gdb/ChangeLog:
[deliverable/binutils-gdb.git] / gdb / parse.c
1 /* Parse expressions for GDB.
2 Copyright 1986, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997,
3 1998, 1999, 2000, 2001 Free Software Foundation, Inc.
4 Modified from expread.y by the Department of Computer Science at the
5 State University of New York at Buffalo, 1991.
6
7 This file is part of GDB.
8
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 2 of the License, or
12 (at your option) any later version.
13
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
18
19 You should have received a copy of the GNU General Public License
20 along with this program; if not, write to the Free Software
21 Foundation, Inc., 59 Temple Place - Suite 330,
22 Boston, MA 02111-1307, USA. */
23
24 /* Parse an expression from text in a string,
25 and return the result as a struct expression pointer.
26 That structure contains arithmetic operations in reverse polish,
27 with constants represented by operations that are followed by special data.
28 See expression.h for the details of the format.
29 What is important here is that it can be built up sequentially
30 during the process of parsing; the lower levels of the tree always
31 come first in the result. */
32
33 #include <ctype.h>
34
35 #include "defs.h"
36 #include "gdb_string.h"
37 #include "symtab.h"
38 #include "gdbtypes.h"
39 #include "frame.h"
40 #include "expression.h"
41 #include "value.h"
42 #include "command.h"
43 #include "language.h"
44 #include "parser-defs.h"
45 #include "gdbcmd.h"
46 #include "symfile.h" /* for overlay functions */
47 #include "inferior.h" /* for NUM_PSEUDO_REGS. NOTE: replace
48 with "gdbarch.h" when appropriate. */
49 #include "doublest.h"
50
51 \f
52 /* Symbols which architectures can redefine. */
53
54 /* Some systems have routines whose names start with `$'. Giving this
55 macro a non-zero value tells GDB's expression parser to check for
56 such routines when parsing tokens that begin with `$'.
57
58 On HP-UX, certain system routines (millicode) have names beginning
59 with `$' or `$$'. For example, `$$dyncall' is a millicode routine
60 that handles inter-space procedure calls on PA-RISC. */
61 #ifndef SYMBOLS_CAN_START_WITH_DOLLAR
62 #define SYMBOLS_CAN_START_WITH_DOLLAR (0)
63 #endif
64
65
66 \f
67 /* Global variables declared in parser-defs.h (and commented there). */
68 struct expression *expout;
69 int expout_size;
70 int expout_ptr;
71 struct block *expression_context_block;
72 struct block *innermost_block;
73 int arglist_len;
74 union type_stack_elt *type_stack;
75 int type_stack_depth, type_stack_size;
76 char *lexptr;
77 char *namecopy;
78 int paren_depth;
79 int comma_terminates;
80 \f
81 static int expressiondebug = 0;
82
83 extern int hp_som_som_object_present;
84
85 static void free_funcalls (void *ignore);
86
87 static void prefixify_expression (struct expression *);
88
89 static void
90 prefixify_subexp (struct expression *, struct expression *, int, int);
91
92 void _initialize_parse (void);
93
94 /* Data structure for saving values of arglist_len for function calls whose
95 arguments contain other function calls. */
96
97 struct funcall
98 {
99 struct funcall *next;
100 int arglist_len;
101 };
102
103 static struct funcall *funcall_chain;
104
105 /* Assign machine-independent names to certain registers
106 (unless overridden by the REGISTER_NAMES table) */
107
108 unsigned num_std_regs = 0;
109 struct std_regs *std_regs;
110
111 /* The generic method for targets to specify how their registers are
112 named. The mapping can be derived from three sources:
113 REGISTER_NAME; std_regs; or a target specific alias hook. */
114
115 int
116 target_map_name_to_register (char *str, int len)
117 {
118 int i;
119
120 /* Search register name space. */
121 for (i = 0; i < NUM_REGS + NUM_PSEUDO_REGS; i++)
122 if (REGISTER_NAME (i) && len == strlen (REGISTER_NAME (i))
123 && STREQN (str, REGISTER_NAME (i), len))
124 {
125 return i;
126 }
127
128 /* Try standard aliases. */
129 for (i = 0; i < num_std_regs; i++)
130 if (std_regs[i].name && len == strlen (std_regs[i].name)
131 && STREQN (str, std_regs[i].name, len))
132 {
133 return std_regs[i].regnum;
134 }
135
136 return -1;
137 }
138
139 /* Begin counting arguments for a function call,
140 saving the data about any containing call. */
141
142 void
143 start_arglist (void)
144 {
145 register struct funcall *new;
146
147 new = (struct funcall *) xmalloc (sizeof (struct funcall));
148 new->next = funcall_chain;
149 new->arglist_len = arglist_len;
150 arglist_len = 0;
151 funcall_chain = new;
152 }
153
154 /* Return the number of arguments in a function call just terminated,
155 and restore the data for the containing function call. */
156
157 int
158 end_arglist (void)
159 {
160 register int val = arglist_len;
161 register struct funcall *call = funcall_chain;
162 funcall_chain = call->next;
163 arglist_len = call->arglist_len;
164 xfree (call);
165 return val;
166 }
167
168 /* Free everything in the funcall chain.
