[GAS][ARM] Fix ARMv8.1 AdvSIMD testism
[binutils-gdb.git] / gdb / go-exp.y
blob7521acc7f52b546f8202b26d09e7ba90a1be1730
1 /* YACC parser for Go expressions, for GDB.
3 Copyright (C) 2012-2018 Free Software Foundation, Inc.
5 This file is part of GDB.
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>. */
20 /* This file is derived from c-exp.y, p-exp.y. */
22 /* Parse a Go expression from text in a string,
23 and return the result as a struct expression pointer.
24 That structure contains arithmetic operations in reverse polish,
25 with constants represented by operations that are followed by special data.
26 See expression.h for the details of the format.
27 What is important here is that it can be built up sequentially
28 during the process of parsing; the lower levels of the tree always
29 come first in the result.
31 Note that malloc's and realloc's in this file are transformed to
32 xmalloc and xrealloc respectively by the same sed command in the
33 makefile that remaps any other malloc/realloc inserted by the parser
34 generator. Doing this with #defines and trying to control the interaction
35 with include files (<malloc.h> and <stdlib.h> for example) just became
36 too messy, particularly when such includes can be inserted at random
37 times by the parser generator. */
39 /* Known bugs or limitations:
41 - Unicode
42 - &^
43 - '_' (blank identifier)
44 - automatic deref of pointers
45 - method expressions
46 - interfaces, channels, etc.
48 And lots of other things.
49 I'm sure there's some cleanup to do.
54 #include "defs.h"
55 #include <ctype.h>
56 #include "expression.h"
57 #include "value.h"
58 #include "parser-defs.h"
59 #include "language.h"
60 #include "c-lang.h"
61 #include "go-lang.h"
62 #include "bfd.h" /* Required by objfiles.h. */
63 #include "symfile.h" /* Required by objfiles.h. */
64 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
65 #include "charset.h"
66 #include "block.h"
68 #define parse_type(ps) builtin_type (parse_gdbarch (ps))
70 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror,
71 etc). */
72 #define GDB_YY_REMAP_PREFIX go_
73 #include "yy-remap.h"
75 /* The state of the parser, used internally when we are parsing the
76 expression. */
78 static struct parser_state *pstate = NULL;
80 int yyparse (void);
82 static int yylex (void);
84 static void yyerror (const char *);
88 /* Although the yacc "value" of an expression is not used,
89 since the result is stored in the structure being created,
90 other node types do have values. */
92 %union
94 LONGEST lval;
95 struct {
96 LONGEST val;
97 struct type *type;
98 } typed_val_int;
99 struct {
100 gdb_byte val[16];
101 struct type *type;
102 } typed_val_float;
103 struct stoken sval;
104 struct symtoken ssym;
105 struct type *tval;
106 struct typed_stoken tsval;
107 struct ttype tsym;
108 int voidval;
109 enum exp_opcode opcode;
110 struct internalvar *ivar;
111 struct stoken_vector svec;
115 /* YYSTYPE gets defined by %union. */
116 static int parse_number (struct parser_state *,
117 const char *, int, int, YYSTYPE *);
120 %type <voidval> exp exp1 type_exp start variable lcurly
121 %type <lval> rcurly
122 %type <tval> type
124 %token <typed_val_int> INT
125 %token <typed_val_float> FLOAT
127 /* Both NAME and TYPENAME tokens represent symbols in the input,
128 and both convey their data as strings.
129 But a TYPENAME is a string that happens to be defined as a type
130 or builtin type name (such as int or char)
131 and a NAME is any other symbol.
132 Contexts where this distinction is not important can use the
133 nonterminal "name", which matches either NAME or TYPENAME. */
135 %token <tsval> RAW_STRING
136 %token <tsval> STRING
137 %token <tsval> CHAR
138 %token <ssym> NAME
139 %token <tsym> TYPENAME /* Not TYPE_NAME cus already taken. */
140 %token <voidval> COMPLETE
141 /*%type <sval> name*/
142 %type <svec> string_exp
143 %type <ssym> name_not_typename
145 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
146 but which would parse as a valid number in the current input radix.
147 E.g. "c" when input_radix==16. Depending on the parse, it will be
148 turned into a name or into a number. */
149 %token <ssym> NAME_OR_INT
151 %token <lval> TRUE_KEYWORD FALSE_KEYWORD
152 %token STRUCT_KEYWORD INTERFACE_KEYWORD TYPE_KEYWORD CHAN_KEYWORD
153 %token SIZEOF_KEYWORD
154 %token LEN_KEYWORD CAP_KEYWORD
155 %token NEW_KEYWORD
156 %token IOTA_KEYWORD NIL_KEYWORD
157 %token CONST_KEYWORD
158 %token DOTDOTDOT
159 %token ENTRY
160 %token ERROR
162 /* Special type cases. */
163 %token BYTE_KEYWORD /* An alias of uint8. */
165 %token <sval> DOLLAR_VARIABLE
167 %token <opcode> ASSIGN_MODIFY
169 %left ','
170 %left ABOVE_COMMA
171 %right '=' ASSIGN_MODIFY
172 %right '?'
