; is this token the start of a type? function token_is_type argument token local c c = *1token if c == TOKEN_IDENTIFIER goto token_is_ident_type if c == KEYWORD_UNSIGNED goto return_1 if c == KEYWORD_CHAR goto return_1 if c == KEYWORD_SHORT goto return_1 if c == KEYWORD_INT goto return_1 if c == KEYWORD_LONG goto return_1 if c == KEYWORD_FLOAT goto return_1 if c == KEYWORD_DOUBLE goto return_1 if c == KEYWORD_VOID goto return_1 if c == KEYWORD_STRUCT goto return_1 if c == KEYWORD_UNION goto return_1 if c == KEYWORD_ENUM goto return_1 goto return_0 :token_is_ident_type token += 8 c = *8token local b b = ident_list_lookup(typedefs, c) if b != 0 goto return_1 goto return_0 function parse_tokens argument tokens local token local ident local type local p local b local c local base_type local base_type_end local name local prefix local prefix_end local suffix local suffix_end local is_extern token = tokens :parse_tokens_loop is_extern = 0 if *1token == TOKEN_EOF goto parse_tokens_eof if *1token == KEYWORD_STATIC goto parse_static_toplevel_decl if *1token == KEYWORD_EXTERN goto parse_extern_toplevel_decl if *1token == KEYWORD_TYPEDEF goto parse_typedef b = token_is_type(token) if b != 0 goto parse_toplevel_decl die(.str_bad_statement) :str_bad_statement string Bad statement. byte 0 :parse_static_toplevel_decl token += 16 ; we don't care that this is static goto parse_toplevel_decl :parse_extern_toplevel_decl token += 16 is_extern = 1 goto parse_toplevel_decl :parse_toplevel_decl base_type = token base_type_end = type_get_base_end(token) token = base_type_end :tl_decl_loop prefix = token prefix_end = type_get_prefix_end(prefix) if *1prefix_end != TOKEN_IDENTIFIER goto tl_decl_no_ident name = prefix_end + 8 name = *8name suffix = prefix_end + 16 suffix_end = type_get_suffix_end(prefix) type = types_bytes_used parse_type_declarators(prefix, prefix_end, suffix, suffix_end) parse_base_type(base_type, base_type_end) token = suffix_end if *1token == SYMBOL_LBRACE goto parse_function_definition if is_extern != 0 goto parse_tl_decl_cont ; ignore external variable declarations ; deal with the initializer if there is one if *1token == SYMBOL_SEMICOLON goto parse_tld_no_initializer if *1token == SYMBOL_COMMA goto parse_tld_no_initializer if *1token == SYMBOL_EQ goto parse_tld_initializer token_error(token, .str_unrecognized_stuff_after_declaration) :str_unrecognized_stuff_after_declaration string Declaration should be followed by one of: { , = byte 32 byte 59 ; semicolon byte 0 :parse_tl_decl_cont ; ensure rwdata_end_addr is aligned to 8 bytes rwdata_end_addr += 7 rwdata_end_addr >= 3 rwdata_end_addr <= 3 if *1token == SYMBOL_SEMICOLON goto tl_decl_loop_done if *1token != SYMBOL_COMMA goto tld_bad_stuff_after_decl token += 16 goto tl_decl_loop :tl_decl_loop_done token += 16 ; skip semicolon goto parse_tokens_loop :tl_decl_no_ident token_error(prefix_end, .str_tl_decl_no_ident) :str_tl_decl_no_ident string No identifier in top-level declaration. byte 0 :tld_bad_stuff_after_decl token_error(token, .str_tld_bad_stuff_after_decl) :str_tld_bad_stuff_after_decl string Declarations should be immediately followed by a comma or semicolon. byte 0 :parse_tld_no_initializer p = types + type if *1p == TYPE_FUNCTION goto parse_tl_decl_cont ; ignore function declarations -- we do two passes anyways b = ident_list_lookup(global_variables, name) if b != 0 goto global_redefinition c = type < 32 c |= rwdata_end_addr ident_list_add(global_variables, name, c) ; just skip forward by the size of this variable -- it'll automatically be filled with 0s. rwdata_end_addr += type_sizeof(type) goto parse_tl_decl_cont :parse_tld_initializer if *1p == TYPE_FUNCTION goto function_initializer b = ident_list_lookup(global_variables, name) if b != 0 goto global_redefinition token += 16 ; skip = c = type < 32 c |= rwdata_end_addr ident_list_add(global_variables, name, c) parse_constant_initializer(&token, type) goto parse_tl_decl_cont :global_redefinition token_error(token, .str_global_redefinition) :str_global_redefinition string Redefinition of global variable. byte 0 :function_initializer token_error(token, .str_function_initializer) :str_function_initializer string Functions should not have initializers. byte 0 :parse_function_definition p = types + type if *1p != TYPE_FUNCTION goto lbrace_after_declaration die(.str_fdNI) ; @TODO :str_fdNI string function definitions not implemented. byte 10 byte 0 :lbrace_after_declaration token_error(token, .str_lbrace_after_declaration) :str_lbrace_after_declaration string Opening { after declaration of non-function. byte 0 :parse_typedef base_type = token + 16 base_type_end = type_get_base_end(base_type) token = base_type_end :typedef_loop prefix = token prefix_end = type_get_prefix_end(prefix) if *1prefix_end != TOKEN_IDENTIFIER goto typedef_no_ident ident = prefix_end + 8 ident = *8ident suffix = prefix_end + 16 suffix_end = type_get_suffix_end(prefix) ;putc('B) ;putc(':) ;print_tokens(base_type, base_type_end) ;putc('P) ;putc(':) ;print_tokens(prefix, prefix_end) ;putc('S) ;putc(':) ;print_tokens(suffix, suffix_end) type = types_bytes_used parse_type_declarators(prefix, prefix_end, suffix, suffix_end) parse_base_type(base_type) puts(.str_typedef) putc(32) print_type(type) putc(10) b = ident_list_lookup(typedefs, ident) if b != 0 goto typedef_redefinition ident_list_add(typedefs, ident, type) token = suffix_end if *1token == SYMBOL_SEMICOLON goto typedef_loop_end if *1token != SYMBOL_COMMA goto bad_typedef token += 16 ; skip comma goto typedef_loop :typedef_loop_end token += 16 ; skip semicolon goto parse_tokens_loop :typedef_no_ident token_error(token, .str_typedef_no_ident) :str_typedef_no_ident string No identifier in typedef declaration. byte 0 :bad_typedef token_error(token, .str_bad_typedef) :str_bad_typedef string Bad typedef. byte 0 :typedef_redefinition token_error(token, .str_typedef_redefinition) :str_typedef_redefinition string typedef redefinition. byte 0 :parse_tokens_eof return ; parse a global variable's initializer ; e.g. int x[5] = {1+8, 2, 3, 4, 5}; ; advances *p_token to the token right after the initializer ; if `type` refers to a sizeless array type (e.g. int x[] = {1,2,3};), it will be altered to the correct size ; outputs the initializer data to rwdata_end_addr, and advances it accordingly. ; after calling this, make sure to align rwdata_end_addr properly function parse_constant_initializer argument p_token argument type local token local end local depth local a local b local c local p local expr local value token = *8p_token depth = 0 end = token ; find the end of the initializer, i.e. the next comma or semicolon not inside braces, ; square brackets, or parentheses: :find_init_end_loop c = *1end if c == TOKEN_EOF goto find_init_end_eof if c == SYMBOL_LPAREN goto find_init_end_incdepth if c == SYMBOL_LSQUARE goto find_init_end_incdepth if c == SYMBOL_LBRACE goto find_init_end_incdepth if c == SYMBOL_RPAREN goto find_init_end_decdepth if c == SYMBOL_RSQUARE goto find_init_end_decdepth if c == SYMBOL_RBRACE goto find_init_end_decdepth if depth > 0 goto find_init_cont if depth < 0 goto init_end_bad_brackets if c == SYMBOL_COMMA goto found_init_end if c == SYMBOL_SEMICOLON goto found_init_end :find_init_cont end += 16 goto find_init_end_loop :find_init_end_incdepth depth += 1 goto find_init_cont :find_init_end_decdepth depth -= 1 goto find_init_cont :found_init_end if *1token == SYMBOL_LBRACE goto parse_braced_initializer if *1token == TOKEN_STRING_LITERAL goto parse_string_literal_initializer ; an ordinary expression p = types + type if *1p > TYPE_POINTER goto expression_initializer_for_nonscalar_type global 8000 dat_const_initializer expr = &dat_const_initializer parse_expression(token, end, expr) evaluate_constant_expression(token, expr, &value) if *1p > TYPE_UNSIGNED_LONG goto init_floating_check :init_good c = type_sizeof(type) p = output_file_data + rwdata_end_addr rwdata_end_addr += c if c == 1 goto write_initializer1 if c == 2 goto write_initializer2 if c == 4 goto write_initializer4 if c == 8 goto write_initializer8 die(.init_somethings_very_wrong) :init_somethings_very_wrong string Scalar type with a weird size. This shouldn't happen. byte 0 :write_initializer1 *1p = value goto const_init_ret :write_initializer2 *2p = value goto const_init_ret :write_initializer4 *4p = value goto const_init_ret :write_initializer8 *8p = value goto const_init_ret :init_floating_check ; also checks pointer types (e.g. don't do static void *x = 1;) if value != 0 goto floating_initializer_other_than_0 goto init_good :const_init_ret *8p_token = end return :parse_braced_initializer byte 0xcc ; @TODO :parse_string_literal_initializer p = token + 16 if p != end goto stuff_after_string_literal p = types + type if *1p == TYPE_ARRAY goto string_literal_array_initializer if *1p != TYPE_POINTER goto string_literal_bad_type p += 1 if *1p != TYPE_CHAR goto string_literal_bad_type token += 8 p = output_file_data + rwdata_end_addr *8p = *8token rwdata_end_addr += 8 goto const_init_ret :string_literal_array_initializer p += 1 c = *8p ; array size token += 8 a = output_file_data + rwdata_end_addr ; destination (where to put the string data) b = output_file_data + *8token ; source (where the string data is now) if c == 0 goto string_literal_sizeless_initializer value = strlen(b) if c < value goto string_literal_init_too_long ; e.g. char x[3] = "hello"; strcpy(a, b) rwdata_end_addr += c ; advance by c, which is possibly more than the length of the string--the remaining bytes will be 0s goto const_init_ret :string_literal_sizeless_initializer ; e.g. char x[] = "hello"; c = strlen(b) c += 1 ; null terminator *8p = c ; set array size strcpy(a, b) rwdata_end_addr += c goto const_init_ret :string_literal_init_too_long token_error(token, .str_string_literal_init_too_long) :str_string_literal_init_too_long string String literal is too long to fit in array. byte 0 :stuff_after_string_literal token_error(token, .str_stuff_after_string_literal) :str_stuff_after_string_literal string Stuff after string literal in initializer. byte 0 :string_literal_bad_type token_error(token, .str_string_literal_bad_type) :str_string_literal_bad_type string Bad type for string literal initializer (i.e. not char* or char[]). byte 0 :find_init_end_eof token_error(token, .str_find_init_end_eof) :str_find_init_end_eof string Can't find end of initializer. byte 0 :init_end_bad_brackets token_error(end, .str_init_end_bad_brackets) :str_init_end_bad_brackets string Too many closing brackets. byte 0 :expression_initializer_for_nonscalar_type token_error(token, .str_expression_initializer_for_nonscalar_type) :str_expression_initializer_for_nonscalar_type string Expression initializer for non-scalar type. byte 0 :floating_initializer_other_than_0 token_error(token, .str_floating_initializer_other_than_0) :str_floating_initializer_other_than_0 string Only 0 is supported as a floating-point/pointer initializer. byte 0 ; *p_token should be pointing to a {, this will advance it to point to the matching } function token_skip_to_matching_rbrace argument p_token local token local depth token = *8p_token depth = 0 :skip_rbrace_loop if *1token == SYMBOL_LBRACE goto skip_rbrace_incdepth if *1token == SYMBOL_RBRACE goto skip_rbrace_decdepth if *1token == TOKEN_EOF goto skip_rbrace_eof :skip_rbrace_next token += 16 goto skip_rbrace_loop :skip_rbrace_incdepth depth += 1 goto skip_rbrace_next :skip_rbrace_decdepth depth -= 1 if depth == 0 goto skip_rbrace_ret goto skip_rbrace_next :skip_rbrace_ret *8p_token = token return :skip_rbrace_eof token_error(*8p_token, .str_skip_rbrace_eof) :str_skip_rbrace_eof string Unmatched { byte 0 ; *p_token should be pointing to a [, this will advance it to point to the matching ] function token_skip_to_matching_rsquare argument p_token local token local depth token = *8p_token depth = 0 :skip_square_loop if *1token == SYMBOL_LSQUARE goto skip_square_incdepth if *1token == SYMBOL_RSQUARE goto skip_square_decdepth if *1token == TOKEN_EOF goto skip_square_eof :skip_square_next token += 16 goto skip_square_loop :skip_square_incdepth depth += 1 goto skip_square_next :skip_square_decdepth depth -= 1 if depth == 0 goto skip_square_ret goto skip_square_next :skip_square_ret *8p_token = token return :skip_square_eof token_error(*8p_token, .str_skip_square_eof) :str_skip_square_eof string Unmatched [ byte 0 ; *p_token should be on a ); this goes back to the corresponding ( ; THERE MUST ACTUALLY BE A MATCHING BRACKET, OTHERWISE THIS WILL DO BAD THINGS function token_reverse_to_matching_lparen argument p_token local token local depth token = *8p_token depth = 0 :reverse_paren_loop if *1token == SYMBOL_LPAREN goto reverse_paren_incdepth if *1token == SYMBOL_RPAREN goto reverse_paren_decdepth :reverse_paren_next token -= 16 goto reverse_paren_loop :reverse_paren_incdepth depth += 1 if depth == 0 goto reverse_paren_ret goto reverse_paren_next :reverse_paren_decdepth depth -= 1 goto reverse_paren_next :reverse_paren_ret *8p_token = token return ; parse things like `int x` or `int f(void, int, char *)` ; advances *p_token ; returns type ID, or 0, in which case you should look at parse_type_result function parse_type ; split types into base (B), prefix (P) and suffix (S) ; struct Thing (*things[5])(void), *something_else[3]; ; BBBBBBBBBBBB PP SSSSSSSSSS P SSS ; Here, we call `struct Thing` the "base type". byte 0xcc ; return the end of the base for this type. function type_get_base_end argument token local c c = *1token if c == KEYWORD_STRUCT goto skip_struct_union_enum if c == KEYWORD_UNION goto skip_struct_union_enum if c == KEYWORD_ENUM goto skip_struct_union_enum ; skip the "base type" token += 16 ; importantly, this skips the typedef'd name if there is one (e.g. typedef int Foo; Foo x;) :skip_base_type_loop c = *1token if c == KEYWORD_UNSIGNED goto skip_base_type_loop_cont ;e.g. int unsigned x; if c == KEYWORD_CHAR goto skip_base_type_loop_cont ;e.g. unsigned char x; if c == KEYWORD_SHORT goto skip_base_type_loop_cont ;e.g. unsigned short x; if c == KEYWORD_INT goto skip_base_type_loop_cont ;e.g. unsigned int x; if c == KEYWORD_LONG goto skip_base_type_loop_cont ;e.g. unsigned long x; if c == KEYWORD_DOUBLE goto skip_base_type_loop_cont ;e.g. long double x; goto skip_base_type_loop_end :skip_base_type_loop_cont token += 16 goto skip_base_type_loop :skip_base_type_loop_end return token :skip_struct_union_enum token += 16 if *1token != TOKEN_IDENTIFIER goto skip_sue_no_name token += 16 ; struct *blah* :skip_sue_no_name if *1token != SYMBOL_LBRACE goto skip_base_type_loop_end ; e.g. struct Something x[5]; ; okay we have something like ; struct { ; int x, y; ; } test; token_skip_to_matching_rbrace(&token) token += 16 goto skip_base_type_loop_end ; return the end of this type prefix function type_get_prefix_end argument token local c :find_prefix_end_loop c = *1token if c == TOKEN_IDENTIFIER goto found_prefix_end if c == KEYWORD_UNSIGNED goto prefix_end_cont if c == KEYWORD_CHAR goto prefix_end_cont if c == KEYWORD_SHORT goto prefix_end_cont if c == KEYWORD_INT goto prefix_end_cont if c == KEYWORD_LONG goto prefix_end_cont if c == KEYWORD_FLOAT goto prefix_end_cont if c == KEYWORD_DOUBLE goto prefix_end_cont if c == SYMBOL_LPAREN goto prefix_end_cont if c == SYMBOL_TIMES goto prefix_end_cont if c == SYMBOL_LSQUARE goto found_prefix_end if c == SYMBOL_RPAREN goto found_prefix_end goto found_prefix_end :prefix_end_cont token += 16 goto find_prefix_end_loop :found_prefix_end return token ; return the end of this type suffix ; NOTE: you must pass in the PREFIX. ; (In general, we can't find the end of the suffix without knowing the prefix.) ; int (*x); ; ^ suffix ends here ; (int *) ; ^ suffix ends here function type_get_suffix_end argument prefix local depth local token local c ; find end of suffix token = prefix depth = 0 ; parenthesis/square bracket depth :suffix_end_loop c = *1token if c == TOKEN_IDENTIFIER goto suffix_end_cont if c == SYMBOL_LSQUARE goto suffix_end_incdepth if c == SYMBOL_RSQUARE goto suffix_end_decdepth if c == SYMBOL_LPAREN goto suffix_end_incdepth if c == SYMBOL_RPAREN goto suffix_end_decdepth if c == SYMBOL_TIMES goto suffix_end_cont if depth == 0 goto suffix_end_found if c == TOKEN_EOF goto type_get_suffix_bad_type goto suffix_end_cont :suffix_end_incdepth depth += 1 goto suffix_end_cont :suffix_end_decdepth depth -= 1 if depth < 0 goto suffix_end_found goto suffix_end_cont :suffix_end_cont token += 16 goto suffix_end_loop :suffix_end_found return token :type_get_suffix_bad_type token_error(prefix, .str_bad_type_suffix) :str_bad_type_suffix string Bad type suffix. byte 0 ; writes to *(types + types_bytes_used), and updates types_bytes_used function parse_type_declarators argument prefix argument prefix_end argument suffix argument suffix_end local p local expr local n local c local depth local out ; main loop for parsing types :type_declarators_loop p = prefix_end - 16 if *1suffix == SYMBOL_LSQUARE goto parse_array_type if *1suffix == SYMBOL_LPAREN goto parse_function_type if *1p == SYMBOL_TIMES goto parse_pointer_type if suffix == suffix_end goto type_declarators_loop_end if *1suffix == SYMBOL_RPAREN goto parse_type_remove_parentheses goto parse_typedecls_bad_type :parse_pointer_type out = types + types_bytes_used *1out = TYPE_POINTER types_bytes_used += 1 prefix_end = p goto type_declarators_loop :parse_array_type out = types + types_bytes_used *1out = TYPE_ARRAY types_bytes_used += 1 p = suffix token_skip_to_matching_rsquare(&p) suffix += 16 ; skip [ if *1suffix == SYMBOL_RSQUARE goto array_no_size ; little hack to avoid screwing up types like double[sizeof(int)] ; temporarily pretend we're using a lot more of types local prev_types_bytes_used prev_types_bytes_used = types_bytes_used types_bytes_used += 4000 expr = malloc(4000) parse_expression(suffix, p, expr) ;print_expression(expr) ;putc(10) evaluate_constant_expression(prefix, expr, &n) if n < 0 goto bad_array_size free(expr) types_bytes_used = prev_types_bytes_used out = types + types_bytes_used *8out = n types_bytes_used += 8 suffix = p + 16 goto type_declarators_loop :bad_array_size token_error(suffix, .str_bad_array_size) :str_bad_array_size string Very large or negative array size. byte 0 :array_no_size ; e.g. int x[] = {1,2,3}; out = types + types_bytes_used *8out = 0 types_bytes_used += 8 suffix += 16 goto type_declarators_loop :parse_function_type local param_base_type local param_prefix local param_prefix_end local param_suffix local param_suffix_end p = suffix + 16 out = types + types_bytes_used *1out = TYPE_FUNCTION types_bytes_used += 1 :function_type_loop param_base_type = p param_prefix = type_get_base_end(param_base_type) param_prefix_end = type_get_prefix_end(param_prefix) param_suffix = param_prefix_end if *1param_suffix != TOKEN_IDENTIFIER goto functype_no_ident param_suffix += 16 :functype_no_ident param_suffix_end = type_get_suffix_end(param_prefix) parse_type_declarators(param_prefix, param_prefix_end, param_suffix, param_suffix_end) parse_base_type(param_base_type) p = param_suffix_end if *1p == SYMBOL_RPAREN goto function_type_loop_end if *1p != SYMBOL_COMMA goto parse_typedecls_bad_type p += 16 goto function_type_loop :function_type_loop_end out = types + types_bytes_used *1out = 0 types_bytes_used += 1 suffix = p + 16 goto type_declarators_loop :parse_type_remove_parentheses if *1p != SYMBOL_LPAREN goto parse_typedecls_bad_type prefix_end = p suffix += 16 goto type_declarators_loop :type_declarators_loop_end return 0 :parse_typedecls_bad_type token_error(prefix, .str_bad_type_declarators) :str_bad_type_declarators string Bad type declarators. byte 0 ; writes to *(types + types_bytes_used), and updates types_bytes_used (no return value) function parse_base_type argument base_type local out local flags local p local c local depth local is_struct is_struct = 0 out = types + types_bytes_used c = *1base_type if c == TOKEN_IDENTIFIER goto base_type_typedef if c == KEYWORD_STRUCT goto base_type_struct if c == KEYWORD_UNION goto base_type_union if c == KEYWORD_ENUM goto base_type_enum if c == KEYWORD_FLOAT goto base_type_float if c == KEYWORD_VOID goto base_type_void ; "normal" type like int, unsigned char, etc. ; annoyingly, all of these are equivalent to `unsigned long`: ; unsigned long int ; long unsigned int ; int long unsigned ; etc. ; so we represent these as PARSETYPE_FLAG_UNSIGNED|PARSETYPE_FLAG_LONG|PARSETYPE_FLAG_INT. #define PARSETYPE_FLAG_UNSIGNED 1 #define PARSETYPE_FLAG_CHAR 2 #define PARSETYPE_FLAG_SHORT 4 #define PARSETYPE_FLAG_INT 8 #define PARSETYPE_FLAG_LONG 16 #define PARSETYPE_FLAG_DOUBLE 32 flags = 0 p = base_type :base_type_normal_loop c = *1p p += 16 if c == KEYWORD_CHAR goto base_type_flag_char if c == KEYWORD_SHORT goto base_type_flag_short if c == KEYWORD_INT goto base_type_flag_int if c == KEYWORD_LONG goto base_type_flag_long if c == KEYWORD_UNSIGNED goto base_type_flag_unsigned if c == KEYWORD_DOUBLE goto base_type_flag_double goto base_type_normal_loop_end :base_type_flag_char c = flags & PARSETYPE_FLAG_CHAR if c != 0 goto repeated_base_type flags |= PARSETYPE_FLAG_CHAR goto base_type_normal_loop :base_type_flag_short c = flags & PARSETYPE_FLAG_SHORT if c != 0 goto repeated_base_type flags |= PARSETYPE_FLAG_SHORT goto base_type_normal_loop :base_type_flag_int c = flags & PARSETYPE_FLAG_INT if c != 0 goto repeated_base_type flags |= PARSETYPE_FLAG_INT goto base_type_normal_loop :base_type_flag_long c = flags & PARSETYPE_FLAG_LONG if c != 0 goto repeated_base_type flags |= PARSETYPE_FLAG_LONG goto base_type_normal_loop :base_type_flag_unsigned c = flags & PARSETYPE_FLAG_UNSIGNED if c != 0 goto repeated_base_type flags |= PARSETYPE_FLAG_UNSIGNED goto base_type_normal_loop :base_type_flag_double c = flags & PARSETYPE_FLAG_DOUBLE if c != 0 goto repeated_base_type flags |= PARSETYPE_FLAG_DOUBLE goto base_type_normal_loop :repeated_base_type token_error(p, .str_repeated_base_type) :str_repeated_base_type string Arithmetic type repeated (e.g. unsigned unsigned int). byte 0 :base_type_normal_loop_end if flags == 8 goto base_type_int ; `int` if flags == 1 goto base_type_uint ; `unsigned` if flags == 9 goto base_type_uint ; `unsigned int` etc. if flags == 2 goto base_type_char ; `char` if flags == 3 goto base_type_uchar ; `unsigned char` etc. if flags == 4 goto base_type_short ; `short` if flags == 12 goto base_type_short `short int` etc. if flags == 5 goto base_type_ushort ; `unsigned short` etc. if flags == 13 goto base_type_ushort ; `unsigned short int` etc. if flags == 16 goto base_type_long ; `long` if flags == 24 goto base_type_long ; `long int` etc. if flags == 17 goto base_type_ulong ; `unsigned long` etc. if flags == 25 goto base_type_ulong ; `unsigned long int` etc. if flags == 32 goto base_type_double ; `double` if flags == 48 goto base_type_double ; `long double` (we use the same type for double and long double) goto bad_base_type :base_type_char *1out = TYPE_CHAR out += 1 goto base_type_done :base_type_uchar *1out = TYPE_UNSIGNED_CHAR out += 1 goto base_type_done :base_type_short *1out = TYPE_SHORT out += 1 goto base_type_done :base_type_ushort *1out = TYPE_UNSIGNED_SHORT out += 1 goto base_type_done :base_type_int *1out = TYPE_INT out += 1 goto base_type_done :base_type_uint *1out = TYPE_UNSIGNED_INT out += 1 goto base_type_done :base_type_long *1out = TYPE_LONG out += 1 goto base_type_done :base_type_ulong *1out = TYPE_UNSIGNED_LONG out += 1 goto base_type_done :base_type_double *1out = TYPE_DOUBLE out += 1 goto base_type_done :base_type_done types_bytes_used = out - types return 0 :base_type_struct is_struct = 1 ; fallthrough :base_type_union local struct_name local struct struct_name = .empty_string p = base_type + 16 if *1p != TOKEN_IDENTIFIER goto base_type_have_name p += 8 struct_name = *8p p += 8 :base_type_have_name c = ident_list_lookup(structures, struct_name) if *1p == SYMBOL_LBRACE goto base_type_struct_definition if c == 0 goto base_type_incomplete_struct ; e.g. struct Foo x; where struct Foo has been defined *1out = TYPE_STRUCT out += 1 *8out = c out += 8 goto base_type_done :base_type_incomplete_struct ; e.g. struct Foo *x; where struct Foo hasn't been defined *1out = TYPE_VOID out += 1 goto base_type_done :base_type_struct_definition ; @NONSTANDARD: we don't handle bit-fields. local member_base_type local member_prefix local member_prefix_end local member_suffix local member_suffix_end local member_name local member_type local member_align local member_size if c != 0 goto struct_redefinition struct = ident_list_create(8000) ; note: maximum "* 127 members in a single structure or union" C89 ยง 2.2.4.1 *1out = TYPE_STRUCT out += 1 *8out = struct out += 8 types_bytes_used = out - types p += 16 ; skip opening { local offset offset = 0 if *1struct_name == 0 goto struct_unnamed ident_list_add(structures, struct_name, struct) :struct_unnamed :struct_defn_loop if *1p == SYMBOL_RBRACE goto struct_defn_loop_end member_base_type = p p = type_get_base_end(member_base_type) :struct_defn_decl_loop ; handle each element of int x, y[5], *z; member_prefix = p member_prefix_end = type_get_prefix_end(member_prefix) if *1member_prefix_end != TOKEN_IDENTIFIER goto member_no_identifier member_name = member_prefix_end + 8 member_name = *8member_name c = ident_list_lookup_check(struct, member_name, 0) if c == 1 goto duplicate_member member_suffix = member_prefix_end + 16 member_suffix_end = type_get_suffix_end(member_prefix) member_type = types_bytes_used parse_type_declarators(member_prefix, member_prefix_end, member_suffix, member_suffix_end) parse_base_type(member_base_type) ; make sure struct member is aligned member_align = type_alignof(member_type) ; offset = ceil(offset / align) * align offset += member_align - 1 offset /= member_align offset *= member_align if offset ] 0xffffffff goto struct_too_large ;putnln(offset) ; data = (type << 32) | offset c = member_type < 32 c |= offset ident_list_add(struct, member_name, c) member_size = type_sizeof(member_type) offset += member_size * is_struct ; keep offset as 0 if this is a union p = member_suffix_end if *1p == SYMBOL_SEMICOLON goto struct_defn_decl_loop_end if *1p != SYMBOL_COMMA goto struct_bad_declaration p += 16 ; skip comma goto struct_defn_decl_loop :duplicate_member token_error(p, .str_duplicate_member) :str_duplicate_member string Duplicate member in struct/union. byte 0 :struct_defn_decl_loop_end p += 16 ; skip semicolon goto struct_defn_loop :struct_defn_loop_end out = types + types_bytes_used goto base_type_done :struct_redefinition token_error(p, .str_struct_redefinition) :str_struct_redefinition string struct redefinition. byte 0 :struct_bad_declaration token_error(p, .str_struct_bad_declaration) :str_struct_bad_declaration string Bad declaration in struct. byte 0 :struct_too_large token_error(p, .str_struct_too_large) :str_struct_too_large string struct too large (maximum is 4GB). byte 0 :member_no_identifier ; e.g. struct { int; }; token_error(p, .str_member_no_identifier) :str_member_no_identifier string No identifier in struct member. byte 0 :base_type_enum local q local expr *1out = TYPE_INT ; treat any enum as int out += 1 types_bytes_used = out - types p = base_type + 16 if *1p == SYMBOL_LBRACE goto enum_definition if *1p != TOKEN_IDENTIFIER goto bad_base_type ; e.g. enum int x; p += 16 if *1p == SYMBOL_LBRACE goto enum_definition goto base_type_done ; just using an enum type, not defining