169 Used when there is an error inside parsing. */
170
171 static void
172 free_funcalls (void *ignore)
173 {
174 register struct funcall *call, *next;
175
176 for (call = funcall_chain; call; call = next)
177 {
178 next = call->next;
179 xfree (call);
180 }
181 }
182 \f
183 /* This page contains the functions for adding data to the struct expression
184 being constructed. */
185
186 /* Add one element to the end of the expression. */
187
188 /* To avoid a bug in the Sun 4 compiler, we pass things that can fit into
189 a register through here */
190
191 void
192 write_exp_elt (union exp_element expelt)
193 {
194 if (expout_ptr >= expout_size)
195 {
196 expout_size *= 2;
197 expout = (struct expression *)
198 xrealloc ((char *) expout, sizeof (struct expression)
199 + EXP_ELEM_TO_BYTES (expout_size));
200 }
201 expout->elts[expout_ptr++] = expelt;
202 }
203
204 void
205 write_exp_elt_opcode (enum exp_opcode expelt)
206 {
207 union exp_element tmp;
208
209 tmp.opcode = expelt;
210
211 write_exp_elt (tmp);
212 }
213
214 void
215 write_exp_elt_sym (struct symbol *expelt)
216 {
217 union exp_element tmp;
218
219 tmp.symbol = expelt;
220
221 write_exp_elt (tmp);
222 }
223
224 void
225 write_exp_elt_block (struct block *b)
226 {
227 union exp_element tmp;
228 tmp.block = b;
229 write_exp_elt (tmp);
230 }
231
232 void
233 write_exp_elt_longcst (LONGEST expelt)
234 {
235 union exp_element tmp;
236
237 tmp.longconst = expelt;
238
239 write_exp_elt (tmp);
240 }
241
242 void
243 write_exp_elt_dblcst (DOUBLEST expelt)
244 {
245 union exp_element tmp;
246
247 tmp.doubleconst = expelt;
248
249 write_exp_elt (tmp);
250 }
251
252 void
253 write_exp_elt_type (struct type *expelt)
254 {
255 union exp_element tmp;
256
257 tmp.type = expelt;
258
259 write_exp_elt (tmp);
260 }
261
262 void
263 write_exp_elt_intern (struct internalvar *expelt)
264 {
265 union exp_element tmp;
266
267 tmp.internalvar = expelt;
268
269 write_exp_elt (tmp);
270 }
271
272 /* Add a string constant to the end of the expression.
273
274 String constants are stored by first writing an expression element
275 that contains the length of the string, then stuffing the string
276 constant itself into however many expression elements are needed
277 to hold it, and then writing another expression element that contains
278 the length of the string. I.E. an expression element at each end of
279 the string records the string length, so you can skip over the
280 expression elements containing the actual string bytes from either
281 end of the string. Note that this also allows gdb to handle
282 strings with embedded null bytes, as is required for some languages.