173 %left OROR
174 %left ANDAND
175 %left '|'
176 %left '^'
177 %left '&'
178 %left ANDNOT
179 %left EQUAL NOTEQUAL
180 %left '<' '>' LEQ GEQ
181 %left LSH RSH
182 %left '@'
183 %left '+' '-'
184 %left '*' '/' '%'
185 %right UNARY INCREMENT DECREMENT
186 %right LEFT_ARROW '.' '[' '('
191 start : exp1
192 | type_exp
195 type_exp: type
196 { write_exp_elt_opcode (pstate, OP_TYPE);
197 write_exp_elt_type (pstate, $1);
198 write_exp_elt_opcode (pstate, OP_TYPE); }
201 /* Expressions, including the comma operator. */
202 exp1 : exp
203 | exp1 ',' exp
204 { write_exp_elt_opcode (pstate, BINOP_COMMA); }
207 /* Expressions, not including the comma operator. */
208 exp : '*' exp %prec UNARY
209 { write_exp_elt_opcode (pstate, UNOP_IND); }
212 exp : '&' exp %prec UNARY
213 { write_exp_elt_opcode (pstate, UNOP_ADDR); }
216 exp : '-' exp %prec UNARY
217 { write_exp_elt_opcode (pstate, UNOP_NEG); }
220 exp : '+' exp %prec UNARY
221 { write_exp_elt_opcode (pstate, UNOP_PLUS); }
224 exp : '!' exp %prec UNARY
225 { write_exp_elt_opcode (pstate, UNOP_LOGICAL_NOT); }
228 exp : '^' exp %prec UNARY
229 { write_exp_elt_opcode (pstate, UNOP_COMPLEMENT); }
232 exp : exp INCREMENT %prec UNARY
233 { write_exp_elt_opcode (pstate, UNOP_POSTINCREMENT); }
236 exp : exp DECREMENT %prec UNARY
237 { write_exp_elt_opcode (pstate, UNOP_POSTDECREMENT); }
240 /* foo->bar is not in Go. May want as a gdb extension. Later. */
242 exp : exp '.' name_not_typename
243 { write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
244 write_exp_string (pstate, $3.stoken);
245 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
248 exp : exp '.' name_not_typename COMPLETE
249 { mark_struct_expression (pstate);
250 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
251 write_exp_string (pstate, $3.stoken);
252 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
255 exp : exp '.' COMPLETE
256 { struct stoken s;
257 mark_struct_expression (pstate);
258 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
259 s.ptr = "";
260 s.length = 0;
261 write_exp_string (pstate, s);
262 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
265 exp : exp '[' exp1 ']'
266 { write_exp_elt_opcode (pstate, BINOP_SUBSCRIPT); }
269 exp : exp '('
270 /* This is to save the value of arglist_len
271 being accumulated by an outer function call. */
272 { start_arglist (); }
273 arglist ')' %prec LEFT_ARROW
274 { write_exp_elt_opcode (pstate, OP_FUNCALL);
275 write_exp_elt_longcst (pstate,
276 (LONGEST) end_arglist ());
277 write_exp_elt_opcode (pstate, OP_FUNCALL); }
280 lcurly : '{'
281 { start_arglist (); }
284 arglist :
287 arglist : exp
288 { arglist_len = 1; }
291 arglist : arglist ',' exp %prec ABOVE_COMMA
292 { arglist_len++; }
295 rcurly : '}'
296 { $$ = end_arglist () - 1; }
299 exp : lcurly type rcurly exp %prec UNARY
300 { write_exp_elt_opcode (pstate, UNOP_MEMVAL);
301 write_exp_elt_type (pstate, $2);
302 write_exp_elt_opcode (pstate, UNOP_MEMVAL); }
305 exp : type '(' exp ')' %prec UNARY
306 { write_exp_elt_opcode (pstate, UNOP_CAST);
307 write_exp_elt_type (pstate, $1);
308 write_exp_elt_opcode (pstate, UNOP_CAST); }
311 exp : '(' exp1 ')'
315 /* Binary operators in order of decreasing precedence. */
317 exp : exp '@' exp
318 { write_exp_elt_opcode (pstate, BINOP_REPEAT); }
321 exp : exp '*' exp
322 { write_exp_elt_opcode (pstate, BINOP_MUL); }
325 exp : exp '/' exp
326 { write_exp_elt_opcode (pstate, BINOP_DIV); }
329 exp : exp '%' exp
330 { write_exp_elt_opcode (pstate, BINOP_REM); }
333 exp : exp '+' exp
334 { write_exp_elt_opcode (pstate, BINOP_ADD); }
337 exp : exp '-' exp
338 { write_exp_elt_opcode (pstate, BINOP_SUB); }
341 exp : exp LSH exp
342 { write_exp_elt_opcode (pstate, BINOP_LSH); }
345 exp : exp RSH exp
346 { write_exp_elt_opcode (pstate, BINOP_RSH); }
349 exp : exp EQUAL exp
350 { write_exp_elt_opcode (pstate, BINOP_EQUAL); }
353 exp : exp NOTEQUAL exp
354 { write_exp_elt_opcode (pstate, BINOP_NOTEQUAL); }
357 exp : exp LEQ exp
358 { write_exp_elt_opcode (pstate, BINOP_LEQ); }
361 exp : exp GEQ exp
362 { write_exp_elt_opcode (pstate, BINOP_GEQ); }
365 exp : exp '<' exp
366 { write_exp_elt_opcode (pstate, BINOP_LESS); }
369 exp : exp '>' exp
370 { write_exp_elt_opcode (pstate, BINOP_GTR); }
373 exp : exp '&' exp
374 { write_exp_elt_opcode (pstate, BINOP_BITWISE_AND); }
377 exp : exp '^' exp
378 { write_exp_elt_opcode (pstate, BINOP_BITWISE_XOR); }
381 exp : exp '|' exp
382 { write_exp_elt_opcode (pstate, BINOP_BITWISE_IOR); }
385 exp : exp ANDAND exp
386 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_AND); }
389 exp : exp OROR exp
390 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_OR); }
393 exp : exp '?' exp ':' exp %prec '?'
