tccasm: Don't ignore # in preprocessor directives

Our preprocessor throws away # line-comments in asm mode. It did so also inside preprocessor directives, thereby removing stringification. Parse defines in non-asm mode (but retain '.' as identifier character inside macro definitions).

Michael Matz committed 10 years ago

Commit dd57a341 parent e7ef0874 files · +50 −26Browse files

Unified · Split

4 files

tcc.h +6 −0
tccasm.c +2 −0
tccpp.c +24 −21
tests/asmtest.S +18 −5

tcc.h+6−0

@@ -1175,10 +1175,16 @@ ST_DATA TokenSym **table_ident; · 1174 unmodified lines
1175	1175	#define PARSE_FLAG_ACCEPT_STRAYS 0x0020 /* next() returns '\\' token */
1176	1176	#define PARSE_FLAG_TOK_STR 0x0040 /* return parsed strings instead of TOK_PPSTR */
1177	1177
	1178	+/* isidnum_table flags: */
	1179	+#define IS_SPC 1
	1180	+#define IS_ID 2
	1181	+#define IS_NUM 4
	1182	+
1178	1183	ST_FUNC TokenSym tok_alloc(const char str, int len);
1179	1184	ST_FUNC const char get_tok_str(int v, CValue cv);
1180	1185	ST_FUNC void begin_macro(TokenString *str, int alloc);
1181	1186	ST_FUNC void end_macro(void);
	1187	+ST_FUNC void set_idnum(int c, int val);
1182	1188	ST_FUNC void save_parse_state(ParseState *s);
1183	1189	ST_FUNC void restore_parse_state(ParseState *s);
1184	1190	ST_INLN void tok_str_new(TokenString *s);

tccasm.c+2−0

@@ -916,6 +916,7 @@ static int tcc_assemble_internal(TCCState *s1, int do_preprocess) · 915 unmodified lines
916	916	ch = file->buf_ptr[0];
917	917	tok_flags = TOK_FLAG_BOL \| TOK_FLAG_BOF;
918	918	parse_flags = PARSE_FLAG_ASM_FILE \| PARSE_FLAG_TOK_STR;
	919	+ set_idnum('.', IS_ID);
919	920	if (do_preprocess)
920	921	parse_flags \|= PARSE_FLAG_PREPROCESS;
921	922	next();
@@ -1036,6 +1037,7 @@ static void tcc_assemble_inline(TCCState s1, char str, int len) · 114 unmodified lines
1036	1037	tcc_close();
1037	1038
1038	1039	parse_flags = saved_parse_flags;
	1040	+ set_idnum('.', (parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0);
1039	1041	macro_ptr = saved_macro_ptr;
1040	1042	}
1041	1043