283
284 Don't be fooled by the fact that the string is null byte terminated,
285 this is strictly for the convenience of debugging gdb itself. Gdb
286 Gdb does not depend up the string being null terminated, since the
287 actual length is recorded in expression elements at each end of the
288 string. The null byte is taken into consideration when computing how
289 many expression elements are required to hold the string constant, of
290 course. */
291
292
293 void
294 write_exp_string (struct stoken str)
295 {
296 register int len = str.length;
297 register int lenelt;
298 register char *strdata;
299
300 /* Compute the number of expression elements required to hold the string
301 (including a null byte terminator), along with one expression element
302 at each end to record the actual string length (not including the
303 null byte terminator). */
304
305 lenelt = 2 + BYTES_TO_EXP_ELEM (len + 1);
306
307 /* Ensure that we have enough available expression elements to store
308 everything. */
309
310 if ((expout_ptr + lenelt) >= expout_size)
311 {
312 expout_size = max (expout_size * 2, expout_ptr + lenelt + 10);
313 expout = (struct expression *)
314 xrealloc ((char *) expout, (sizeof (struct expression)
315 + EXP_ELEM_TO_BYTES (expout_size)));
316 }
317
318 /* Write the leading length expression element (which advances the current
319 expression element index), then write the string constant followed by a
320 terminating null byte, and then write the trailing length expression
321 element. */
322
323 write_exp_elt_longcst ((LONGEST) len);
324 strdata = (char *) &expout->elts[expout_ptr];
325 memcpy (strdata, str.ptr, len);
326 *(strdata + len) = '\0';
327 expout_ptr += lenelt - 2;
328 write_exp_elt_longcst ((LONGEST) len);
329 }
330
331 /* Add a bitstring constant to the end of the expression.
332
333 Bitstring constants are stored by first writing an expression element
334 that contains the length of the bitstring (in bits), then stuffing the
335 bitstring constant itself into however many expression elements are
336 needed to hold it, and then writing another expression element that
337 contains the length of the bitstring. I.E. an expression element at
338 each end of the bitstring records the bitstring length, so you can skip
339 over the expression elements containing the actual bitstring bytes from
340 either end of the bitstring. */
341
342 void
343 write_exp_bitstring (struct stoken str)
344 {
345 register int bits = str.length; /* length in bits */
346 register int len = (bits + HOST_CHAR_BIT - 1) / HOST_CHAR_BIT;
347 register int lenelt;
348 register char *strdata;
349
350 /* Compute the number of expression elements required to hold the bitstring,
351 along with one expression element at each end to record the actual
352 bitstring length in bits. */
353
354 lenelt = 2 + BYTES_TO_EXP_ELEM (len);
355
356 /* Ensure that we have enough available expression elements to store
357 everything. */
358
359 if ((expout_ptr + lenelt) >= expout_size)
360 {
361 expout_size = max (expout_size * 2, expout_ptr + lenelt + 10);
362 expout = (struct expression *)
363 xrealloc ((char *) expout, (sizeof (struct expression)
364 + EXP_ELEM_TO_BYTES (expout_size)));
365 }
366
367 /* Write the leading length expression element (which advances the current
368 expression element index), then write the bitstring constant, and then