394 { write_exp_elt_opcode (pstate, TERNOP_COND); }
397 exp : exp '=' exp
398 { write_exp_elt_opcode (pstate, BINOP_ASSIGN); }
401 exp : exp ASSIGN_MODIFY exp
402 { write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY);
403 write_exp_elt_opcode (pstate, $2);
404 write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY); }
407 exp : INT
408 { write_exp_elt_opcode (pstate, OP_LONG);
409 write_exp_elt_type (pstate, $1.type);
410 write_exp_elt_longcst (pstate, (LONGEST)($1.val));
411 write_exp_elt_opcode (pstate, OP_LONG); }
414 exp : CHAR
416 struct stoken_vector vec;
417 vec.len = 1;
418 vec.tokens = &$1;
419 write_exp_string_vector (pstate, $1.type, &vec);
423 exp : NAME_OR_INT
424 { YYSTYPE val;
425 parse_number (pstate, $1.stoken.ptr,
426 $1.stoken.length, 0, &val);
427 write_exp_elt_opcode (pstate, OP_LONG);
428 write_exp_elt_type (pstate, val.typed_val_int.type);
429 write_exp_elt_longcst (pstate, (LONGEST)
430 val.typed_val_int.val);
431 write_exp_elt_opcode (pstate, OP_LONG);
436 exp : FLOAT
437 { write_exp_elt_opcode (pstate, OP_FLOAT);
438 write_exp_elt_type (pstate, $1.type);
439 write_exp_elt_floatcst (pstate, $1.val);
440 write_exp_elt_opcode (pstate, OP_FLOAT); }
443 exp : variable
446 exp : DOLLAR_VARIABLE
448 write_dollar_variable (pstate, $1);
452 exp : SIZEOF_KEYWORD '(' type ')' %prec UNARY
454 /* TODO(dje): Go objects in structs. */
455 write_exp_elt_opcode (pstate, OP_LONG);
456 /* TODO(dje): What's the right type here? */
457 write_exp_elt_type
458 (pstate,
459 parse_type (pstate)->builtin_unsigned_int);
460 $3 = check_typedef ($3);
461 write_exp_elt_longcst (pstate,
462 (LONGEST) TYPE_LENGTH ($3));
463 write_exp_elt_opcode (pstate, OP_LONG);
467 exp : SIZEOF_KEYWORD '(' exp ')' %prec UNARY
469 /* TODO(dje): Go objects in structs. */
470 write_exp_elt_opcode (pstate, UNOP_SIZEOF);
473 string_exp:
474 STRING
476 /* We copy the string here, and not in the
477 lexer, to guarantee that we do not leak a
478 string. */
479 /* Note that we NUL-terminate here, but just
480 for convenience. */
481 struct typed_stoken *vec = XNEW (struct typed_stoken);
482 $$.len = 1;
483 $$.tokens = vec;
485 vec->type = $1.type;
486 vec->length = $1.length;
487 vec->ptr = (char *) malloc ($1.length + 1);
488 memcpy (vec->ptr, $1.ptr, $1.length + 1);
491 | string_exp '+' STRING
493 /* Note that we NUL-terminate here, but just
494 for convenience. */
495 char *p;
496 ++$$.len;
497 $$.tokens = XRESIZEVEC (struct typed_stoken,
498 $$.tokens, $$.len);
500 p = (char *) malloc ($3.length + 1);
501 memcpy (p, $3.ptr, $3.length + 1);
503 $$.tokens[$$.len - 1].type = $3.type;
504 $$.tokens[$$.len - 1].length = $3.length;
505 $$.tokens[$$.len - 1].ptr = p;
509 exp : string_exp %prec ABOVE_COMMA
511 int i;
513 write_exp_string_vector (pstate, 0 /*always utf8*/,
514 &$1);
515 for (i = 0; i < $1.len; ++i)
516 free ($1.tokens[i].ptr);
517 free ($1.tokens);
521 exp : TRUE_KEYWORD
522 { write_exp_elt_opcode (pstate, OP_BOOL);
523 write_exp_elt_longcst (pstate, (LONGEST) $1);
524 write_exp_elt_opcode (pstate, OP_BOOL); }
527 exp : FALSE_KEYWORD
528 { write_exp_elt_opcode (pstate, OP_BOOL);
529 write_exp_elt_longcst (pstate, (LONGEST) $1);
530 write_exp_elt_opcode (pstate, OP_BOOL); }
533 variable: name_not_typename ENTRY
534 { struct symbol *sym = $1.sym.symbol;
536 if (sym == NULL
537 || !SYMBOL_IS_ARGUMENT (sym)
538 || !symbol_read_needs_frame (sym))
539 error (_("@entry can be used only for function "
540 "parameters, not for \"%s\""),
541 copy_name ($1.stoken));
543 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE);
544 write_exp_elt_sym (pstate, sym);
545 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE);
549 variable: name_not_typename
550 { struct block_symbol sym = $1.sym;
552 if (sym.symbol)
554 if (symbol_read_needs_frame (sym.symbol))
555 innermost_block.update (sym);
557 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
558 write_exp_elt_block (pstate, sym.block);
559 write_exp_elt_sym (pstate, sym.symbol);
560 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
562 else if ($1.is_a_field_of_this)
564 /* TODO(dje): Can we get here?