tccpp.c+24−21

@@ -53,11 +53,6 @@ static struct TinyAlloc *toksym_alloc; · 52 unmodified lines
53	53	static struct TinyAlloc *tokstr_alloc;
54	54	static struct TinyAlloc *cstr_alloc;
55	55
56		−/* isidnum_table flags: */
57		−#define IS_SPC 1
58		−#define IS_ID 2
59		−#define IS_NUM 4
60		−
61	56	static TokenString *macro_stack;
62	57
63	58	static const char tcc_keywords[] =
@@ -819,6 +814,11 @@ ST_FUNC uint8_t parse_comment(uint8_t p) · 755 unmodified lines
819	814	return p;
820	815	}
821	816
	817	+ST_FUNC void set_idnum(int c, int val)
	818	+{
	819	+ isidnum_table[c - CH_EOF] = val;
	820	+}
	821	+
822	822	#define cinp minp
823	823
824	824	static inline void skip_spaces(void)
@@ -1453,13 +1453,15 @@ ST_FUNC void parse_define(void) · 628 unmodified lines
1453	1453	/* XXX: should check if same macro (ANSI) */
1454	1454	first = NULL;
1455	1455	t = MACRO_OBJ;
	1456	+ /* We have to parse the whole define as if not in asm mode, in particular
	1457	+ no line comment with '#' must be ignored. Also for function
	1458	+ macros the argument list must be parsed without '.' being an ID
	1459	+ character. */
	1460	+ parse_flags = ((parse_flags & ~PARSE_FLAG_ASM_FILE) \| PARSE_FLAG_SPACES);
1456	1461	/* '(' must be just after macro definition for MACRO_FUNC */
1457		− parse_flags \|= PARSE_FLAG_SPACES;
1458	1462	next_nomacro_spc();
1459	1463	if (tok == '(') {
1460		− /* must be able to parse TOK_DOTS (in asm mode '.' can be part of identifier) */
1461		− parse_flags &= ~PARSE_FLAG_ASM_FILE;
1462		− isidnum_table['.' - CH_EOF] = 0;
	1464	+ set_idnum('.', 0);
1463	1465	next_nomacro();
1464	1466	ps = &first;
1465	1467	if (tok != ')') for (;;) {
@@ -1487,14 +1489,17 @@ ST_FUNC void parse_define(void) · 21 unmodified lines
1487	1489	}
1488	1490	next_nomacro_spc();
1489	1491	t = MACRO_FUNC;
1490		− parse_flags \|= (saved_parse_flags & PARSE_FLAG_ASM_FILE);
1491		− isidnum_table['.' - CH_EOF] =
1492		− (parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0;
1493	1492	}
1494	1493
1495	1494	tokstr_buf.len = 0;
1496	1495	spc = 2;
1497	1496	parse_flags \|= PARSE_FLAG_ACCEPT_STRAYS \| PARSE_FLAG_SPACES \| PARSE_FLAG_LINEFEED;
	1497	+ /* The body of a macro definition should be parsed such that identifiers
	1498	+ are parsed like the file mode determines (i.e. with '.' being an
	1499	+ ID character in asm mode). But '#' should be retained instead of
	1500	+ regarded as line comment leader, so still don't set ASM_FILE
	1501	+ in parse_flags. */
	1502	+ set_idnum('.', (saved_parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0);
1498	1503	while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1499	1504	/* remove spaces around ## and after '#' */
1500	1505	if (TOK_TWOSHARPS == tok) {
@@ -2673,7 +2678,7 @@ maybe_newline: · 1172 unmodified lines
2673	2678	if (isnum(c)) {
2674	2679	t = '.';
2675	2680	goto parse_num;
2676		− } else if ((parse_flags & PARSE_FLAG_ASM_FILE)
	2681	+ } else if ((isidnum_table['.' - CH_EOF] & IS_ID)
2677	2682	&& (isidnum_table[c - CH_EOF] & (IS_ID\|IS_NUM))) {
2678	2683	*--p = c = '.';
2679	2684	goto parse_ident_fast;
@@ -3470,10 +3475,8 @@ ST_FUNC void preprocess_start(TCCState *s1) · 790 unmodified lines
3470	3475	s1->pack_stack[0] = 0;
3471	3476	s1->pack_stack_ptr = s1->pack_stack;
3472	3477
3473		− isidnum_table['$' - CH_EOF] =
3474		− s1->dollars_in_identifiers ? IS_ID : 0;
3475		− isidnum_table['.' - CH_EOF] =
3476		− (parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0;
	3478	+ set_idnum('$', s1->dollars_in_identifiers ? IS_ID : 0);
	3479	+ set_idnum('.', (parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0);
3477	3480	buf = tcc_malloc(3 + strlen(file->filename));
3478	3481	sprintf(buf, "\"%s\"", file->filename);
3479	3482	tcc_undefine_symbol(s1, "__BASE_FILE__");
@@ -3505,14 +3508,14 @@ ST_FUNC void tccpp_new(TCCState *s) · 25 unmodified lines
3505	3508
3506	3509	/* init isid table */
3507	3510	for(i = CH_EOF; i<128; i++)
3508		− isidnum_table[i - CH_EOF]
3509		− = is_space(i) ? IS_SPC
	3511	+ set_idnum(i,
	3512	+ is_space(i) ? IS_SPC
3510	3513	: isid(i) ? IS_ID
3511	3514	: isnum(i) ? IS_NUM
3512		− : 0;
	3515	+ : 0);
3513	3516
3514	3517	for(i = 128; i<256; i++)
3515		− isidnum_table[i - CH_EOF] = IS_ID;
	3518	+ set_idnum(i, IS_ID);
3516	3519
3517	3520	/* init allocators */
3518	3521	tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE);

tests/asmtest.S+18−5

@@ -16,10 +16,17 @@ · 15 unmodified lines
16	16	.skip 3
17	17	.skip 15, 0x90
18	18	.string "hello\0world"
19		−
	19	+/* Macro expansion should work like with C, the #n shouldn't be parsed
	20	+ as asm line comment */
	21	+#define __stringify(n) #n
	22	+#define stringify(n) __stringify(n)
	23	+ .skip 8,0x90
	24	+ .asciz stringify(BLA)
	25	+ .skip 8,0x90
	26	+
	27	+# 28 "asmtest.S" # a line directive (and a line comment)
	28	+ movl %eax, %ebx # some more asm comment
20	29	/* some label tests */
21		−
22		− movl %eax, %ebx
23	30	L1:
24	31	movl %eax, %ebx
25	32	mov 0x10000, %eax
@@ -572,8 +579,8 @@ int $0x10 · 546 unmodified lines
572	579	pusha
573	580	popa
574	581	#endif
575		− clc
576		− cld
	582	+ clc # another comment
	583	+ cld # a comment with embedded ' tick
577	584	cli
578	585	clts
579	586	cmc
@@ -676,8 +683,10 @@ int $0x10 · 96 unmodified lines
676	683	lar %eax,%dx
677	684	lar %ax,%edx
678	685	lar %eax,%edx
	686	+#ifdef __x86_64__
679	687	lar %ax,%rdx
680	688	lar %eax,%rdx
	689	+#endif
681	690	emms
682	691	movd %edx, %mm3
683	692	movd 0x1000, %mm2
@@ -820,7 +829,11 @@ nop · 136 unmodified lines
820	829	.long 145 + 2b
821	830	.word 164, 0
822	831	.org 2b+32
	832	+#ifdef __x86_64__
823	833	.quad 1b
	834	+#else
	835	+ .long 1b
	836	+#endif
824	837	.popsection
825	838	3: mov %eax,%ecx
826	839	4:

tccasm: Don't ignore # in preprocessor directives

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code