X-Git-Url: https://git.xonotic.org/?a=blobdiff_plain;f=misc%2Ftools%2Fprogs-analyzer.pl;h=ff6edec11065dbf16ef81700a8343f56b28e8d72;hb=16e64e8fe3ee1595fcc85da57a939a3123f6c487;hp=91eb67f13f7bb4d2b414fe1d9ba51504cc7e644e;hpb=49682cfb866bb80df5fba491c2888b1dd1da58de;p=xonotic%2Fxonotic.git diff --git a/misc/tools/progs-analyzer.pl b/misc/tools/progs-analyzer.pl index 91eb67f1..ff6edec1 100644 --- a/misc/tools/progs-analyzer.pl +++ b/misc/tools/progs-analyzer.pl @@ -1,5 +1,6 @@ use strict; use warnings; +use Digest::SHA; sub id() { @@ -117,7 +118,7 @@ sub checkop($) } if($op =~ /^DONE$|^RETURN$/) { - return { a => 'inglobal', isreturn => 1 }; + return { a => 'inglobalvec', isreturn => 1 }; } return { a => 'inglobal', b => 'inglobal', c => 'outglobal' }; } @@ -126,7 +127,7 @@ use constant TYPES => { int => ['V', 4, signed 32], ushort => ['v', 2, id], short => ['v', 2, signed 16], - opcode => ['v', 2, sub { OPCODE_E->[$_[0]] or die "Invalid opcode: $_[0]"; }], + opcode => ['v', 2, sub { OPCODE_E->[$_[0]] or do { warn "Invalid opcode: $_[0]"; "INVALID#$_[0]"; }; }], float => ['f', 4, id], uchar8 => ['a8', 8, sub { [unpack 'C8', $_[0]] }], global => ['i', 4, sub { { int => $_[0], float => unpack "f", pack "L", $_[0] }; }], @@ -226,10 +227,19 @@ sub parse_section($$$$$) return $out[0]; } +sub nfa_default_state_checker() +{ + my %seen; + return sub + { + my ($ip, $state) = @_; + return $seen{"$ip $state"}++; + }; +} + sub run_nfa($$$$$$) { - my ($progs, $ip, $state, $copy_handler, $state_hasher, $instruction_handler) = @_; - my %seen = (); + my ($progs, $ip, $state, $copy_handler, $state_checker, $instruction_handler) = @_; my $statements = $progs->{statements}; @@ -239,17 +249,17 @@ sub run_nfa($$$$$$) no warnings 'recursion'; my ($ip, $state) = @_; + my $ret = 0; for(;;) { - my $statestr = $state_hasher->($state); - return - if $seen{"$ip:$statestr"}++; + return $ret + if $state_checker->($ip, $state); my $s = $statements->[$ip]; my $c = checkop $s->{op}; - if($instruction_handler->($ip, $state, $s, $c)) + if(($ret = $instruction_handler->($ip, $state, $s, $c))) { # abort execution last; @@ -259,12 +269,31 @@ sub run_nfa($$$$$$) { last; } + elsif($c->{iscall}) + { + my $func = $s->{a}; + my $funcid = $progs->{globals}[$func]{v}{int}; + last + if $progs->{error_func}{$funcid}; + $ip += 1; + } elsif($c->{isjump}) { if($c->{isconditional}) { - $nfa->($ip+1, $copy_handler->($state)); - $ip += $s->{$c->{isjump}}; + if(rand 2) + { + if(($ret = $nfa->($ip+$s->{$c->{isjump}}, $copy_handler->($state))) < 0) + { + last; + } + $ip += 1; + } + else + { + $nfa->($ip+1, $copy_handler->($state)); + $ip += $s->{$c->{isjump}}; + } } else { @@ -276,6 +305,8 @@ sub run_nfa($$$$$$) $ip += 1; } } + + return $ret; }; $nfa->($ip, $copy_handler->($state)); @@ -286,9 +317,9 @@ sub get_constant($$) my ($progs, $g) = @_; if($g->{int} == 0) { - return undef; + return 0; } - elsif($g->{int} > 0 && $g->{int} < 16777216) + elsif($g->{int} > 0 && $g->{int} < 8388608) { if($g->{int} < length $progs->{strings} && $g->{int} > 0) { @@ -305,10 +336,10 @@ sub get_constant($$) } } -use constant PRE_MARK_STATEMENT => "\e[1m"; -use constant POST_MARK_STATEMENT => "\e[m"; -use constant PRE_MARK_OPERAND => "\e[41m"; -use constant POST_MARK_OPERAND => "\e[49m"; +use constant PRE_MARK_STATEMENT => ""; +use constant POST_MARK_STATEMENT => ""; +use constant PRE_MARK_OPERAND => "*** "; +use constant POST_MARK_OPERAND => " ***"; use constant INSTRUCTION_FORMAT => "%8s %3s | %-12s "; use constant OPERAND_FORMAT => "%s"; @@ -348,14 +379,14 @@ sub disassemble_function($$;$) my $p = $func->{parm_start}; for(0..