369 write the trailing length expression element. */
370
371 write_exp_elt_longcst ((LONGEST) bits);
372 strdata = (char *) &expout->elts[expout_ptr];
373 memcpy (strdata, str.ptr, len);
374 expout_ptr += lenelt - 2;
375 write_exp_elt_longcst ((LONGEST) bits);
376 }
377
378 /* Add the appropriate elements for a minimal symbol to the end of
379 the expression. The rationale behind passing in text_symbol_type and
380 data_symbol_type was so that Modula-2 could pass in WORD for
381 data_symbol_type. Perhaps it still is useful to have those types vary
382 based on the language, but they no longer have names like "int", so
383 the initial rationale is gone. */
384
385 static struct type *msym_text_symbol_type;
386 static struct type *msym_data_symbol_type;
387 static struct type *msym_unknown_symbol_type;
388
389 void
390 write_exp_msymbol (struct minimal_symbol *msymbol,
391 struct type *text_symbol_type,
392 struct type *data_symbol_type)
393 {
394 CORE_ADDR addr;
395
396 write_exp_elt_opcode (OP_LONG);
397 /* Let's make the type big enough to hold a 64-bit address. */
398 write_exp_elt_type (builtin_type_CORE_ADDR);
399
400 addr = SYMBOL_VALUE_ADDRESS (msymbol);
401 if (overlay_debugging)
402 addr = symbol_overlayed_address (addr, SYMBOL_BFD_SECTION (msymbol));
403 write_exp_elt_longcst ((LONGEST) addr);
404
405 write_exp_elt_opcode (OP_LONG);
406
407 write_exp_elt_opcode (UNOP_MEMVAL);
408 switch (msymbol->type)
409 {
410 case mst_text:
411 case mst_file_text:
412 case mst_solib_trampoline:
413 write_exp_elt_type (msym_text_symbol_type);
414 break;
415
416 case mst_data:
417 case mst_file_data:
418 case mst_bss:
419 case mst_file_bss:
420 write_exp_elt_type (msym_data_symbol_type);
421 break;
422
423 default:
424 write_exp_elt_type (msym_unknown_symbol_type);
425 break;
426 }
427 write_exp_elt_opcode (UNOP_MEMVAL);
428 }
429 \f
430 /* Recognize tokens that start with '$'. These include:
431
432 $regname A native register name or a "standard
433 register name".
434
435 $variable A convenience variable with a name chosen
436 by the user.
437
438 $digits Value history with index <digits>, starting
439 from the first value which has index 1.
440
441 $$digits Value history with index <digits> relative
442 to the last value. I.E. $$0 is the last
443 value, $$1 is the one previous to that, $$2
444 is the one previous to $$1, etc.
445
446 $ | $0 | $$0 The last value in the value history.
447
448 $$ An abbreviation for the second to the last
449 value in the value history, I.E. $$1
450
451 */
452
453 void
454 write_dollar_variable (struct stoken str)
455 {
456 /* Handle the tokens $digits; also $ (short for $0) and $$ (short for $$1)
457 and $$digits (equivalent to $<-digits> if you could type that). */
458
459 int negate = 0;
460 int i = 1;
461 /* Double dollar means negate the number and add -1 as well.
462 Thus $$ alone means -1. */
463 if (str.length >= 2 && str.ptr[1] == '$')
464 {
465 negate = 1;
466 i = 2;
467 }
468 if (i == str.length)
469 {
470 /* Just dollars (one or two) */
471 i = -negate;
472 goto handle_last;
473 }
474 /* Is the rest of the token digits? */
475 for (; i < str.length; i++)
476 if (!(str.ptr[i] >= '0' && str.ptr[i] <= '9'))
477 break;
478 if (i == str.length)
479 {
480 i = atoi (str.ptr + 1 + negate);
481 if (negate)
482 i = -i;
483 goto handle_last;
484 }
485