565 E.g., via a mix of c++ and go? */
566 gdb_assert_not_reached ("go with `this' field");
568 else
570 struct bound_minimal_symbol msymbol;
571 char *arg = copy_name ($1.stoken);
573 msymbol =
574 lookup_bound_minimal_symbol (arg);
575 if (msymbol.minsym != NULL)
576 write_exp_msymbol (pstate, msymbol);
577 else if (!have_full_symbols ()
578 && !have_partial_symbols ())
579 error (_("No symbol table is loaded. "
580 "Use the \"file\" command."));
581 else
582 error (_("No symbol \"%s\" in current context."),
583 copy_name ($1.stoken));
588 /* TODO
589 method_exp: PACKAGENAME '.' name '.' name
595 type /* Implements (approximately): [*] type-specifier */
596 : '*' type
597 { $$ = lookup_pointer_type ($2); }
598 | TYPENAME
599 { $$ = $1.type; }
601 | STRUCT_KEYWORD name
602 { $$ = lookup_struct (copy_name ($2),
603 expression_context_block); }
605 | BYTE_KEYWORD
606 { $$ = builtin_go_type (parse_gdbarch (pstate))
607 ->builtin_uint8; }
610 /* TODO
611 name : NAME { $$ = $1.stoken; }
612 | TYPENAME { $$ = $1.stoken; }
613 | NAME_OR_INT { $$ = $1.stoken; }
617 name_not_typename
618 : NAME
619 /* These would be useful if name_not_typename was useful, but it is just
620 a fake for "variable", so these cause reduce/reduce conflicts because
621 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
622 =exp) or just an exp. If name_not_typename was ever used in an lvalue
623 context where only a name could occur, this might be useful.
624 | NAME_OR_INT
630 /* Take care of parsing a number (anything that starts with a digit).
631 Set yylval and return the token type; update lexptr.
632 LEN is the number of characters in it. */
634 /* FIXME: Needs some error checking for the float case. */
635 /* FIXME(dje): IWBN to use c-exp.y's parse_number if we could.
636 That will require moving the guts into a function that we both call
637 as our YYSTYPE is different than c-exp.y's */
639 static int
640 parse_number (struct parser_state *par_state,
641 const char *p, int len, int parsed_float, YYSTYPE *putithere)
643 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
644 here, and we do kind of silly things like cast to unsigned. */
645 LONGEST n = 0;
646 LONGEST prevn = 0;
647 ULONGEST un;
649 int i = 0;
650 int c;
651 int base = input_radix;
652 int unsigned_p = 0;
654 /* Number of "L" suffixes encountered. */
655 int long_p = 0;
657 /* We have found a "L" or "U" suffix. */
658 int found_suffix = 0;
660 ULONGEST high_bit;
661 struct type *signed_type;
662 struct type *unsigned_type;
664 if (parsed_float)
666 const struct builtin_go_type *builtin_go_types
667 = builtin_go_type (parse_gdbarch (par_state));
669 /* Handle suffixes: 'f' for float32, 'l' for long double.
670 FIXME: This appears to be an extension -- do we want this? */
671 if (len >= 1 && tolower (p[len - 1]) == 'f')
673 putithere->typed_val_float.type
674 = builtin_go_types->builtin_float32;
675 len--;
677 else if (len >= 1 && tolower (p[len - 1]) == 'l')
679 putithere->typed_val_float.type
680 = parse_type (par_state)->builtin_long_double;
681 len--;
683 /* Default type for floating-point literals is float64. */
684 else
686 putithere->typed_val_float.type
687 = builtin_go_types->builtin_float64;
690 if (!parse_float (p, len,
691 putithere->typed_val_float.type,
692 putithere->typed_val_float.val))
693 return ERROR;
694 return FLOAT;
697 /* Handle base-switching prefixes 0x, 0t, 0d, 0. */
698 if (p[0] == '0')
699 switch (p[1])
701 case 'x':
702 case 'X':
703 if (len >= 3)
705 p += 2;
706 base = 16;
707 len -= 2;
709 break;
711 case 'b':
712 case 'B':
713 if (len >= 3)
715 p += 2;
716 base = 2;
717 len -= 2;
719 break;
721 case 't':
722 case 'T':
723 case 'd':
724 case 'D':
725 if (len >= 3)
727 p += 2;
728 base = 10;
729 len -= 2;
731 break;
733 default:
734 base = 8;
735 break;
738 while (len-- > 0)
740 c = *p++;
741 if (c >= 'A' && c <= 'Z')
742 c += 'a' - 'A';
743 if (c != 'l' && c != 'u')
744 n *= base;
745 if (c >= '0' && c <= '9')
747 if (found_suffix)
748 return ERROR;
749 n += i = c - '0';
751 else
753 if (base > 10 && c >= 'a' && c <= 'f')
755 if (found_suffix)
756 return ERROR;
757 n += i = c - 'a' + 10;
759 else if (c == 'l')
761 ++long_p;
762 found_suffix = 1;
764 else if (c == 'u')
766 unsigned_p = 1;
767 found_suffix = 1;
769 else
770 return ERROR; /* Char not a digit */
772 if (i >= base)
773 return ERROR; /* Invalid digit in this base. */
775 /* Portably test for overflow (only works for nonzero values, so make
776 a second check for zero). FIXME: Can't we just make n and prevn
777 unsigned and avoid this? */
778 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
779 unsigned_p = 1; /* Try something unsigned. */
781 /* Portably test for unsigned overflow.