($func->{numparms}-1)) { - $override_locals{$p} //= "argv[$_]"; + $override_locals{$p} //= "argv_$_"; for my $comp(0..($func->{parm_size}[$_]-1)) { - $override_locals{$p} //= "argv[$_][$comp]"; + $override_locals{$p} //= "argv_$_\[$comp]"; ++$p; } printf INSTRUCTION_FORMAT, '', '', '.ARG'; - printf OPERAND_FORMAT, "argv[$_]"; + printf OPERAND_FORMAT, "argv_$_"; print OPERAND_SEPARATOR; printf OPERAND_FORMAT, $func->{parm_size}[$_]; print INSTRUCTION_SEPARATOR; @@ -364,10 +395,10 @@ sub disassemble_function($$;$) { next if exists $override_locals{$_}; - $override_locals{$_} = "\@$_"; + $override_locals{$_} = "local_$_"; printf INSTRUCTION_FORMAT, '', '', '.LOCAL'; - printf OPERAND_FORMAT, "\@$_"; + printf OPERAND_FORMAT, "local_$_"; $initializer->($_); print INSTRUCTION_SEPARATOR; } @@ -375,7 +406,6 @@ sub disassemble_function($$;$) my $getname = sub { my ($ofs) = @_; - $ofs &= 0xFFFF; return $override_locals{$ofs} if exists $override_locals{$ofs}; my $def = $progs->{globaldef_byoffset}->($ofs); @@ -420,25 +450,11 @@ sub disassemble_function($$;$) } }; - my %statements = (); - my %come_from = (); - run_nfa $progs, $func->{first_statement}, "", id, id, - sub - { - my ($ip, $state, $s, $c) = @_; - ++$statements{$ip}; - - if(my $j = $c->{isjump}) - { - my $t = $ip + $s->{$j}; - $come_from{$t}{$ip} = $c->{isconditional}; - } - - return 0; - }; + my $statements = $func->{statements}; + my $come_from = $func->{come_from}; my $ipprev = undef; - for my $ip(sort { $a <=> $b } keys %statements) + for my $ip(sort { $a <=> $b } keys %$statements) { if($ip == $func->{first_statement}) { @@ -451,7 +467,7 @@ sub disassemble_function($$;$) 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; @@ -468,11 +484,25 @@ sub disassemble_function($$;$) my $ipt = $progs->{statements}[$ip]; my $opprop = checkop $op; + if($highlight and $highlight->{$ip}) + { + for(values %{$highlight->{$ip}}) + { + for(@$_) + { + print PRE_MARK_STATEMENT; + printf INSTRUCTION_FORMAT, '', '', '.WARN'; + printf OPERAND_FORMAT, "$_ (in $func->{debugname})"; + print INSTRUCTION_SEPARATOR; + } + } + } + print PRE_MARK_STATEMENT if $highlight and $highlight->{$ip}; my $showip = $opprop->{isjump}; - printf INSTRUCTION_FORMAT, $showip ? $ip : '', $highlight->{$ip} ? "" : "", $op; + printf INSTRUCTION_FORMAT, $showip ? $ip : '', $highlight->{$ip} ? '' : '', $op; my $cnt = 0; for my $o(qw(a b c)) @@ -499,6 +529,7 @@ sub find_uninitialized_locals($$) { my ($progs, $func) = @_; + return if $func->{first_statement} < 0; # builtin @@ -518,45 +549,38 @@ sub find_uninitialized_locals($$) for(keys %{$progs->{temps}}) { - $watchme{$_} = WATCHME_T | WATCHME_X - if not exists $watchme{$_}; + next + if exists $watchme{$_}; + if($progs->{temps}{$_}) + { + # shared temp + $watchme{$_} = WATCHME_T | WATCHME_X + } + else + { + # unique temp + $watchme{$_} = WATCHME_X + } } - run_nfa $progs, $func->{first_statement}, "", id, id, - sub + $watchme{$_} |= WATCHME_R + for keys %{$func->{globals_read}}; + $watchme{$_} |= WATCHME_W + for keys %{$func->{globals_written}}; + + my %write_places = (); + for my $ofs(keys %{$func->{globals_written}}) + { + next + unless exists $watchme{$ofs} and $watchme{$ofs} & WATCHME_X; + for my $ip(keys %{$func->{globals_written}{$ofs}}) { - my ($ip, $state, $s, $c) = @_; - for(qw(a