486 /* Handle tokens that refer to machine registers:
487 $ followed by a register name. */
488 i = target_map_name_to_register (str.ptr + 1, str.length - 1);
489 if (i >= 0)
490 goto handle_register;
491
492 if (SYMBOLS_CAN_START_WITH_DOLLAR)
493 {
494 struct symbol *sym = NULL;
495 struct minimal_symbol *msym = NULL;
496
497 /* On HP-UX, certain system routines (millicode) have names beginning
498 with $ or $$, e.g. $$dyncall, which handles inter-space procedure
499 calls on PA-RISC. Check for those, first. */
500
501 /* This code is not enabled on non HP-UX systems, since worst case
502 symbol table lookup performance is awful, to put it mildly. */
503
504 sym = lookup_symbol (copy_name (str), (struct block *) NULL,
505 VAR_NAMESPACE, (int *) NULL, (struct symtab **) NULL);
506 if (sym)
507 {
508 write_exp_elt_opcode (OP_VAR_VALUE);
509 write_exp_elt_block (block_found); /* set by lookup_symbol */
510 write_exp_elt_sym (sym);
511 write_exp_elt_opcode (OP_VAR_VALUE);
512 return;
513 }
514 msym = lookup_minimal_symbol (copy_name (str), NULL, NULL);
515 if (msym)
516 {
517 write_exp_msymbol (msym,
518 lookup_function_type (builtin_type_int),
519 builtin_type_int);
520 return;
521 }
522 }
523
524 /* Any other names starting in $ are debugger internal variables. */
525
526 write_exp_elt_opcode (OP_INTERNALVAR);
527 write_exp_elt_intern (lookup_internalvar (copy_name (str) + 1));
528 write_exp_elt_opcode (OP_INTERNALVAR);
529 return;
530 handle_last:
531 write_exp_elt_opcode (OP_LAST);
532 write_exp_elt_longcst ((LONGEST) i);
533 write_exp_elt_opcode (OP_LAST);
534 return;
535 handle_register:
536 write_exp_elt_opcode (OP_REGISTER);
537 write_exp_elt_longcst (i);
538 write_exp_elt_opcode (OP_REGISTER);
539 return;
540 }
541
542
543 /* Parse a string that is possibly a namespace / nested class
544 specification, i.e., something of the form A::B::C::x. Input
545 (NAME) is the entire string; LEN is the current valid length; the
546 output is a string, TOKEN, which points to the largest recognized
547 prefix which is a series of namespaces or classes. CLASS_PREFIX is
548 another output, which records whether a nested class spec was
549 recognized (= 1) or a fully qualified variable name was found (=
550 0). ARGPTR is side-effected (if non-NULL) to point to beyond the
551 string recognized and consumed by this routine.
552
553 The return value is a pointer to the symbol for the base class or
554 variable if found, or NULL if not found. Callers must check this
555 first -- if NULL, the outputs may not be correct.
556
557 This function is used c-exp.y. This is used specifically to get
558 around HP aCC (and possibly other compilers), which insists on
559 generating names with embedded colons for namespace or nested class
560 members.
561
562 (Argument LEN is currently unused. 1997-08-27)
563
564 Callers must free memory allocated for the output string TOKEN. */
565
566 static const char coloncolon[2] =
567 {':', ':'};
568
569 struct symbol *
570 parse_nested_classes_for_hpacc (char *name, int len, char **token,
571 int *class_prefix, char **argptr)
572 {
573 /* Comment below comes from decode_line_1 which has very similar
574 code, which is called for "break" command parsing. */
575
576 /* We have what looks like a class or namespace
577 scope specification (A::B), possibly with many
578 levels of namespaces or classes (A::B::C::D).