782 FIXME: This check is wrong; for example it doesn't find overflow
783 on 0x123456789 when LONGEST is 32 bits. */
784 if (c != 'l' && c != 'u' && n != 0)
786 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
787 error (_("Numeric constant too large."));
789 prevn = n;
792 /* An integer constant is an int, a long, or a long long. An L
793 suffix forces it to be long; an LL suffix forces it to be long
794 long. If not forced to a larger size, it gets the first type of
795 the above that it fits in. To figure out whether it fits, we
796 shift it right and see whether anything remains. Note that we
797 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
798 operation, because many compilers will warn about such a shift
799 (which always produces a zero result). Sometimes gdbarch_int_bit
800 or gdbarch_long_bit will be that big, sometimes not. To deal with
801 the case where it is we just always shift the value more than
802 once, with fewer bits each time. */
804 un = (ULONGEST)n >> 2;
805 if (long_p == 0
806 && (un >> (gdbarch_int_bit (parse_gdbarch (par_state)) - 2)) == 0)
808 high_bit
809 = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch (par_state)) - 1);
811 /* A large decimal (not hex or octal) constant (between INT_MAX
812 and UINT_MAX) is a long or unsigned long, according to ANSI,
813 never an unsigned int, but this code treats it as unsigned
814 int. This probably should be fixed. GCC gives a warning on
815 such constants. */
817 unsigned_type = parse_type (par_state)->builtin_unsigned_int;
818 signed_type = parse_type (par_state)->builtin_int;
820 else if (long_p <= 1
821 && (un >> (gdbarch_long_bit (parse_gdbarch (par_state)) - 2)) == 0)
823 high_bit
824 = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch (par_state)) - 1);
825 unsigned_type = parse_type (par_state)->builtin_unsigned_long;
826 signed_type = parse_type (par_state)->builtin_long;
828 else
830 int shift;
831 if (sizeof (ULONGEST) * HOST_CHAR_BIT
832 < gdbarch_long_long_bit (parse_gdbarch (par_state)))
833 /* A long long does not fit in a LONGEST. */
834 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
835 else
836 shift = (gdbarch_long_long_bit (parse_gdbarch (par_state)) - 1);
837 high_bit = (ULONGEST) 1 << shift;
838 unsigned_type = parse_type (par_state)->builtin_unsigned_long_long;
839 signed_type = parse_type (par_state)->builtin_long_long;
842 putithere->typed_val_int.val = n;
844 /* If the high bit of the worked out type is set then this number
845 has to be unsigned. */
847 if (unsigned_p || (n & high_bit))
849 putithere->typed_val_int.type = unsigned_type;
851 else
853 putithere->typed_val_int.type = signed_type;
856 return INT;
859 /* Temporary obstack used for holding strings. */
860 static struct obstack tempbuf;
861 static int tempbuf_init;
863 /* Parse a string or character literal from TOKPTR. The string or
864 character may be wide or unicode. *OUTPTR is set to just after the
865 end of the literal in the input string. The resulting token is
866 stored in VALUE. This returns a token value, either STRING or
867 CHAR, depending on what was parsed. *HOST_CHARS is set to the
868 number of host characters in the literal. */
870 static int
871 parse_string_or_char (const char *tokptr, const char **outptr,
872 struct typed_stoken *value, int *host_chars)
874 int quote;
876 /* Build the gdb internal form of the input string in tempbuf. Note
877 that the buffer is null byte terminated *only* for the
878 convenience of debugging gdb itself and printing the buffer
879 contents when the buffer contains no embedded nulls. Gdb does
880 not depend upon the buffer being null byte terminated, it uses
881 the length string instead. This allows gdb to handle C strings
882 (as well as strings in other languages) with embedded null
883 bytes */
885 if (!tempbuf_init)
886 tempbuf_init = 1;
887 else
888 obstack_free (&tempbuf, NULL);
889 obstack_init (&tempbuf);
891 /* Skip the quote. */
892 quote = *tokptr;
893 ++tokptr;
895 *host_chars = 0;
897 while (*tokptr)
899 char c = *tokptr;
900 if (c == '\\')
902 ++tokptr;
903 *host_chars += c_parse_escape (&tokptr, &tempbuf);
905 else if (c == quote)
906 break;
907 else
909 obstack_1grow (&tempbuf, c);
910 ++tokptr;
911 /* FIXME: this does the wrong thing with multi-byte host
912 characters. We could use mbrlen here, but that would
913 make "set host-charset" a bit less useful. */
914 ++*host_chars;
918 if (*tokptr != quote)
920 if (quote == '"')
921 error (_("Unterminated string in expression."));
922 else
923 error (_("Unmatched single quote."));
925 ++tokptr;
927 value->type = C_STRING | (quote == '\'' ? C_CHAR : 0); /*FIXME*/
928 value->ptr = (char *) obstack_base (&tempbuf);
929 value->length = obstack_object_size (&tempbuf);
931 *outptr = tokptr;
933 return quote == '\'' ? CHAR : STRING;