b c)) + for my $op(keys %{$func->{globals_written}{$ofs}{$ip}}) { - my $type = $c->{$_}; - next - unless defined $type; - - my $ofs = $s->{$_}; - if($type eq 'inglobal' || $type eq 'inglobalfunc') - { - $watchme{$ofs} |= WATCHME_R; - } - elsif($type eq 'inglobalvec') - { - $watchme{$ofs} |= WATCHME_R; - $watchme{$ofs+1} |= WATCHME_R; - $watchme{$ofs+2} |= WATCHME_R; - } - elsif($type eq 'outglobal') - { - $watchme{$ofs} |= WATCHME_W; - } - elsif($type eq 'outglobalvec') - { - $watchme{$ofs} |= WATCHME_W; - $watchme{$ofs+1} |= WATCHME_W; - $watchme{$ofs+2} |= WATCHME_W; - } + push @{$write_places{$ip}{$op}}, $ofs; } - - return 0; - }; + } + } for(keys %watchme) { @@ -571,56 +595,72 @@ sub find_uninitialized_locals($$) { $watchme{$_} = { flags => $watchme{$_}, - valid => 0 + valid => [0, undef, undef] }; } # mark parameters as initialized for($func->{parm_start} .. ($p-1)) { - $watchme{$_}{valid} = 1 + $watchme{$_}{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 - if defined $watchme{$s->{a}}; - $watchme{$s->{a}+1}{valid} = 1 - if defined $watchme{$s->{a}+1}; - $watchme{$s->{a}+2}{valid} = 1 - if defined $watchme{$s->{a}+2}; - } - elsif($s->{op} =~ /^STORE_/) - { - $watchme{$s->{a}}{valid} = 1 - if defined $watchme{$s->{a}}; - } - else - { - last; - } - } my %warned = (); + my %ip_seen = (); run_nfa $progs, $func->{first_statement}, \%watchme, sub { my ($h) = @_; return { map { $_ => { %{$h->{$_}} } } keys %$h }; }, sub { - my ($h) = @_; - return join ' ', map { $h->{$_}->{valid}; } sort keys %$h; + my ($ip, $state) = @_; + + my $s = $ip_seen{$ip}; + if($s) + { + # if $state is stronger or equal to $s, return 1 + + for(keys %$state) + { + if($state->{$_}{valid}[0] < $s->{$_}) + { + # The current state is LESS valid than the previously run one. We NEED to run this. + # The saved state can safely become the intersection [citation needed]. + for(keys %$state) + { + $s->{$_} = $state->{$_}{valid}[0] + if $state->{$_}{valid}[0] < $s->{$_}; + } + return 0; + } + } + # if we get here, $state is stronger or equal. No need to try it. + return 1; + } + else + { + # Never seen this IP yet. + $ip_seen{$ip} = { map { ($_ => $state->{$_}{valid}[0]); } keys %$state }; + return 0; + } }, sub { my ($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; + } + } + for(qw(a b c)) { my $type = $c->{$_}; @@ -632,42 +672,52 @@ sub find_uninitialized_locals($$) my $read = sub { my ($ofs) = @_; + ++$return_hack + if $return_hack; return if not exists $state->{$ofs}; my $valid = $state->{$ofs}{valid}; - if($valid == 0) + if($valid->[0] == 0) { - print "; Use of uninitialized value $ofs in $func->{debugname} at $ip.$_\n"; - ++$warned{$ip}{$_}; + # COMPILER BUG of FTEQCC: AND and OR may take uninitialized as second argument (logicops) + if($return_hack <= 2 and ($op ne 'OR' && $op ne 'AND' || $_ ne 'b')) + { + push @{$warned{$ip}{$_}}, "Use of uninitialized value"; + } } - elsif($valid < 0) + elsif($valid->[0] < 0) { - print "; Use of temporary $ofs across CALL in $func->{debugname} at $ip.