579
580 Some versions of the HP ANSI C++ compiler (as also possibly
581 other compilers) generate class/function/member names with
582 embedded double-colons if they are inside namespaces. To
583 handle this, we loop a few times, considering larger and
584 larger prefixes of the string as though they were single
585 symbols. So, if the initially supplied string is
586 A::B::C::D::foo, we have to look up "A", then "A::B",
587 then "A::B::C", then "A::B::C::D", and finally
588 "A::B::C::D::foo" as single, monolithic symbols, because
589 A, B, C or D may be namespaces.
590
591 Note that namespaces can nest only inside other
592 namespaces, and not inside classes. So we need only
593 consider *prefixes* of the string; there is no need to look up
594 "B::C" separately as a symbol in the previous example. */
595
596 register char *p;
597 char *start, *end;
598 char *prefix = NULL;
599 char *tmp;
600 struct symbol *sym_class = NULL;
601 struct symbol *sym_var = NULL;
602 struct type *t;
603 int prefix_len = 0;
604 int done = 0;
605 char *q;
606
607 /* Check for HP-compiled executable -- in other cases
608 return NULL, and caller must default to standard GDB
609 behaviour. */
610
611 if (!hp_som_som_object_present)
612 return (struct symbol *) NULL;
613
614 p = name;
615
616 /* Skip over whitespace and possible global "::" */
617 while (*p && (*p == ' ' || *p == '\t'))
618 p++;
619 if (p[0] == ':' && p[1] == ':')
620 p += 2;
621 while (*p && (*p == ' ' || *p == '\t'))
622 p++;
623
624 while (1)
625 {
626 /* Get to the end of the next namespace or class spec. */
627 /* If we're looking at some non-token, fail immediately */
628 start = p;
629 if (!(isalpha (*p) || *p == '$' || *p == '_'))
630 return (struct symbol *) NULL;
631 p++;
632 while (*p && (isalnum (*p) || *p == '$' || *p == '_'))
633 p++;
634
635 if (*p == '<')
636 {
637 /* If we have the start of a template specification,
638 scan right ahead to its end */
639 q = find_template_name_end (p);
640 if (q)
641 p = q;
642 }
643
644 end = p;
645
646 /* Skip over "::" and whitespace for next time around */
647 while (*p && (*p == ' ' || *p == '\t'))
648 p++;
649 if (p[0] == ':' && p[1] == ':')
650 p += 2;
651 while (*p && (*p == ' ' || *p == '\t'))
652 p++;
653
654 /* Done with tokens? */
655 if (!*p || !(isalpha (*p) || *p == '$' || *p == '_'))
656 done = 1;
657
658 tmp = (char *) alloca (prefix_len + end - start + 3);
659 if (prefix)
660 {
661 memcpy (tmp, prefix, prefix_len);
662 memcpy (tmp + prefix_len, coloncolon, 2);
663 memcpy (tmp + prefix_len + 2, start, end - start);
664 tmp[prefix_len + 2 + end - start] = '\000';
665 }
666 else
667 {
668 memcpy (tmp, start, end - start);
669 tmp[end - start] = '\000';
670 }
671
672 prefix = tmp;
673 prefix_len = strlen (prefix);
674
675 /* See if the prefix we have now is something we know about */
676
677 if (!done)
678 {
679 /* More tokens to process, so this must be a class/namespace */
680 sym_class = lookup_symbol (prefix, 0, STRUCT_NAMESPACE,
681 0, (struct symtab **) NULL);
682 }
683 else
684 {
685 /* No more tokens, so try as a variable first */
686 sym_var = lookup_symbol (prefix, 0, VAR_NAMESPACE,
687 0, (struct symtab **) NULL);
688 /* If failed, try as class/namespace */
689 if (!sym_var)
690 sym_class = lookup_symbol (prefix, 0, STRUCT_NAMESPACE,
691 0, (struct symtab **) NULL);
692 }
693
694 if (sym_var ||
695 (sym_class &&
696 (t = check_typedef (SYMBOL_TYPE (sym_class)),
697 (TYPE_CODE (t) == TYPE_CODE_STRUCT
698 || TYPE_CODE (t) == TYPE_CODE_UNION))))
699 {