936 struct token
938 const char *oper;
939 int token;
940 enum exp_opcode opcode;
943 static const struct token tokentab3[] =
945 {">>=", ASSIGN_MODIFY, BINOP_RSH},
946 {"<<=", ASSIGN_MODIFY, BINOP_LSH},
947 /*{"&^=", ASSIGN_MODIFY, BINOP_BITWISE_ANDNOT}, TODO */
948 {"...", DOTDOTDOT, OP_NULL},
951 static const struct token tokentab2[] =
953 {"+=", ASSIGN_MODIFY, BINOP_ADD},
954 {"-=", ASSIGN_MODIFY, BINOP_SUB},
955 {"*=", ASSIGN_MODIFY, BINOP_MUL},
956 {"/=", ASSIGN_MODIFY, BINOP_DIV},
957 {"%=", ASSIGN_MODIFY, BINOP_REM},
958 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR},
959 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND},
960 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR},
961 {"++", INCREMENT, BINOP_END},
962 {"--", DECREMENT, BINOP_END},
963 /*{"->", RIGHT_ARROW, BINOP_END}, Doesn't exist in Go. */
964 {"<-", LEFT_ARROW, BINOP_END},
965 {"&&", ANDAND, BINOP_END},
966 {"||", OROR, BINOP_END},
967 {"<<", LSH, BINOP_END},
968 {">>", RSH, BINOP_END},
969 {"==", EQUAL, BINOP_END},
970 {"!=", NOTEQUAL, BINOP_END},
971 {"<=", LEQ, BINOP_END},
972 {">=", GEQ, BINOP_END},
973 /*{"&^", ANDNOT, BINOP_END}, TODO */
976 /* Identifier-like tokens. */
977 static const struct token ident_tokens[] =
979 {"true", TRUE_KEYWORD, OP_NULL},
980 {"false", FALSE_KEYWORD, OP_NULL},
981 {"nil", NIL_KEYWORD, OP_NULL},
982 {"const", CONST_KEYWORD, OP_NULL},
983 {"struct", STRUCT_KEYWORD, OP_NULL},
984 {"type", TYPE_KEYWORD, OP_NULL},
985 {"interface", INTERFACE_KEYWORD, OP_NULL},
986 {"chan", CHAN_KEYWORD, OP_NULL},
987 {"byte", BYTE_KEYWORD, OP_NULL}, /* An alias of uint8. */
988 {"len", LEN_KEYWORD, OP_NULL},
989 {"cap", CAP_KEYWORD, OP_NULL},
990 {"new", NEW_KEYWORD, OP_NULL},
991 {"iota", IOTA_KEYWORD, OP_NULL},
994 /* This is set if a NAME token appeared at the very end of the input
995 string, with no whitespace separating the name from the EOF. This
996 is used only when parsing to do field name completion. */
997 static int saw_name_at_eof;
999 /* This is set if the previously-returned token was a structure
1000 operator -- either '.' or ARROW. This is used only when parsing to
1001 do field name completion. */
1002 static int last_was_structop;
1004 /* Read one token, getting characters through lexptr. */
1006 static int
1007 lex_one_token (struct parser_state *par_state)
1009 int c;
1010 int namelen;
1011 unsigned int i;
1012 const char *tokstart;
1013 int saw_structop = last_was_structop;
1014 char *copy;
1016 last_was_structop = 0;
1018 retry:
1020 prev_lexptr = lexptr;
1022 tokstart = lexptr;
1023 /* See if it is a special token of length 3. */
1024 for (i = 0; i < sizeof (tokentab3) / sizeof (tokentab3[0]); i++)
1025 if (strncmp (tokstart, tokentab3[i].oper, 3) == 0)
1027 lexptr += 3;
1028 yylval.opcode = tokentab3[i].opcode;
1029 return tokentab3[i].token;
1032 /* See if it is a special token of length 2. */
1033 for (i = 0; i < sizeof (tokentab2) / sizeof (tokentab2[0]); i++)
1034 if (strncmp (tokstart, tokentab2[i].oper, 2) == 0)
1036 lexptr += 2;
1037 yylval.opcode = tokentab2[i].opcode;
1038 /* NOTE: -> doesn't exist in Go, so we don't need to watch for
1039 setting last_was_structop here. */
1040 return tokentab2[i].token;
1043 switch (c = *tokstart)
1045 case 0:
1046 if (saw_name_at_eof)
1048 saw_name_at_eof = 0;
1049 return COMPLETE;
1051 else if (saw_structop)
1052 return COMPLETE;
1053 else
1054 return 0;
1056 case ' ':
1057 case '\t':
1058 case '\n':
1059 lexptr++;
1060 goto retry;
1062 case '[':
1063 case '(':
1064 paren_depth++;
1065 lexptr++;
1066 return c;
1068 case ']':
1069 case ')':
1070 if (paren_depth == 0)
1071 return 0;
1072 paren_depth--;
1073 lexptr++;
1074 return c;
1076 case ',':
1077 if (comma_terminates
1078 && paren_depth == 0)
1079 return 0;
1080 lexptr++;
1081 return c;
1083 case '.':
1084 /* Might be a floating point number. */
1085 if (lexptr[1] < '0' || lexptr[1] > '9')
1087 if (parse_completion)
1088 last_was_structop = 1;
1089 goto symbol; /* Nope, must be a symbol. */
1091 /* FALL THRU. */
1093 case '0':
1094 case '1':
1095 case '2':
1096 case '3':
1097 case '4':
1098 case '5':
1099 case '6':
1100 case '7':
1101 case '8':
1102 case '9':
1104 /* It's a number. */
1105 int got_dot = 0, got_e = 0, toktype;
1106 const char *p = tokstart;
1107 int hex = input_radix > 10;
1109 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1111 p += 2;
1112 hex = 1;
1115 for (;; ++p)
1117 /* This test includes !hex because 'e' is a valid hex digit
1118 and thus does not indicate a floating point number when
1119 the radix is hex. */
1120 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1121 got_dot = got_e = 1;
1122 /* This test does not include !hex, because a '.' always indicates
1123 a decimal floating point number regardless of the radix. */
1124 else if (!got_dot && *p == '.')