it. :enum_definition local name local value value = -1 ; consider initial previous value as -1, because -1 + 1 = 0 p += 16 ; skip opening { :enum_defn_loop if *1p == SYMBOL_RBRACE goto enum_defn_loop_end if *1p != TOKEN_IDENTIFIER goto bad_enum_definition p += 8 name = *8p p += 8 if *1p == SYMBOL_COMMA goto enum_defn_no_equals if *1p == SYMBOL_RBRACE goto enum_defn_no_equals if *1p != SYMBOL_EQ goto bad_enum_definition ; e.g. enum { X ! }; ; value provided, e.g. X = 5, p += 16 depth = 0 ; parenthesis depth q = p ; find matching comma/right brace ; -- yes, a comma can appear in an enumerator expression, e.g. ; enum { X = sizeof(struct{int x, y;}) }; ; or enum { X = (enum {A,B})3 }; ; find associated comma or right-brace :enum_comma_loop if depth > 0 goto enum_comma_deep if *1q == SYMBOL_COMMA goto enum_comma_loop_end if *1q == SYMBOL_RBRACE goto enum_comma_loop_end :enum_comma_deep if *1q == TOKEN_EOF goto bad_base_type c = *1q q += 16 if c == SYMBOL_LPAREN goto enum_comma_incdepth if c == SYMBOL_RPAREN goto enum_comma_decdepth goto enum_comma_loop :enum_comma_incdepth depth += 1 goto enum_comma_loop :enum_comma_decdepth depth -= 1 goto enum_comma_loop :enum_comma_loop_end expr = malloc(4000) parse_expression(p, q, expr) evaluate_constant_expression(p, expr, &value) free(expr) if value < -0x80000000 goto bad_enumerator if value > 0x7fffffff goto bad_enumerator ident_list_add(enumerators, name, value) p = q if *1p == SYMBOL_RBRACE goto enum_defn_loop_end p += 16 ; skip , goto enum_defn_loop :bad_enumerator token_error(p, .str_bad_enumerator) :str_bad_enumerator string Enumerators too large for int. byte 0 :enum_defn_no_equals ; no value provided, e.g. X, ; the value of this enumerator is one more than the value of the last one value += 1 ident_list_add(enumerators, name, value) if *1p == SYMBOL_RBRACE goto enum_defn_loop_end p += 16 ; skip , goto enum_defn_loop :enum_defn_loop_end out = types + types_bytes_used ; fix stuff in case there were any types in the enumerator expressions goto base_type_done :bad_enum_definition token_error(base_type, .str_bad_enum_defn) :str_bad_enum_defn string Bad enum definition. byte 0 :base_type_float *1out = TYPE_FLOAT out += 1 goto base_type_done :base_type_void *1out = TYPE_VOID out += 1 goto base_type_done :base_type_typedef p = base_type + 8 c = ident_list_lookup(typedefs, *8p) if c == 0 goto bad_base_type local len len = type_length(c) c += types memcpy(out, c, len) out += len goto base_type_done :bad_base_type token_error(base_type, .str_bad_base_type) :str_bad_base_type string Bad base type. byte 0 ; how many bytes does it take to encode this type? function type_length argument type local p local n p = types + type if *1p <= TYPE_DOUBLE goto return_1 if *1p != TYPE_POINTER goto type_length_not_pointer type += 1 n = type_length(type) return n + 1 :type_length_not_pointer if *1p != TYPE_ARRAY goto type_length_not_array type += 9 n = type_length(type) return n + 9 :type_length_not_array if *1p == TYPE_STRUCT goto return_9 if *1p != TYPE_FUNCTION goto type_length_not_function local start start = type type += 1 :type_length_function_loop p = types + type if *1p == 0 goto type_length_function_loop_end type += type_length(type) goto type_length_function_loop :type_length_function_loop_end type += 1 type += type_length(type) return type - start :type_length_not_function fputs(2, .str_type_length_bad_type) exit(1) :str_type_length_bad_type string Bad type passed to type_length. This shouldn't happen. byte 10 byte 0 ; returns length of type function type_copy_ids argument dest argument src local n n = type_length(src) dest += types src += types memcpy(dest, src, n) return n function type_create_pointer argument type local id local p id = types_bytes_used p = types + id *1p = TYPE_POINTER types_bytes_used += 1 p = id + 1 types_bytes_used += type_copy_ids(p, type) return id function parse_expression argument tokens argument tokens_end argument out local in local a local b local c local p local n local type local best local best_precedence local depth local value local first_token :parse_expression_top ;print_tokens(tokens, tokens_end) type = out + 4 if tokens == tokens_end goto empty_expression p = tokens + 16 if p == tokens_end goto single_token_expression if *1tokens != SYMBOL_LPAREN goto parse_expression_not_entirely_in_parens p = tokens_end - 16 if *1p != SYMBOL_RPAREN goto parse_expression_not_entirely_in_parens depth = 1 ; bracket depth p = tokens + 16 a = tokens_end - 16 ; stop point :expr_paren_check_loop if p >= a goto expr_paren_check_loop_end c = *1p p += 16 if c == SYMBOL_LPAREN goto expr_paren_check_loop_incdepth if c == SYMBOL_RPAREN goto expr_paren_check_loop_decdepth goto expr_paren_check_loop :expr_paren_check_loop_incdepth depth += 1 goto expr_paren_check_loop :expr_paren_check_loop_decdepth depth -= 1 if depth == 0 goto parse_expression_not_entirely_in_parens goto expr_paren_check_loop :expr_paren_check_loop_end ; if we made it this far, the expression is entirely in parenthesis, e.g. (x+2) tokens += 16 tokens_end -= 16 goto parse_expression_top :parse_expression_not_entirely_in_parens ; look for the operator with the lowest precedence not in brackets depth = 0 ; paren/square bracket depth first_token = 1 p = tokens best = 0 best_precedence = 1000 goto expr_find_operator_loop_first :expr_find_operator_loop first_token = 0 :expr_find_operator_loop_first if p >= tokens_end goto expr_find_operator_loop_end n = p c = *1p p += 16 if depth > 0 goto expr_findop_not_new_best if depth < 0 goto expr_too_many_closing_brackets a = operator_precedence(n, first_token) n = a if a == 0xe0 goto select_leftmost ; ensure that the leftmost unary operator is processed first b = operator_right_associative(c) if b != 0 goto select_leftmost ; ensure that the leftmost += / -= / etc. is processed first goto select_rightmost :select_leftmost n += 1 ; fallthrough :select_rightmost if n > best_precedence goto expr_findop_not_new_best ; new best! best = p - 16 ;putc('O) ;putc(':) ;putn(*1best) ;putc(32) ;putc('P) ;putc(':) ;putnln(a) best_precedence = a :expr_findop_not_new_best if c == SYMBOL_LPAREN goto expr_findop_incdepth if c == SYMBOL_RPAREN goto expr_findop_decdepth if c == SYMBOL_LSQUARE goto expr_findop_incdepth if c == SYMBOL_RSQUARE goto expr_findop_decdepth goto expr_find_operator_loop :expr_findop_incdepth depth += 1 goto expr_find_operator_loop :expr_findop_decdepth depth -= 1 goto expr_find_operator_loop :expr_find_operator_loop_end if best == 0 goto unrecognized_expression n = best - tokens c = *1best if best == tokens goto parse_expr_unary ; it's a binary expression. if c == SYMBOL_PLUS_PLUS goto parse_postincrement if c == SYMBOL_MINUS_MINUS goto parse_postdecrement if c == SYMBOL_QUESTION goto parse_conditional *1out = binop_symbol_to_expression_type(c) c = *1out out += 8 if c == EXPRESSION_DOT goto parse_expr_member if c == EXPRESSION_ARROW goto parse_expr_member a = out + 4 ; type of first operand out = parse_expression(tokens, best, out) ; first operand p = best + 16 b = out + 4 ; type of second operand if c != EXPRESSION_SUBSCRIPT goto binary_not_subscript tokens_end -= 16 if *1tokens_end != SYMBOL_RSQUARE goto unrecognized_expression :binary_not_subscript out = parse_expression(p, tokens_end, out) ; second operand if c == EXPRESSION_LSHIFT goto type_shift if c == EXPRESSION_RSHIFT goto type_shift if c == EXPRESSION_SUBSCRIPT goto type_subscript if c == EXPRESSION_EQ goto type_int if c == EXPRESSION_NEQ goto type_int if c == EXPRESSION_LEQ goto type_int if c == EXPRESSION_GEQ goto type_int if c == EXPRESSION_LT goto type_int if c == EXPRESSION_GT goto type_int if c == EXPRESSION_COMMA goto type_binary_right if c == EXPRESSION_EQ goto type_binary_left if c == EXPRESSION_ASSIGN_ADD goto type_binary_left if c == EXPRESSION_ASSIGN_SUB goto type_binary_left if c == EXPRESSION_ASSIGN_MUL goto type_binary_left if c == EXPRESSION_ASSIGN_DIV goto type_binary_left if c == EXPRESSION_ASSIGN_REMAINDER goto type_binary_left if c == EXPRESSION_ASSIGN_AND goto type_binary_left_integer if c == EXPRESSION_ASSIGN_XOR goto type_binary_left_integer if c == EXPRESSION_ASSIGN_OR goto type_binary_left_integer if c == EXPRESSION_ASSIGN_LSHIFT goto type_binary_left_integer if c == EXPRESSION_ASSIGN_RSHIFT goto type_binary_left_integer if c == EXPRESSION_LOGICAL_OR goto type_int if c == EXPRESSION_LOGICAL_AND goto type_int if c == EXPRESSION_BITWISE_AND goto type_binary_usual_integer if c == EXPRESSION_BITWISE_XOR goto type_binary_usual_integer if c == EXPRESSION_BITWISE_OR goto type_binary_usual_integer if c == EXPRESSION_ADD goto type_plus if c == EXPRESSION_SUB goto type_minus if c == EXPRESSION_MUL goto type_binary_usual if c == EXPRESSION_DIV goto type_binary_usual if c == EXPRESSION_REMAINDER goto type_binary_usual_integer fputs(2, .str_binop_this_shouldnt_happen) exit(1) :str_binop_this_shouldnt_happen string Bad binop symbol (this shouldn't happen). byte 10 byte 0 :type_plus p = types + *4a if *1p == TYPE_POINTER goto type_binary_left ; pointer plus integer p = types + *4b if *1p == TYPE_POINTER goto type_binary_right ; integer plus pointer goto type_binary_usual :type_minus p = types + *4a if *1p == TYPE_POINTER goto type_minus_left_ptr goto type_binary_usual :type_minus_left_ptr p = types + *4b if *1p == TYPE_POINTER goto type_long ; pointer difference goto type_binary_left ; pointer minus integer :type_subscript p = types + *4a if *1p == TYPE_POINTER goto type_subscript_pointer if *1p == TYPE_ARRAY goto type_subscript_array goto subscript_bad_type :type_subscript_pointer *4type = *4a + 1 return out :type_subscript_array *4type = *4a + 9 return out :subscript_bad_type token_error(tokens, .str_subscript_bad_type) :str_subscript_bad_type string Subscript of non-pointer type. byte 0 ; apply the "usual conversions" :type_binary_usual *4type = expr_binary_type_usual_conversions(tokens, *4a, *4b) return out ; like type_binary_usual, but the operands must be integers :type_binary_usual_integer *4type = expr_binary_type_usual_conversions(tokens, *4a, *4b) p = types + *4type if *1p >= TYPE_FLOAT goto expr_binary_bad_types return out :type_binary_left_integer p = types + *4a if *1p >= TYPE_FLOAT goto expr_binary_bad_types p = types + *4b if *1p >= TYPE_FLOAT goto expr_binary_bad_types goto type_binary_left :type_binary_left *4type = *4a return out :type_binary_right *4type = *4b return out :type_shift p = types + *4a if *1p >= TYPE_FLOAT goto expr_binary_bad_types p = types + *4b if *1p >= TYPE_FLOAT goto expr_binary_bad_types *4type = type_promotion(*4a) return out ; the type here is just int :type_int *4type = TYPE_INT return out :type_long *4type = TYPE_LONG return out :expr_binary_bad_types bad_types_to_operator(tokens, *4a, *4b) :parse_expr_unary if c == KEYWORD_SIZEOF goto parse_sizeof *1out = unary_op_to_expression_type(c) c = *1out if c == EXPRESSION_CAST goto parse_cast out += 8 a = out + 4 ; type of operand p = tokens + 16 out = parse_expression(p, tokens_end, out) p = types + *4a if c == EXPRESSION_BITWISE_NOT goto unary_type_integral if c == EXPRESSION_UNARY_PLUS goto unary_type_promote if c == EXPRESSION_UNARY_MINUS goto unary_type_promote if c == EXPRESSION_LOGICAL_NOT goto unary_type_logical_not if c == EXPRESSION_ADDRESS_OF goto unary_address_of if c == EXPRESSION_DEREFERENCE goto unary_dereference if c == EXPRESSION_PRE_INCREMENT goto unary_type_arithmetic_nopromote if c == EXPRESSION_PRE_DECREMENT goto unary_type_arithmetic_nopromote fputs(2, .str_unop_this_shouldnt_happen) exit(1) :str_unop_this_shouldnt_happen string Bad unary symbol (this shouldn't happen). byte 10 byte 0 :parse_cast local cast_base_type local cast_prefix local cast_suffix local cast_suffix_end cast_base_type = best + 16 cast_prefix = type_get_base_end(cast_base_type) cast_suffix = type_get_prefix_end(cast_prefix) cast_suffix_end = type_get_suffix_end(cast_prefix) a = types_bytes_used parse_type_declarators(cast_prefix, cast_suffix, cast_suffix, cast_suffix_end) parse_base_type(cast_base_type) p = cast_suffix_end if *1p != SYMBOL_RPAREN goto bad_cast ; e.g. (int ,)5 out += 4 *4out = a out += 4 p += 16 out = parse_expression(p, tokens_end, out) return out :bad_cast token_error(tokens, .str_bad_cast) :str_bad_cast string Bad cast. byte 0 :unary_address_of *4type = type_create_pointer(*4a) return out :unary_dereference if *1p != TYPE_POINTER goto unary_bad_type *4type = *4a + 1 return out :unary_type_logical_not if *1p > TYPE_POINTER goto unary_bad_type *4type = TYPE_INT return out :unary_type_integral if *1p >= TYPE_FLOAT goto unary_bad_type goto unary_type_promote :unary_type_promote if *1p > TYPE_DOUBLE goto unary_bad_type *4type = type_promotion(*4a) return out :unary_type_arithmetic_nopromote if *1p > TYPE_DOUBLE goto unary_bad_type *4type = *4a return out :unary_bad_type fprint_token_location(1, tokens) puts(.str_unary_bad_type) print_type(*4a) putc(10) exit(1) :str_unary_bad_type string : Bad type for unary operator: byte 32 byte 0 :parse_sizeof local sizeof_base_type local sizeof_prefix local sizeof_suffix local sizeof_suffix_end *1out = EXPRESSION_CONSTANT_INT out += 4 *1out = TYPE_UNSIGNED_LONG out += 4 p = best + 16 if *1p != SYMBOL_LPAREN goto parse_sizeof_expr p += 16 b = token_is_type(p) if b == 0 goto parse_sizeof_expr ; it's a type, e.g. sizeof(int) sizeof_base_type = p sizeof_prefix = type_get_base_end(sizeof_base_type) sizeof_suffix = type_get_prefix_end(sizeof_prefix) sizeof_suffix_end = type_get_suffix_end(sizeof_prefix) p = sizeof_suffix_end a = types_bytes_used parse_type_declarators(sizeof_prefix, sizeof_suffix, sizeof_suffix, sizeof_suffix_end) parse_base_type(sizeof_base_type) if *1p != SYMBOL_RPAREN goto bad_expression ; e.g. sizeof(int ,) *8out = type_sizeof(a) goto parse_sizeof_finish :parse_sizeof_expr ; it's an expression, e.g. sizeof(x+3) local temp temp = malloc(4000) p = best + 16 parse_expression(p, tokens_end, temp) p = temp + 4 *8out = type_sizeof(*4p) free(temp) :parse_sizeof_finish out += 8 return out :parse_expr_member ; -> or . p = best + 16 if *1p != TOKEN_IDENTIFIER goto bad_expression a = out + 4 ; pointer to type ID out = parse_expression(tokens, best, out) a = types + *4a if c == EXPRESSION_DOT goto type_dot if *1a != TYPE_POINTER goto arrow_non_pointer a += 1 :type_dot if *1a != TYPE_STRUCT goto member_non_struct a += 1 a = *8a ; pointer to struct data p += 8 c = ident_list_lookup(a, *8p) if c == 0 goto member_not_in_struct *8out = c & 0xffffffff ; offset *4type = c > 32 ; type out += 8 p += 8 if p != tokens_end goto bad_expression ; e.g. foo->bar hello return out :arrow_non_pointer token_error(p, .str_arrow_non_pointer) :str_arrow_non_pointer string Trying to use -> operator on a non-pointer type. byte 0 :member_non_struct token_error(p, .str_member_non_struct) :str_member_non_struct string Trying to access member of something other than a (complete) structure/union. byte 0 :member_not_in_struct token_error(p, .str_member_not_in_struct) :str_member_not_in_struct string Trying to access non-existent member of structure or union. byte 0 :parse_conditional depth = 0 ; bracket depth n = 0 ; ? : depth ; find : associated with this ? p = best + 16 :parse_conditional_loop if p >= tokens_end goto bad_expression if *1p == SYMBOL_QUESTION goto parse_cond_incn if *1p == SYMBOL_COLON goto parse_cond_decn if *1p == SYMBOL_LPAREN goto parse_cond_incdepth if *1p == SYMBOL_RPAREN goto parse_cond_decdepth if *1p == SYMBOL_LSQUARE goto parse_cond_incdepth if *1p == SYMBOL_RSQUARE goto parse_cond_decdepth :parse_cond_cont p += 16 goto parse_conditional_loop :parse_cond_incdepth depth += 1 goto parse_cond_cont :parse_cond_decdepth depth -= 1 goto parse_cond_cont :parse_cond_incn n += 1 goto parse_cond_cont :parse_cond_decn n -= 1 if n >= 0 goto parse_cond_cont if depth > 0 goto parse_cond_cont ; okay, q now points to the : *1out = EXPRESSION_CONDITIONAL out += 8 out = parse_expression(tokens, best, out) a = out + 4 ; type of left branch of conditional best += 16 out = parse_expression(best, p, out) b = out + 4 ; type of right branch of conditional p += 16 out = parse_expression(p, tokens_end, out) p = types + *4a if *1p == TYPE_STRUCT goto parse_cond_ltype if *1p == TYPE_VOID goto parse_cond_ltype if *1p == TYPE_POINTER goto parse_cond_ltype ; @NONSTANDARD: we don't handle sizeof *(0 ? (void*)0 : "hello") correctly--it should be 1 (a standard-compliant implementation is annoyingly complicated) *4type = expr_binary_type_usual_conversions(tokens, *4a, *4b) return out :parse_cond_ltype ; no conversions *4type = *4a return out :parse_postincrement *1out = EXPRESSION_POST_INCREMENT p = tokens_end - 16 if *1p != SYMBOL_PLUS_PLUS goto bad_expression ; e.g. a ++ b out += 8 a = out + 4 ; type of operand out = parse_expression(tokens, p, out) *4type = *4a ; this expression's type is the operand's type (yes even for types smaller than int) return out :parse_postdecrement *1out = EXPRESSION_POST_DECREMENT p = tokens_end - 16 if *1p != SYMBOL_MINUS_MINUS goto bad_expression ; e.g. a -- b out += 8 a = out + 4 ; type of operand out = parse_expression(tokens, p, out) *4type = *4a ; type of this = type of operand return out :single_token_expression in = tokens c = *1in if c == TOKEN_CONSTANT_INT goto expression_integer if c == TOKEN_CONSTANT_CHAR goto expression_integer ; character constants are basically the same as integer constants if c == TOKEN_CONSTANT_FLOAT goto expression_float if c == TOKEN_STRING_LITERAL goto expression_string_literal if c == TOKEN_IDENTIFIER goto expression_identifier goto unrecognized_expression :expression_identifier in += 8 a = *8in in += 8 ; check if it's an enumerator c = ident_list_lookup_check(enumerators, a, &n) if c == 0 goto not_enumerator ; it is an enumerator *1out = EXPRESSION_CONSTANT_INT out += 4 *4out = TYPE_INT out += 4 *8out = n out += 8 return out :not_enumerator ; @TODO: check if it's a local variable ; check if it's a global c = ident_list_lookup(global_variables, a) if c == 0 goto not_global ; it is a global variable *1out = EXPRESSION_GLOBAL_VARIABLE out += 4 *4out = c > 32 ; extract type out += 4 *8out = c & 0xffffffff ; extract address out += 8 return out :not_global in -= 16 token_error(in, .str_undeclared_variable) :str_undeclared_variable string Undeclared variable. byte 0 :expression_integer *1out = EXPRESSION_CONSTANT_INT p = in + 8 value = *8p p = out + 8 *8p = value p = in + 1 a = int_suffix_to_type(*1p) ; what the suffix says the type should be b = int_value_to_type(value) ; what the value says the type should be (if the value is too large to fit in int) a = max_signed(a, b) ; take the maximum of the two types ; make sure that if the integer has a u suffix, the type will be unsigned a &= b | 0xfe p = out + 4 *4p = a in += 16 out += 16 return out :expression_float *1out = EXPRESSION_CONSTANT_FLOAT p = in + 8 value = *8p p = out + 8 *8p = value p = in + 1 a = float_suffix_to_type(*1p) p = out + 4 *4p = a in += 16 out += 16 return out :expression_string_literal *1out = EXPRESSION_STRING_LITERAL p = in + 8 value = *8p p = out + 8 *8p = value ; we already know this is char* p = out + 4 *4p = TYPE_POINTER_TO_CHAR in += 16 out += 16 return out :empty_expression token_error(tokens, .str_empty_expression) :str_empty_expression string Empty expression. byte 0 :bad_expression token_error(tokens, .str_bad_expression) :str_bad_expression string Bad expression. byte 0 :unrecognized_expression token_error(tokens, .str_unrecognized_expression) :str_unrecognized_expression string Unrecognized expression. byte 0 :expr_too_many_closing_brackets token_error(tokens, .str_too_many_closing_brackets) :str_too_many_closing_brackets string Too many closing brackets. byte 0 :return_type_int return TYPE_INT :return_type_long return TYPE_LONG :return_type_unsigned_int return TYPE_UNSIGNED_INT :return_type_unsigned_long return TYPE_UNSIGNED_LONG :return_type_float return TYPE_FLOAT :return_type_double return TYPE_DOUBLE function type_sizeof argument type local p local c p = types + type c = *1p if c == TYPE_CHAR goto return_1 if c == TYPE_UNSIGNED_CHAR goto return_1 if c == TYPE_SHORT goto return_2 if c == TYPE_UNSIGNED_SHORT goto return_2 if c == TYPE_INT goto return_4 if c == TYPE_UNSIGNED_INT goto return_4 if c == TYPE_LONG goto return_8 if c == TYPE_UNSIGNED_LONG goto return_8 if c == TYPE_FLOAT goto return_4 if c == TYPE_DOUBLE goto return_8 if c == TYPE_VOID goto return_1 if c == TYPE_POINTER goto return_8 if c == TYPE_FUNCTION goto return_8 if c == TYPE_ARRAY goto sizeof_array if c == TYPE_STRUCT goto sizeof_struct fputs(2, .str_sizeof_bad) exit(1) :str_sizeof_bad string type_sizeof bad type. byte 10 byte 0 :sizeof_array local n p += 1 n = *8p p += 8 p -= types c = type_sizeof(p) return n * c :sizeof_struct ; size of struct is offset of last member + size of last member, ; rounded up to fit alignment local align local offset local member align = type_alignof(type) p += 1 member = *8p :sizeof_struct_loop if *1member == 0 goto sizeof_struct_loop_end member = memchr(member, 0) ; don't care about name member += 1 ; skip null terminator c = *8member member += 8 offset = c & 0xffffffff c >= 32 ; extract type offset += type_sizeof(c) goto sizeof_struct_loop :sizeof_struct_loop_end offset += align - 1 offset /= align offset *= align return offset function type_alignof argument type local p local c p = types + type c = *1p if c == TYPE_CHAR goto return_1 if c == TYPE_UNSIGNED_CHAR goto return_1 if c == TYPE_SHORT goto return_2 if c == TYPE_UNSIGNED_SHORT goto return_2 if c == TYPE_INT goto return_4 if c == TYPE_UNSIGNED_INT goto return_4 if c == TYPE_LONG goto return_8 if c == TYPE_UNSIGNED_LONG goto return_8 if c == TYPE_FLOAT goto return_4 if c == TYPE_DOUBLE goto return_8 if c == TYPE_VOID goto return_1 if c == TYPE_POINTER goto return_8 if c == TYPE_FUNCTION goto return_8 if c == TYPE_ARRAY goto alignof_array if c == TYPE_STRUCT goto alignof_struct fputs(2, .str_alignof_bad) exit(1) :str_alignof_bad string type_alignof bad type. byte 10 byte 0 :alignof_struct ; alignment of struct is max alignment of members local align local member local a align = 1 p += 1 member = *8p :alignof_struct_loop if *1member == 0 goto alignof_struct_loop_end member = memchr(member, 0) ; don't care about name member += 1 ; skip null terminator c = *8member member += 8 c >= 32 ; ignore offset a = type_alignof(c) if a <= align goto alignof_struct_loop align = a goto alignof_struct_loop :alignof_struct_loop_end return align :alignof_array p = type + 9 ; skip TYPE_ARRAY and size return type_alignof(p) ; evaluate an expression which can be the size of an array, e.g. ; enum { A, B, C }; ; int x[A * sizeof(float) + 3 << 5]; ; @NONSTANDARD: doesn't handle floats, but really why would you use floats in an array size ; e.g. SomeType x[(int)3.3]; ; this is also used for #if evaluation ; token is used for error messages (e.g. if this "constant" expression is *x or something) ; NOTE: this returns the end of the expression, not the value (which is stored in *8p_value) function evaluate_constant_expression argument token argument expr argument p_value local a local b local c local p local mask local type type = expr + 4 type = *4type c = *1expr if c == EXPRESSION_CONSTANT_INT goto eval_constant_int if c == EXPRESSION_UNARY_PLUS goto eval_unary_plus if c == EXPRESSION_UNARY_MINUS goto eval_unary_minus if c == EXPRESSION_BITWISE_NOT goto eval_bitwise_not if c == EXPRESSION_LOGICAL_NOT goto eval_logical_not if c == EXPRESSION_CAST goto eval_cast if c == EXPRESSION_ADD goto eval_add if c == EXPRESSION_SUB goto eval_sub if c == EXPRESSION_MUL goto eval_mul if c == EXPRESSION_DIV goto eval_div if c == EXPRESSION_REMAINDER goto eval_remainder if c == EXPRESSION_LSHIFT goto eval_lshift if c == EXPRESSION_RSHIFT goto eval_rshift if c == EXPRESSION_EQ goto eval_eq if c == EXPRESSION_NEQ goto eval_neq if c == EXPRESSION_LT goto eval_lt if c == EXPRESSION_GT goto eval_gt if c == EXPRESSION_LEQ goto eval_leq if c == EXPRESSION_GEQ goto eval_geq if c == EXPRESSION_BITWISE_AND goto eval_bitwise_and if c == EXPRESSION_BITWISE_OR goto eval_bitwise_or if c == EXPRESSION_BITWISE_XOR goto eval_bitwise_xor if c == EXPRESSION_LOGICAL_AND goto eval_logical_and if c == EXPRESSION_LOGICAL_OR goto eval_logical_or if c == EXPRESSION_CONDITIONAL goto eval_conditional token_error(token, .str_eval_bad_exprtype) :str_eval_bad_exprtype string Can't evaluate constant expression. byte 0 :eval_cast p = types + type c = *1p if c == TYPE_VOID goto eval_cast_bad_type ; @NONSTANDARD: we don't support, for example, int x[(int)(float)5]; if c == TYPE_FLOAT goto eval_cast_bad_type if c == TYPE_DOUBLE goto eval_cast_bad_type if c > TYPE_POINTER goto eval_cast_bad_type expr += 8 expr = evaluate_constant_expression(token, expr, p_value) goto eval_fit_to_type :eval_cast_bad_type token_error(token, .str_eval_cast_bad_type) :str_eval_cast_bad_type string Bad type for constant cast (note: floating-point casts are not supported even though they are standard). byte 0 :eval_constant_int expr += 8 *8p_value = *8expr expr += 8 return expr :eval_unary_plus expr += 8 expr = evaluate_constant_expression(token, expr, p_value) return expr :eval_unary_minus expr += 8 expr = evaluate_constant_expression(token, expr, &a) *8p_value = 0 - a goto eval_fit_to_type :eval_bitwise_not expr += 8 expr = evaluate_constant_expression(token, expr, &a) *8p_value = ~a goto eval_fit_to_type :eval_logical_not expr += 8 expr = evaluate_constant_expression(token, expr, &a) if a == 0 goto eval_value_1 goto eval_value_0 :eval_add expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a + b goto eval_fit_to_type :eval_sub expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a - b goto eval_fit_to_type :eval_mul expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a * b goto eval_fit_to_type :eval_div expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) if *1p == TYPE_UNSIGNED_LONG goto eval_div_unsigned ; division is signed or uses a small type, so we can use 64-bit signed division *8p_value = a / b goto eval_fit_to_type :eval_div_unsigned ; must use unsigned division divmod_unsigned(a, b, p_value, &a) goto eval_fit_to_type :eval_remainder expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) p = types + type if *1p == TYPE_UNSIGNED_LONG goto eval_rem_unsigned *8p_value = a % b goto eval_fit_to_type :eval_rem_unsigned divmod_unsigned(a, b, &a, p_value) goto eval_fit_to_type :eval_lshift expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a < b goto eval_fit_to_type :eval_rshift expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) p = types + type p = *1p p &= 1 ; signed types are odd if p == 1 goto eval_signed_rshift *8p_value = a > b goto eval_fit_to_type :eval_signed_rshift local v mask = a > 63 ; sign bit ; sign extension mask <= b mask -= 1 mask <= 64 - b v = a > b v += mask *8p_value = v goto eval_fit_to_type ; comparison masks: ; 1 = less than ; 2 = equal to ; 4 = greater than ; e.g. not-equal is 1|4 = 5 because not equal = less than or greater than :eval_eq mask = 2 goto eval_comparison :eval_neq mask = 5 goto eval_comparison :eval_lt mask = 1 goto eval_comparison :eval_gt mask = 4 goto eval_comparison :eval_leq mask = 3 goto eval_comparison :eval_geq mask = 6 goto eval_comparison :eval_comparison expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) p = types + type p = *1p p &= 1 if a == b goto eval_comparison_eq ; for checking < and >, we care about whether a and b