$_\n"; - ++$warned{$ip}{$_}; + # COMPILER BUG of FTEQCC: AND and OR may take uninitialized as second argument (logicops) + if($return_hack <= 2 and ($op ne 'OR' && $op ne 'AND' || $_ ne 'b')) + { + push @{$warned{$ip}{$_}}, "Use of temporary across CALL"; + } + } + else + { + # it's VALID + if(defined $valid->[1]) + { + delete $write_places{$valid->[1]}{$valid->[2]}; + } } }; my $write = sub { my ($ofs) = @_; - $state->{$ofs}{valid} = 1 + $state->{$ofs}{valid} = [1, $ip, $_] if exists $state->{$ofs}; }; if($type eq 'inglobal' || $type eq 'inglobalfunc') { - if($op ne 'OR' && $op ne 'AND') # fteqcc logicops cause this - { - $read->($ofs); - } + $read->($ofs); } elsif($type eq 'inglobalvec') { - if($op ne 'OR' && $op ne 'AND') # fteqcc logicops cause this - { - $read->($ofs); - $read->($ofs+1); - $read->($ofs+2); - } + $read->($ofs); + $read->($ofs+1); + $read->($ofs+2); } elsif($type eq 'outglobal') { @@ -686,59 +736,125 @@ sub find_uninitialized_locals($$) my $func = $s->{a}; my $funcid = $progs->{globals}[$func]{v}{int}; my $funcobj = $progs->{functions}[$funcid]; - if($funcobj->{first_statement} >= 0) + if(!$funcobj || $funcobj->{first_statement} >= 0) { # invalidate temps for(values %$state) { if($_->{flags} & WATCHME_T) { - $_->{valid} = -1; + $_->{valid} = [-1, undef, undef]; } } } - elsif($funcobj->{debugname} =~ /(^|:)error$/) - { - return 1; - } } return 0; }; + + for my $ip(keys %write_places) + { + for my $operand(keys %{$write_places{$ip}}) + { + # TODO verify it + my %left = map { $_ => 1 } @{$write_places{$ip}{$operand}}; + my $isread = 0; + + my %writeplace_seen = (); + run_nfa $progs, $ip+1, \%left, + sub + { + return { %{$_[0]} }; + }, + sub + { + my ($ip, $state) = @_; + return $writeplace_seen{"$ip " . join " ", sort keys %$state}++; + }, + sub + { + my ($ip, $state, $s, $c) = @_; + for(qw(a b c)) + { + my $type = $c->{$_}; + next + unless defined $type; + + my $ofs = $s->{$_}; + if($type eq 'inglobal' || $type eq 'inglobalfunc') + { + if($state->{$ofs}) + { + $isread = 1; + return -1; # exit TOTALLY + } + } + elsif($type eq 'inglobalvec') + { + if($state->{$ofs} || $state->{$ofs+1} || $state->{$ofs+2}) + { + $isread = 1; + return -1; # exit TOTALLY + } + } + elsif($type eq 'outglobal') + { + delete $state->{$ofs}; + return 1 + if !%$state; + } + elsif($type eq 'outglobalvec') + { + delete $state->{$ofs}; + delete $state->{$ofs+1}; + delete $state->{$ofs+2}; + return 1 + if !%$state; + } + } + return 0; + }; + + if(!$isread) + { + push @{$warned{$ip}{$operand}}, "Value is never used"; + } + } + } disassemble_function($progs, $func, \%warned) if keys %warned; } use constant DEFAULTGLOBALS => [ - "", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]", - "", - "[1]", - "[2]" + "OFS_NULL", + "OFS_RETURN", + "OFS_RETURN[1]", + "OFS_RETURN[2]", + "OFS_PARM0", + "OFS_PARM0[1]", + "OFS_PARM0[2]", + "OFS_PARM1", + "OFS_PARM1[1]", + "OFS_PARM1[2]", + "OFS_PARM2", + "OFS_PARM2[1]", + "OFS_PARM2[2]", + "OFS_PARM3", + "OFS_PARM3[1]", + "OFS_PARM3[2]", + "OFS_PARM4", + "OFS_PARM4[1]", + "OFS_PARM4[2]", + "OFS_PARM5", + "OFS_PARM5[1]", + "OFS_PARM5[2]", + "OFS_PARM6", + "OFS_PARM6[1]", + "OFS_PARM6[2]", + "OFS_PARM7", + "OFS_PARM7[1]", + "OFS_PARM7[2]" ]; sub defaultglobal($) @@ -751,139 +867,331 @@ 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 + my @globalflags = (GLOBALFLAG_Q | GLOBALFLAG_U) x @{$progs->{globals}}; + + for(@{$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}}; + 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 $name eq 'IMMEDIATE'; + if($type->{save}) { - ++$offsets_saved{$_->{ofs}+$i}; + for my $i(0..(typesize($_->{type}{type})-1)) + { + $globalflags[$_->{ofs}] |= GLOBALFLAG_S; + } + } + if($name ne "") + { + for my $i(0..