1125 got_dot = 1;
1126 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
1127 && (*p == '-' || *p == '+'))
1128 /* This is the sign of the exponent, not the end of the
1129 number. */
1130 continue;
1131 /* We will take any letters or digits. parse_number will
1132 complain if past the radix, or if L or U are not final. */
1133 else if ((*p < '0' || *p > '9')
1134 && ((*p < 'a' || *p > 'z')
1135 && (*p < 'A' || *p > 'Z')))
1136 break;
1138 toktype = parse_number (par_state, tokstart, p - tokstart,
1139 got_dot|got_e, &yylval);
1140 if (toktype == ERROR)
1142 char *err_copy = (char *) alloca (p - tokstart + 1);
1144 memcpy (err_copy, tokstart, p - tokstart);
1145 err_copy[p - tokstart] = 0;
1146 error (_("Invalid number \"%s\"."), err_copy);
1148 lexptr = p;
1149 return toktype;
1152 case '@':
1154 const char *p = &tokstart[1];
1155 size_t len = strlen ("entry");
1157 while (isspace (*p))
1158 p++;
1159 if (strncmp (p, "entry", len) == 0 && !isalnum (p[len])
1160 && p[len] != '_')
1162 lexptr = &p[len];
1163 return ENTRY;
1166 /* FALLTHRU */
1167 case '+':
1168 case '-':
1169 case '*':
1170 case '/':
1171 case '%':
1172 case '|':
1173 case '&':
1174 case '^':
1175 case '~':
1176 case '!':
1177 case '<':
1178 case '>':
1179 case '?':
1180 case ':':
1181 case '=':
1182 case '{':
1183 case '}':
1184 symbol:
1185 lexptr++;
1186 return c;
1188 case '\'':
1189 case '"':
1190 case '`':
1192 int host_len;
1193 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval,
1194 &host_len);
1195 if (result == CHAR)
1197 if (host_len == 0)
1198 error (_("Empty character constant."));
1199 else if (host_len > 2 && c == '\'')
1201 ++tokstart;
1202 namelen = lexptr - tokstart - 1;
1203 goto tryname;
1205 else if (host_len > 1)
1206 error (_("Invalid character constant."));
1208 return result;
1212 if (!(c == '_' || c == '$'
1213 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1214 /* We must have come across a bad character (e.g. ';'). */
1215 error (_("Invalid character '%c' in expression."), c);
1217 /* It's a name. See how long it is. */
1218 namelen = 0;
1219 for (c = tokstart[namelen];
1220 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1221 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));)
1223 c = tokstart[++namelen];
1226 /* The token "if" terminates the expression and is NOT removed from
1227 the input stream. It doesn't count if it appears in the
1228 expansion of a macro. */
1229 if (namelen == 2
1230 && tokstart[0] == 'i'
1231 && tokstart[1] == 'f')
1233 return 0;
1236 /* For the same reason (breakpoint conditions), "thread N"
1237 terminates the expression. "thread" could be an identifier, but
1238 an identifier is never followed by a number without intervening
1239 punctuation.
1240 Handle abbreviations of these, similarly to
1241 breakpoint.c:find_condition_and_thread.
1242 TODO: Watch for "goroutine" here? */
1243 if (namelen >= 1
1244 && strncmp (tokstart, "thread", namelen) == 0
1245 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t'))
1247 const char *p = tokstart + namelen + 1;
1249 while (*p == ' ' || *p == '\t')
1250 p++;
1251 if (*p >= '0' && *p <= '9')
1252 return 0;
1255 lexptr += namelen;
1257 tryname:
1259 yylval.sval.ptr = tokstart;
1260 yylval.sval.length = namelen;
1262 /* Catch specific keywords. */
1263 copy = copy_name (yylval.sval);
1264 for (i = 0; i < sizeof (ident_tokens) / sizeof (ident_tokens[0]); i++)
1265 if (strcmp (copy, ident_tokens[i].oper) == 0)
1267 /* It is ok to always set this, even though we don't always
1268 strictly need to. */
1269 yylval.opcode = ident_tokens[i].opcode;
1270 return ident_tokens[i].token;
1273 if (*tokstart == '$')
1274 return DOLLAR_VARIABLE;
1276 if (parse_completion && *lexptr == '\0')
1277 saw_name_at_eof = 1;
1278 return NAME;
1281 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
1282 struct token_and_value
1284 int token;
1285 YYSTYPE value;
1288 /* A FIFO of tokens that have been read but not yet returned to the
1289 parser. */
1290 static std::vector<token_and_value> token_fifo;
1292 /* Non-zero if the lexer should return tokens from the FIFO. */
1293 static int popping;
1295 /* Temporary storage for yylex; this holds symbol names as they are
1296 built up. */
1297 static auto_obstack name_obstack;
1299 /* Build "package.name" in name_obstack.
1300 For convenience of the caller, the name is NUL-terminated,
1301 but the NUL is not included in the recorded length. */
1303 static struct stoken
1304 build_packaged_name (const char *package, int package_len,
1305 const char *name, int name_len)
1307 struct stoken result;
1309 name_obstack.clear ();
1310 obstack_grow (&name_obstack, package, package_len);
1311 obstack_grow_str (&name_obstack, ".");
1312 obstack_grow (&name_obstack, name, name_len);
1313 obstack_grow (&name_obstack, "", 1);
1314 result.ptr = (char *) obstack_base (&name_obstack);
1315 result.length = obstack_object_size (&name_obstack) - 1;
1317 return result;
1320 /* Return non-zero if NAME is a package name.
1321 BLOCK is the scope in which to interpret NAME; this can be NULL
1322 to mean the global scope. */
1324 static int
1325 package_name_p (const char *name, const struct block *block)
1327 struct symbol *sym;
1328 struct field_of_this_result is_a_field_of_this;
1330 sym = lookup_symbol (name, block, STRUCT_DOMAIN, &is_a_field_of_this).symbol;
1332 if (sym
1333 && SYMBOL_CLASS (sym) == LOC_TYPEDEF
1334 && TYPE_CODE (SYMBOL_TYPE (sym)) == TYPE_CODE_MODULE)
1335 return 1;
1337 return 0;
1340 /* Classify a (potential) function in the "unsafe" package.
1341 We fold these into "keywords" to keep things simple, at least until
1342 something more complex is warranted. */
1344 static int
1345 classify_unsafe_function (struct stoken function_name)
1347 char *copy = copy_name (function_name);
1349 if (strcmp (copy, "Sizeof") == 0)
1351 yylval.sval = function_name;
1352 return SIZEOF_KEYWORD;
1355 error (_("Unknown function in `unsafe' package: %s"), copy);
1358 /* Classify token(s) "name1.name2" where name1 is known to be a package.
1359 The contents of the token are in `yylval'.
1360 Updates yylval and returns the new token type.
1362 The result is one of NAME, NAME_OR_INT, or TYPENAME. */
1364 static int
1365 classify_packaged_name (const struct block *block)
1367 char *copy;
1368 struct block_symbol sym;
1369 struct field_of_this_result is_a_field_of_this;
1371 copy = copy_name (yylval.sval);
1373 sym = lookup_symbol (copy, block, VAR_DOMAIN, &is_a_field_of_this);
1375 if (sym.symbol)
1377 yylval.ssym.sym = sym;
1378 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
1381 return NAME;
1384 /* Classify a NAME token.