700 /* We found a valid token */
701 *token = (char *) xmalloc (prefix_len + 1);
702 memcpy (*token, prefix, prefix_len);
703 (*token)[prefix_len] = '\000';
704 break;
705 }
706
707 /* No variable or class/namespace found, no more tokens */
708 if (done)
709 return (struct symbol *) NULL;
710 }
711
712 /* Out of loop, so we must have found a valid token */
713 if (sym_var)
714 *class_prefix = 0;
715 else
716 *class_prefix = 1;
717
718 if (argptr)
719 *argptr = done ? p : end;
720
721 return sym_var ? sym_var : sym_class; /* found */
722 }
723
724 char *
725 find_template_name_end (char *p)
726 {
727 int depth = 1;
728 int just_seen_right = 0;
729 int just_seen_colon = 0;
730 int just_seen_space = 0;
731
732 if (!p || (*p != '<'))
733 return 0;
734
735 while (*++p)
736 {
737 switch (*p)
738 {
739 case '\'':
740 case '\"':
741 case '{':
742 case '}':
743 /* In future, may want to allow these?? */
744 return 0;
745 case '<':
746 depth++; /* start nested template */
747 if (just_seen_colon || just_seen_right || just_seen_space)
748 return 0; /* but not after : or :: or > or space */
749 break;
750 case '>':
751 if (just_seen_colon || just_seen_right)
752 return 0; /* end a (nested?) template */
753 just_seen_right = 1; /* but not after : or :: */
754 if (--depth == 0) /* also disallow >>, insist on > > */
755 return ++p; /* if outermost ended, return */
756 break;
757 case ':':
758 if (just_seen_space || (just_seen_colon > 1))
759 return 0; /* nested class spec coming up */
760 just_seen_colon++; /* we allow :: but not :::: */
761 break;
762 case ' ':
763 break;
764 default:
765 if (!((*p >= 'a' && *p <= 'z') || /* allow token chars */
766 (*p >= 'A' && *p <= 'Z') ||
767 (*p >= '0' && *p <= '9') ||
768 (*p == '_') || (*p == ',') || /* commas for template args */
769 (*p == '&') || (*p == '*') || /* pointer and ref types */
770 (*p == '(') || (*p == ')') || /* function types */
771 (*p == '[') || (*p == ']'))) /* array types */
772 return 0;
773 }
774 if (*p != ' ')
775 just_seen_space = 0;
776 if (*p != ':')
777 just_seen_colon = 0;
778 if (*p != '>')
779 just_seen_right = 0;
780 }
781 return 0;
782 }
783 \f
784
785
786 /* Return a null-terminated temporary copy of the name
787 of a string token. */
788
789 char *
790 copy_name (struct stoken token)
791 {
792 memcpy (namecopy, token.ptr, token.length);
793 namecopy[token.length] = 0;
794 return namecopy;
795 }
796 \f
797 /* Reverse an expression from suffix form (in which it is constructed)
798 to prefix form (in which we can conveniently print or execute it). */
799
800 static void
801 prefixify_expression (register struct expression *expr)
802 {
803 register int len =
804 sizeof (struct expression) + EXP_ELEM_TO_BYTES (expr->nelts);
805 register struct expression *temp;
806 register int inpos = expr->nelts, outpos = 0;
807
808 temp = (struct expression *) alloca (len);
809
810 /* Copy the original expression into temp. */
811 memcpy (temp, expr, len);
812
813 prefixify_subexp (temp, expr, inpos, outpos);
814 }
815
816 /* Return the number of exp_elements in the subexpression of EXPR
817 whose last exp_element is at index ENDPOS - 1 in EXPR. */
818
819 int
820 length_of_subexp (register struct expression *expr, register int endpos)
821 {
822 register int oplen = 1;
823 register int args = 0;
824 register int i;
825
826 if (endpos < 1)
827 error ("?error in length_of_subexp");
828
829 i = (int) expr->elts[endpos - 1].opcode;
830
831 switch (i)
832 {
833 /* C++ */
834 case OP_SCOPE:
835 oplen = longest_to_int (expr->elts[endpos - 2].longconst);