are signed if p == 1 goto eval_signed_comparison if a ] b goto eval_comparison_gt goto eval_comparison_lt :eval_signed_comparison if a > b goto eval_comparison_gt goto eval_comparison_lt :eval_comparison_eq ; a == b mask &= 2 goto eval_comparison_done :eval_comparison_lt ; a < b mask &= 1 goto eval_comparison_done :eval_comparison_gt ; a > b mask &= 4 goto eval_comparison_done :eval_comparison_done if mask != 0 goto eval_value_1 goto eval_value_0 :eval_bitwise_and expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a & b goto eval_fit_to_type :eval_bitwise_or expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a | b goto eval_fit_to_type :eval_bitwise_xor expr += 8 expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) *8p_value = a ^ b goto eval_fit_to_type :eval_logical_and expr += 8 expr = evaluate_constant_expression(token, expr, &a) if a == 0 goto eval_value_0 expr = evaluate_constant_expression(token, expr, &b) if b == 0 goto eval_value_0 goto eval_value_1 :eval_logical_or expr += 8 expr = evaluate_constant_expression(token, expr, &a) if a != 0 goto eval_value_1 expr = evaluate_constant_expression(token, expr, &b) if b != 0 goto eval_value_1 goto eval_value_0 :eval_conditional expr += 8 expr = evaluate_constant_expression(token, expr, &mask) expr = evaluate_constant_expression(token, expr, &a) expr = evaluate_constant_expression(token, expr, &b) if mask == 0 goto eval_conditional_b *8p_value = a goto eval_fit_to_type :eval_conditional_b *8p_value = b goto eval_fit_to_type :eval_fit_to_type *8p_value = fit_to_type(*8p_value, type) return expr :eval_value_0 *8p_value = 0 return expr :eval_value_1 *8p_value = 1 return expr ; value is the output of some arithmetic expression; correct it to be within the range of type. function fit_to_type argument value argument type local c local s c = types + type c = *1c if c == TYPE_CHAR goto fit_to_type_char if c == TYPE_UNSIGNED_CHAR goto fit_to_type_uchar if c == TYPE_SHORT goto fit_to_type_short if c == TYPE_UNSIGNED_SHORT goto fit_to_type_ushort if c == TYPE_INT goto fit_to_type_int if c == TYPE_UNSIGNED_INT goto fit_to_type_uint if c == TYPE_LONG goto fit_to_type_long if c == TYPE_UNSIGNED_LONG goto fit_to_type_ulong if c == TYPE_POINTER goto fit_to_type_ulong fputs(2, .str_bad_fit_to_type) exit(1) :str_bad_fit_to_type string Bad type passed to fit_to_type. byte 10 byte 0 ; yes, signed integer overflow is undefined behavior and ; casting to a signed integer is implementation-defined; ; i'm going to play it safe and implement it properly :fit_to_type_char value &= 0xff s = value > 7 ; sign bit value += s * 0xffffffffffffff00 ; sign-extend return value :fit_to_type_uchar value &= 0xff return value :fit_to_type_short value &= 0xffff s = value > 15 ; sign bit value += s * 0xffffffffffff0000 ; sign-extend return value :fit_to_type_ushort value &= 0xffff return value :fit_to_type_int value &= 0xffffffff s = value > 31 ; sign bit value += s * 0xffffffff00000000 ; sign-extend return value :fit_to_type_uint value &= 0xffffffff return value :fit_to_type_long :fit_to_type_ulong return value ; the "usual conversions" for binary operators, as the C standard calls it function expr_binary_type_usual_conversions argument token ; for errors argument type1 argument type2 local ptype1 local ptype2 local kind1 local kind2 if type1 == 0 goto return_0 if type2 == 0 goto return_0 ptype1 = types + type1 ptype2 = types + type2 kind1 = *1ptype1 kind2 = *1ptype2 if kind1 > TYPE_DOUBLE goto usual_bad_types_to_operator if kind2 > TYPE_DOUBLE goto usual_bad_types_to_operator ; "if either operand has type double, the other operand is converted to double" if kind1 == TYPE_DOUBLE goto return_type_double if kind2 == TYPE_DOUBLE goto return_type_double ; "if either operand has type float, the other operand is converted to float" if kind1 == TYPE_FLOAT goto return_type_float if kind2 == TYPE_FLOAT goto return_type_float ; "If either operand has type unsigned long int, the other operand is converted to unsigned long int" if kind1 == TYPE_UNSIGNED_LONG goto return_type_unsigned_long if kind2 == TYPE_UNSIGNED_LONG goto return_type_unsigned_long ; "if either operand has type long int, the other operand is converted to long int" if kind1 == TYPE_LONG goto return_type_long if kind2 == TYPE_LONG goto return_type_long ; "if either operand has type unsigned int, the other operand is converted to unsigned int." if kind1 == TYPE_UNSIGNED_INT goto return_type_unsigned_int if kind2 == TYPE_UNSIGNED_INT goto return_type_unsigned_int ; "Otherwise, both operands have type int." goto return_type_int :str_space_and_space string and byte 32 byte 0 :usual_bad_types_to_operator bad_types_to_operator(token, type1, type2) function bad_types_to_operator argument token argument type1 argument type2 fprint_token_location(1, token) puts(.str_bad_types_to_operator) print_type(type1) puts(.str_space_and_space) print_type(type2) putc(10) exit(1) :str_bad_types_to_operator string : Bad types to operator: byte 32 byte 0 function type_promotion argument type local p p = types + type if *1p < TYPE_INT goto return_type_int return type ; return precedence of given operator token, or 0xffff if not an operator function operator_precedence argument token argument is_first local op local b if is_first != 0 goto operator_precedence_unary ; if an operator is preceded by another, it must be a unary operator, e.g. ; in 5 + *x, * is a unary operator op = token - 16 op = *1op if op == SYMBOL_RPAREN goto figre_out_rparen_arity op = is_operator(op) ; if an operator is immediately followed by another (including lparen), the second must be ; unary. if op != 0 goto operator_precedence_unary :operator_precedence_binary op = *1token ; see "C OPERATOR PRECEDENCE" in constants.b if op == SYMBOL_COMMA goto return_0x10 if op == SYMBOL_EQ goto return_0x20 if op == SYMBOL_PLUS_EQ goto return_0x20 if op == SYMBOL_MINUS_EQ goto return_0x20 if op == SYMBOL_TIMES_EQ goto return_0x20 if op == SYMBOL_DIV_EQ goto return_0x20 if op == SYMBOL_PERCENT_EQ goto return_0x20 if op == SYMBOL_LSHIFT_EQ goto return_0x20 if op == SYMBOL_RSHIFT_EQ goto return_0x20 if op == SYMBOL_AND_EQ goto return_0x20 if op == SYMBOL_OR_EQ goto return_0x20 if op == SYMBOL_XOR_EQ goto return_0x20 if op == SYMBOL_QUESTION goto return_0x30 if op == SYMBOL_OR_OR goto return_0x40 if op == SYMBOL_AND_AND goto return_0x50 if op == SYMBOL_OR goto return_0x60 if op == SYMBOL_XOR goto return_0x70 if op == SYMBOL_AND goto return_0x80 if op == SYMBOL_EQ_EQ goto return_0x90 if op == SYMBOL_NOT_EQ goto return_0x90 if op == SYMBOL_LT goto return_0xa0 if op == SYMBOL_GT goto return_0xa0 if op == SYMBOL_LT_EQ goto return_0xa0 if op == SYMBOL_GT_EQ goto return_0xa0 if op == SYMBOL_LSHIFT goto return_0xb0 if op == SYMBOL_RSHIFT goto return_0xb0 if op == SYMBOL_PLUS goto return_0xc0 if op == SYMBOL_MINUS goto return_0xc0 if op == SYMBOL_TIMES goto return_0xd0 if op == SYMBOL_DIV goto return_0xd0 if op == SYMBOL_PERCENT goto return_0xd0 if op == SYMBOL_ARROW goto return_0xf0 if op == SYMBOL_DOT goto return_0xf0 if op == SYMBOL_LPAREN goto return_0xf0 ; function call if op == SYMBOL_LSQUARE goto return_0xf0 ; subscript if op == SYMBOL_PLUS_PLUS goto return_0xf0 if op == SYMBOL_MINUS_MINUS goto return_0xf0 return 0xffff :operator_precedence_unary op = *1token if op == KEYWORD_SIZEOF goto return_0xe0 if op == SYMBOL_PLUS_PLUS goto return_0xe0 if op == SYMBOL_MINUS_MINUS goto return_0xe0 if op == SYMBOL_AND goto return_0xe0 if op == SYMBOL_TIMES goto return_0xe0 if op == SYMBOL_PLUS goto return_0xe0 if op == SYMBOL_MINUS goto return_0xe0 if op == SYMBOL_TILDE goto return_0xe0 if op == SYMBOL_NOT goto return_0xe0 if op == SYMBOL_LPAREN goto cast_precedence return 0xffff :cast_precedence ; make sure this actually is a cast ; this is necessary to handle both ; - (x)->something ; and - (int)x->something ; correctly (in the first case, the arrow is the top-level operator, but in the second, the cast is) token += 16 b = token_is_type(token) if b == 0 goto return_0xffff goto return_0xd8 ; it's a cast :figre_out_rparen_arity ; given that the token before this one is a right-parenthesis, figure out if ; this is a unary or binary operator. this is (annoyingly) necessary, because: ; (int)-x; /* cast processed first */ ; (y)-x; /* subtraction processed first */ local p p = token - 16 token_reverse_to_matching_lparen(&p) p += 16 b = token_is_type(p) if b != 0 goto operator_precedence_unary ; e.g. (int)-x; goto operator_precedence_binary ; e.g. (y)-x; function unary_op_to_expression_type argument op if op == SYMBOL_PLUS_PLUS goto return_EXPRESSION_PRE_INCREMENT if op == SYMBOL_MINUS_MINUS goto return_EXPRESSION_PRE_DECREMENT if op == SYMBOL_AND goto return_EXPRESSION_ADDRESS_OF if op == SYMBOL_TIMES goto return_EXPRESSION_DEREFERENCE if op == SYMBOL_PLUS goto return_EXPRESSION_UNARY_PLUS if op == SYMBOL_MINUS goto return_EXPRESSION_UNARY_MINUS if op == SYMBOL_TILDE goto return_EXPRESSION_BITWISE_NOT if op == SYMBOL_NOT goto return_EXPRESSION_LOGICAL_NOT if op == SYMBOL_LPAREN goto return_EXPRESSION_CAST return 0 :return_EXPRESSION_PRE_INCREMENT return EXPRESSION_PRE_INCREMENT :return_EXPRESSION_PRE_DECREMENT return EXPRESSION_PRE_INCREMENT :return_EXPRESSION_ADDRESS_OF return EXPRESSION_ADDRESS_OF :return_EXPRESSION_DEREFERENCE return EXPRESSION_DEREFERENCE :return_EXPRESSION_UNARY_PLUS return EXPRESSION_UNARY_PLUS :return_EXPRESSION_UNARY_MINUS return EXPRESSION_UNARY_MINUS :return_EXPRESSION_BITWISE_NOT return EXPRESSION_BITWISE_NOT :return_EXPRESSION_LOGICAL_NOT return EXPRESSION_LOGICAL_NOT :return_EXPRESSION_CAST return EXPRESSION_CAST ; is this operator right-associative? most C operators are left associative, ; but += / -= / etc. are not function operator_right_associative argument op if op == SYMBOL_QUESTION goto return_1 if op < SYMBOL_EQ goto return_0 if op > SYMBOL_OR_EQ goto return_0 goto return_1 :binop_table byte SYMBOL_COMMA byte EXPRESSION_COMMA byte SYMBOL_EQ byte EXPRESSION_ASSIGN byte SYMBOL_PLUS_EQ byte EXPRESSION_ASSIGN_ADD byte SYMBOL_MINUS_EQ byte EXPRESSION_ASSIGN_SUB byte SYMBOL_TIMES_EQ byte EXPRESSION_ASSIGN_MUL byte SYMBOL_DIV_EQ byte EXPRESSION_ASSIGN_DIV byte SYMBOL_PERCENT_EQ byte EXPRESSION_ASSIGN_REMAINDER byte SYMBOL_LSHIFT_EQ byte EXPRESSION_ASSIGN_LSHIFT byte SYMBOL_RSHIFT_EQ byte EXPRESSION_ASSIGN_RSHIFT byte SYMBOL_AND_EQ byte EXPRESSION_ASSIGN_AND byte SYMBOL_OR_EQ byte EXPRESSION_ASSIGN_OR byte SYMBOL_XOR_EQ byte EXPRESSION_ASSIGN_XOR byte SYMBOL_OR_OR byte EXPRESSION_LOGICAL_OR byte SYMBOL_AND_AND byte EXPRESSION_LOGICAL_AND byte SYMBOL_OR byte EXPRESSION_BITWISE_OR byte SYMBOL_XOR byte EXPRESSION_BITWISE_XOR byte SYMBOL_AND byte EXPRESSION_BITWISE_AND byte SYMBOL_EQ_EQ byte EXPRESSION_EQ byte SYMBOL_NOT_EQ byte EXPRESSION_NEQ byte SYMBOL_LT byte EXPRESSION_LT byte SYMBOL_GT byte EXPRESSION_GT byte SYMBOL_LT_EQ byte EXPRESSION_LEQ byte SYMBOL_GT_EQ byte EXPRESSION_GEQ byte SYMBOL_LSHIFT byte EXPRESSION_LSHIFT byte SYMBOL_RSHIFT byte EXPRESSION_RSHIFT byte SYMBOL_PLUS byte EXPRESSION_ADD byte SYMBOL_MINUS byte EXPRESSION_SUB byte SYMBOL_TIMES byte EXPRESSION_MUL byte SYMBOL_DIV byte EXPRESSION_DIV byte SYMBOL_PERCENT byte EXPRESSION_REMAINDER byte SYMBOL_ARROW byte EXPRESSION_ARROW byte SYMBOL_DOT byte EXPRESSION_DOT byte SYMBOL_LSQUARE byte EXPRESSION_SUBSCRIPT byte 0 byte 0 function binop_symbol_to_expression_type argument op local p p = .binop_table :binop_symbol_to_expression_type_loop if *1p == op goto binop_symbol_to_expression_type_found p += 2 if *1p != 0 goto binop_symbol_to_expression_type_loop return 0 :binop_symbol_to_expression_type_found p += 1 return *1p function is_operator argument symbol local b b = binop_symbol_to_expression_type(symbol) if b != 0 goto return_1 b = unary_op_to_expression_type(symbol) if b != 0 goto return_1 goto return_0 function binop_expression_type_to_symbol argument exprtype local p p = .binop_table :binop_expr2symb_type_loop p += 1 if *1p == exprtype goto binop_expr2symb_type_found p += 1 if *1p != 0 goto binop_expr2symb_type_loop return 0 :binop_expr2symb_type_found p -= 1 return *1p function int_suffix_to_type argument suffix if suffix == NUMBER_SUFFIX_L goto return_type_long if suffix == NUMBER_SUFFIX_U goto return_type_unsigned_int if suffix == NUMBER_SUFFIX_UL goto return_type_unsigned_long goto return_type_int function float_suffix_to_type argument suffix if suffix == NUMBER_SUFFIX_F goto return_type_float goto return_type_double ; smallest integer type which can fit this value, only using unsigned if necessary function int_value_to_type argument value if value [ 0x80000000 goto return_type_int if value [ 0x8000000000000000 goto return_type_long goto return_type_unsigned_long ; returns pointer to end of expression function print_expression argument expression local c local b local p p = expression + 4 if *4p == 0 goto print_expr_skip_type putc(40) print_type(*4p) putc(41) :print_expr_skip_type c = *1expression if c == EXPRESSION_CONSTANT_INT goto print_expr_int if c == EXPRESSION_CONSTANT_FLOAT goto print_expr_float if c == EXPRESSION_STRING_LITERAL goto print_expr_str if c == EXPRESSION_POST_INCREMENT goto print_post_increment if c == EXPRESSION_POST_DECREMENT goto print_post_decrement if c == EXPRESSION_DOT goto print_expr_dot if c == EXPRESSION_ARROW goto print_expr_arrow if c == EXPRESSION_PRE_INCREMENT goto print_pre_increment if c == EXPRESSION_PRE_DECREMENT goto print_pre_decrement if c == EXPRESSION_ADDRESS_OF goto print_address_of if c == EXPRESSION_DEREFERENCE goto print_dereference if c == EXPRESSION_UNARY_PLUS goto print_unary_plus if c == EXPRESSION_UNARY_MINUS goto print_unary_minus if c == EXPRESSION_BITWISE_NOT goto print_bitwise_not if c == EXPRESSION_LOGICAL_NOT goto print_logical_not if c == EXPRESSION_CAST goto print_cast b = binop_expression_type_to_symbol(c) if b != 0 goto print_expr_binop puts(.str_print_bad_expr) exit(1) :str_print_bad_expr string Bad expression passed to print_expression. byte 10 byte 0 :print_cast ; we've already printed the type expression += 8 expression = print_expression(expression) return expression :print_expr_int expression += 8 putn_signed(*8expression) expression += 8 return expression :print_expr_float expression += 8 putx64(*8expression) expression += 8 return expression :print_expr_str expression += 8 putc('0) putc('x) putx32(*8expression) expression += 8 return expression :print_expr_binop putc(40) expression += 8 expression = print_expression(expression) ; 1st operand b = get_keyword_str(b) puts(b) expression = print_expression(expression) ; 2nd operand putc(41) return expression :print_expr_dot putc(40) expression += 8 expression = print_expression(expression) puts(.str_dot) putn(*8expression) expression += 8 putc(41) return expression :print_expr_arrow putc(40) expression += 8 expression = print_expression(expression) puts(.str_arrow) putn(*8expression) expression += 8 putc(41) return expression :print_post_increment putc(40) expression += 8 expression = print_expression(expression) putc('+) putc('+) putc(41) return expression :print_post_decrement putc(40) expression += 8 expression = print_expression(expression) putc('-) putc('-) putc(41) return expression :print_pre_increment putc(40) putc('+) putc('+) expression += 8 expression = print_expression(expression) putc(41) return expression :print_pre_decrement putc(40) putc('-) putc('-) expression += 8 expression = print_expression(expression) putc(41) return expression :print_address_of putc(40) putc('&) expression += 8 expression = print_expression(expression) putc(41) return expression :print_dereference putc(40) putc('*) expression += 8 expression = print_expression(expression) putc(41) return expression :print_unary_plus putc(40) putc('+) expression += 8 expression = print_expression(expression) putc(41) return expression :print_unary_minus putc(40) putc('-) expression += 8 expression = print_expression(expression) putc(41) return expression :print_bitwise_not putc(40) putc('~) expression += 8 expression = print_expression(expression) putc(41) return expression :print_logical_not putc(40) putc('!) expression += 8 expression = print_expression(expression) putc(41) return expression ; NOTE: to make things easier, the format which this outputs isn't the same as C's, specifically we have ; *int for pointer to int and [5]int for array of 5 ints function print_type argument type local c :print_type_top c = types + type c = *1c if c == TYPE_VOID goto print_type_void if c == TYPE_CHAR goto print_type_char if c == TYPE_UNSIGNED_CHAR goto print_type_unsigned_char if c == TYPE_SHORT goto print_type_short if c == TYPE_UNSIGNED_SHORT goto print_type_unsigned_short if c == TYPE_INT goto print_type_int if c == TYPE_UNSIGNED_INT goto print_type_unsigned_int if c == TYPE_LONG goto print_type_long if c == TYPE_UNSIGNED_LONG goto print_type_unsigned_long if c == TYPE_FLOAT goto print_type_float if c == TYPE_DOUBLE goto print_type_double if c == TYPE_POINTER goto print_type_pointer if c == TYPE_ARRAY goto print_type_array if c == TYPE_STRUCT goto print_type_struct if c == TYPE_FUNCTION goto print_type_function fputs(2, .str_bad_print_type) exit(1) :str_bad_print_type string Bad type passed to print_type. byte 10 byte 0 :print_type_void return puts(.str_void) :print_type_char return puts(.str_char) :print_type_unsigned_char return puts(.str_unsigned_char) :print_type_short return puts(.str_short) :print_type_unsigned_short return puts(.str_unsigned_short) :print_type_int return puts(.str_int) :print_type_unsigned_int return puts(.str_unsigned_int) :print_type_long return puts(.str_long) :print_type_unsigned_long return puts(.str_unsigned_long) :print_type_float return puts(.str_float) :print_type_double return puts(.str_double) :print_type_pointer putc('*) type += 1 goto print_type_top :print_type_array putc('[) type += 1 c = types + type putn(*8c) ; UNALIGNED putc(']) type += 8 goto print_type_top :print_type_struct return puts(.str_struct) :print_type_function type += 1 putc(40) putc(40) :print_type_function_loop c = types + type if *1c == 0 goto print_type_function_loop_end print_type(type) putc(44) type += type_length(type) goto print_type_function_loop :print_type_function_loop_end type += 1 ; 0 terminator putc(41) putc(32) putc('-) putc('>) putc(32) print_type(type) putc(41) return