1385 The contents of the token are in `yylval'.
1386 Updates yylval and returns the new token type.
1387 BLOCK is the block in which lookups start; this can be NULL
1388 to mean the global scope.
1390 The result is one of NAME, NAME_OR_INT, or TYPENAME. */
1392 static int
1393 classify_name (struct parser_state *par_state, const struct block *block)
1395 struct type *type;
1396 struct block_symbol sym;
1397 char *copy;
1398 struct field_of_this_result is_a_field_of_this;
1400 copy = copy_name (yylval.sval);
1402 /* Try primitive types first so they win over bad/weird debug info. */
1403 type = language_lookup_primitive_type (parse_language (par_state),
1404 parse_gdbarch (par_state),
1405 copy);
1406 if (type != NULL)
1408 /* NOTE: We take advantage of the fact that yylval coming in was a
1409 NAME, and that struct ttype is a compatible extension of struct
1410 stoken, so yylval.tsym.stoken is already filled in. */
1411 yylval.tsym.type = type;
1412 return TYPENAME;
1415 /* TODO: What about other types? */
1417 sym = lookup_symbol (copy, block, VAR_DOMAIN, &is_a_field_of_this);
1419 if (sym.symbol)
1421 yylval.ssym.sym = sym;
1422 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
1423 return NAME;
1426 /* If we didn't find a symbol, look again in the current package.
1427 This is to, e.g., make "p global_var" work without having to specify
1428 the package name. We intentionally only looks for objects in the
1429 current package. */
1432 char *current_package_name = go_block_package_name (block);
1434 if (current_package_name != NULL)
1436 struct stoken sval =
1437 build_packaged_name (current_package_name,
1438 strlen (current_package_name),
1439 copy, strlen (copy));
1441 xfree (current_package_name);
1442 sym = lookup_symbol (sval.ptr, block, VAR_DOMAIN,
1443 &is_a_field_of_this);
1444 if (sym.symbol)
1446 yylval.ssym.stoken = sval;
1447 yylval.ssym.sym = sym;
1448 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
1449 return NAME;
1454 /* Input names that aren't symbols but ARE valid hex numbers, when
1455 the input radix permits them, can be names or numbers depending
1456 on the parse. Note we support radixes > 16 here. */
1457 if ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
1458 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10))
1460 YYSTYPE newlval; /* Its value is ignored. */
1461 int hextype = parse_number (par_state, copy, yylval.sval.length,
1462 0, &newlval);
1463 if (hextype == INT)
1465 yylval.ssym.sym.symbol = NULL;
1466 yylval.ssym.sym.block = NULL;
1467 yylval.ssym.is_a_field_of_this = 0;
1468 return NAME_OR_INT;
1472 yylval.ssym.sym.symbol = NULL;
1473 yylval.ssym.sym.block = NULL;
1474 yylval.ssym.is_a_field_of_this = 0;
1475 return NAME;
1478 /* This is taken from c-exp.y mostly to get something working.
1479 The basic structure has been kept because we may yet need some of it. */
1481 static int
1482 yylex (void)
1484 token_and_value current, next;
1486 if (popping && !token_fifo.empty ())
1488 token_and_value tv = token_fifo[0];
1489 token_fifo.erase (token_fifo.begin ());
1490 yylval = tv.value;
1491 /* There's no need to fall through to handle package.name
1492 as that can never happen here. In theory. */
1493 return tv.token;
1495 popping = 0;
1497 current.token = lex_one_token (pstate);
1499 /* TODO: Need a way to force specifying name1 as a package.
1500 .name1.name2 ? */
1502 if (current.token != NAME)
1503 return current.token;
1505 /* See if we have "name1 . name2". */
1507 current.value = yylval;
1508 next.token = lex_one_token (pstate);
1509 next.value = yylval;
1511 if (next.token == '.')
1513 token_and_value name2;
1515 name2.token = lex_one_token (pstate);
1516 name2.value = yylval;
1518 if (name2.token == NAME)
1520 /* Ok, we have "name1 . name2". */
1521 char *copy;
1523 copy = copy_name (current.value.sval);
1525 if (strcmp (copy, "unsafe") == 0)
1527 popping = 1;
1528 return classify_unsafe_function (name2.value.sval);
1531 if (package_name_p (copy, expression_context_block))
1533 popping = 1;
1534 yylval.sval = build_packaged_name (current.value.sval.ptr,
1535 current.value.sval.length,
1536 name2.value.sval.ptr,
1537 name2.value.sval.length);
1538 return classify_packaged_name (expression_context_block);
1542 token_fifo.push_back (next);
1543 token_fifo.push_back (name2);
1545 else
1546 token_fifo.push_back (next);
1548 /* If we arrive here we don't have a package-qualified name. */
1550 popping = 1;
1551 yylval = current.value;
1552 return classify_name (pstate, expression_context_block);
1556 go_parse (struct parser_state *par_state)
1558 /* Setting up the parser state. */
1559 scoped_restore pstate_restore = make_scoped_restore (&pstate);
1560 gdb_assert (par_state != NULL);
1561 pstate = par_state;
1563 scoped_restore restore_yydebug = make_scoped_restore (&yydebug,
1564 parser_debug);
1566 /* Initialize some state used by the lexer. */
1567 last_was_structop = 0;
1568 saw_name_at_eof = 0;
1570 token_fifo.clear ();
1571 popping = 0;
1572 name_obstack.clear ();
1574 return yyparse ();
1577 static void
1578 yyerror (const char *msg)
1580 if (prev_lexptr)
1581 lexptr = prev_lexptr;
1583 error (_("A %s in expression, near `%s'."), msg, lexptr);