(typesize($_->{type}{type})-1)) + { + $globalflags[$_->{ofs}] |= GLOBALFLAG_N; + } } } 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 @globaldefs = (undef) x @{$progs->{globaldefs}}; + for(@{$progs->{globaldefs}}) { - $_->{debugname} = $p{getstring}->($_->{s_name}); + my $s = $progs->{getstring}->($_->{s_name}); + $s = '' + if $s eq 'IMMEDIATE'; + $_->{debugname} //= "\$" . "$s" + if length $s; } - for(@{$p{globaldefs}}) + for(@{$progs->{globaldefs}}) { $globaldefs[$_->{ofs}] //= $_ - if $_->{debugname} ne ""; + 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 => "" + debugname => undef }; } - my %globaldefs = (); + for(0..(@{(DEFAULTGLOBALS)}-1)) + { + $globaldefs[$_] = { ofs => $_, s_name => undef, debugname => DEFAULTGLOBALS->[$_], type => undef }; + $globaltypes[$_] = 'defglobal'; + } + my %globaldefs_namecount = (); for(@globaldefs) { - if($_->{debugname} eq "") + $_->{globaltype} = $globaltypes[$_->{ofs}]; + if(defined $_->{debugname}) { - if($istemp{$_->{ofs}}) - { - $_->{debugname} = "\@$_->{ofs}"; - } - elsif($isconst{$_->{ofs}}) - { - $_->{debugname} = "<" . get_constant(\%p, $p{globals}[$_->{ofs}]{v}) . ">\@$_->{ofs}"; - } - else - { - $_->{debugname} = "\@$_->{ofs}"; - } + # already has debugname + } + elsif($_->{globaltype} eq 'const') + { + $_->{debugname} = get_constant($progs, $progs->{globals}[$_->{ofs}]{v}); } - ++$globaldefs{$_->{debugname}}; + else + { + $_->{debugname} = "$_->{globaltype}_$_->{ofs}"; + } + ++$globaldefs_namecount{$_->{debugname}}; } for(@globaldefs) { next - if $globaldefs{$_->{debugname}} <= 1; + if $globaldefs_namecount{$_->{debugname}} <= 1; + #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) = @_; - # functions - my %functions = (); + 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}; + $p{getstring} = sub + { + my ($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}]; + + print STDERR "Fixing statements...\n"; + for my $s(@{$p{statements}}) + { + 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; + } + elsif($type eq 'inglobalvec') + { + $s->{$_} &= 0xFFFF; + } + elsif($type eq 'outglobal') + { + $s->{$_} &= 0xFFFF; + } + elsif($type eq 'outglobalvec') + { + $s->{$_} &= 0xFFFF; + } + } + } + + print STDERR "Parsing globaldefs...\n"; + $p{globaldefs} = [parse_section $fh, DDEF_T, $p{header}{ofs_globaldefs}, undef, $p{header}{numglobaldefs}]; + + print STDERR "Parsing fielddefs...\n"; + $p{fielddefs} = [parse_section $fh, DDEF_T, $p{header}{ofs_fielddefs}, undef, $p{header}{numfielddefs}]; + + print STDERR "Parsing globals...\n"; + $p{globals} = [parse_section $fh, DGLOBAL_T, $p{header}{ofs_globals}, undef, $p{header}{numglobals}]; + + print STDERR "Parsing functions...\n"; + $p{functions} = [parse_section $fh, DFUNCTION_T, $p{header}{ofs_functions}, undef, $p{header}{numfunctions}]; + + print STDERR "Looking for error()...\n"; + $p{error_func} = {}; + for(@{$p{globaldefs}}) + { + next + if $p{getstring}($_->{s_name}) ne 'error'; + 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 error() at offset $_->{ofs} (builtin #@{[-$first]})\n"; + $p{error_func}{$_->{ofs}} = 1; + } + + print STDERR "Scanning functions...\n"; for(@{$p{functions}}) { my $file = $p{getstring}->($_->{s_file}); @@ -891,13 +1199,86 @@ 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 = (); + + 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 } - $p{function_byoffset} = sub - { - my ($ofs) = @_; - return $functions{$ofs}; - }; + + print STDERR "Detecting constants and temps, and naming...\n"; + detect_constants \%p; # what do we want to do? my $checkfunc = \&find_uninitialized_locals;