset in statement $ip - cannot continue" + if $s->{$_} >= @{$p{globals}}; + } + elsif($type eq 'outglobalvec') + { + $s->{$_} &= 0xFFFF; + die "Out of range global offset in statement $ip - cannot continue" + if $s->{$_} >= @{$p{globals}}-2; + } + elsif($type eq 'ipoffset') + { + die "Out of range GOTO/IF/IFNOT in statement $ip - cannot continue" + if $ip + $s->{$_} < 0 || $ip + $s->{$_} >= @{$p{statements}}; + } + } + } + + print STDERR "Looking for error(), setmodel(), setsize()...\n"; + $p{builtins} = { error => {}, setmodel => {}, setsize => {} }; + for(@{$p{globaldefs}}) + { + my $name = $p{getstring}($_->{s_name}); + next + if not exists $p{builtins}{$name}; + my $v = $p{globals}[$_->{ofs}]{v}{int}; + next + if $v <= 0 || $v >= @{$p{functions}}; + my $first = $p{functions}[$v]{first_statement}; + next + if $first >= 0; + print STDERR "Detected $name() at offset $_->{ofs} (builtin #@{[-$first]})\n"; + $p{builtins}{$name}{$_->{ofs}} = 1; + } + + print STDERR "Scanning functions...\n"; for(@{$p{functions}}) { my $file = $p{getstring}->($_->{s_file}); @@ -1046,22 +1439,118 @@ sub parse_progs($) $name = "$file:$name" if length $file; $_->{debugname} = $name; - $functions{$_->{first_statement}} = $_; + + next + if $_->{first_statement} < 0; + + my %statements = (); + my %come_from = (); + my %go_to = (); + my %globals_read = (); + my %globals_written = (); + my %globals_used = (); + + if($_->{first_statement} >= 0) + { + run_nfa \%p, $_->{first_statement}, "", id, nfa_default_state_checker, + sub + { + my ($ip, $state, $s, $c) = @_; + ++$statements{$ip}; + + if(my $j = $c->{isjump}) + { + my $t = $ip + $s->{$j}; + $come_from{$t}{$ip} = $c->{isconditional}; + $go_to{$ip}{$t} = $c->{isconditional}; + } + + for my $o(qw(a b c)) + { + my $type = $c->{$o} + or next; + my $ofs = $s->{$o}; + + my $read = sub + { + my ($ofs) = @_; + $globals_read{$ofs}{$ip}{$o} = 1; + $globals_used{$ofs} = 1; + }; + my $write = sub + { + my ($ofs) = @_; + $globals_written{$ofs}{$ip}{$o} = 1; + $globals_used{$ofs} = 1; + }; + + if($type eq 'inglobal' || $type eq 'inglobalfunc') + { + $read->($ofs); + } + elsif($type eq 'inglobalvec') + { + $read->($ofs); + $read->($ofs+1); + $read->($ofs+2); + } + elsif($type eq 'outglobal') + { + $write->($ofs); + } + elsif($type eq 'outglobalvec') + { + $write->($ofs); + $write->($ofs+1); + $write->($ofs+2); + } + } + + return 0; + }; + } + + $_->{statements} = \%statements; + $_->{come_from} = \%come_from; + $_->{go_to} = \%go_to; + $_->{globals_read} = \%globals_read; + $_->{globals_written} = \%globals_written; + $_->{globals_used} = \%globals_used; + + # using this info, we could now identify basic blocks } - $p{function_byoffset} = sub + + print STDERR "Detecting constants and temps, and naming...\n"; + detect_constants \%p; + + if($ENV{DUMP}) { - my ($ofs) = @_; - return $functions{$ofs}; - }; + use Data::Dumper; + print Dumper \%p; + return; + } # what do we want to do? my $checkfunc = \&find_uninitialized_locals; - #my $checkfunc = \&disassemble_function; + if($ENV{DISASSEMBLE}) + { + $checkfunc = \&disassemble_function; + } for(sort { $a->{debugname} cmp $b->{debugname} } @{$p{functions}}) { $checkfunc->(\%p, $_); } } -open my $fh, '<', $ARGV[0]; -parse_progs $fh; +for my $progs (@ARGV) { + my $lno = "$progs.lno"; + $lno =~ s/\.dat\.lno$/.lno/; + + open my $fh, '<', $progs + or die "$progs: $!"; + + open my $lnofh, '<', $lno + or warn "$lno: $!"; + + parse_progs $fh, $lnofh; +}