836 oplen = 5 + BYTES_TO_EXP_ELEM (oplen + 1);
837 break;
838
839 case OP_LONG:
840 case OP_DOUBLE:
841 case OP_VAR_VALUE:
842 oplen = 4;
843 break;
844
845 case OP_TYPE:
846 case OP_BOOL:
847 case OP_LAST:
848 case OP_REGISTER:
849 case OP_INTERNALVAR:
850 oplen = 3;
851 break;
852
853 case OP_COMPLEX:
854 oplen = 1;
855 args = 2;
856 break;
857
858 case OP_FUNCALL:
859 case OP_F77_UNDETERMINED_ARGLIST:
860 oplen = 3;
861 args = 1 + longest_to_int (expr->elts[endpos - 2].longconst);
862 break;
863
864 case UNOP_MAX:
865 case UNOP_MIN:
866 oplen = 3;
867 break;
868
869 case BINOP_VAL:
870 case UNOP_CAST:
871 case UNOP_MEMVAL:
872 oplen = 3;
873 args = 1;
874 break;
875
876 case UNOP_ABS:
877 case UNOP_CAP:
878 case UNOP_CHR:
879 case UNOP_FLOAT:
880 case UNOP_HIGH:
881 case UNOP_ODD:
882 case UNOP_ORD:
883 case UNOP_TRUNC:
884 oplen = 1;
885 args = 1;
886 break;
887
888 case OP_LABELED:
889 case STRUCTOP_STRUCT:
890 case STRUCTOP_PTR:
891 args = 1;
892 /* fall through */
893 case OP_M2_STRING:
894 case OP_STRING:
895 case OP_NAME:
896 case OP_EXPRSTRING:
897 oplen = longest_to_int (expr->elts[endpos - 2].longconst);
898 oplen = 4 + BYTES_TO_EXP_ELEM (oplen + 1);
899 break;
900
901 case OP_BITSTRING:
902 oplen = longest_to_int (expr->elts[endpos - 2].longconst);
903 oplen = (oplen + HOST_CHAR_BIT - 1) / HOST_CHAR_BIT;
904 oplen = 4 + BYTES_TO_EXP_ELEM (oplen);
905 break;
906
907 case OP_ARRAY:
908 oplen = 4;
909 args = longest_to_int (expr->elts[endpos - 2].longconst);
910 args -= longest_to_int (expr->elts[endpos - 3].longconst);
911 args += 1;
912 break;
913
914 case TERNOP_COND:
915 case TERNOP_SLICE:
916 case TERNOP_SLICE_COUNT:
917 args = 3;
918 break;
919
920 /* Modula-2 */
921 case MULTI_SUBSCRIPT:
922 oplen = 3;
923 args = 1 + longest_to_int (expr->elts[endpos - 2].longconst);
924 break;
925
926 case BINOP_ASSIGN_MODIFY:
927 oplen = 3;
928 args = 2;
929 break;
930
931 /* C++ */
932 case OP_THIS:
933 oplen = 2;
934 break;
935
936 default:
937 args = 1 + (i < (int) BINOP_END);
938 }
939
940 while (args > 0)
941 {
942 oplen += length_of_subexp (expr, endpos - oplen);
943 args--;
944 }
945
946 return oplen;
947 }
948
949 /* Copy the subexpression ending just before index INEND in INEXPR
950 into OUTEXPR, starting at index OUTBEG.
951 In the process, convert it from suffix to prefix form. */
952
953 static void
954 prefixify_subexp (register struct expression *inexpr,
955 struct expression *outexpr, register int inend, int outbeg)
956 {
957 register int oplen = 1;
958 register int args = 0;
959 register int i;
960 int *arglens;
961 enum exp_opcode opcode;
962
963 /* Compute how long the last operation is (in OPLEN),
964 and also how many preceding subexpressions serve as
965 arguments for it (in ARGS). */
966
967 opcode = inexpr->elts[inend - 1].opcode;
968 switch (opcode)
969 {
970 /* C++ */
971 case OP_SCOPE:
972 oplen = longest_to_int (inexpr->elts[inend - 2].longconst);
973 oplen = 5 + BYTES_TO_EXP_ELEM (oplen + 1);
974 break;
975
976 case OP_LONG:
977 case OP_DOUBLE:
978 case OP_VAR_VALUE:
979 oplen = 4;
980 break;
981
982 case OP_TYPE:
983 case OP_BOOL:
984 case OP_LAST:
985 case OP_REGISTER:
986 case OP_INTERNALVAR:
987 oplen = 3;
988 break;
989
990 case OP_COMPLEX:
991 oplen = 1;
992 args = 2;
993 break;
994
995 case OP_FUNCALL:
996 case OP_F77_UNDETERMINED_ARGLIST:
997 oplen = 3;
998 args = 1 + longest_to_int (inexpr->elts[inend - 2].longconst);
999 break;
1000
1001 case UNOP_MIN:
1002 case UNOP_MAX:
1003 oplen = 3;
1004 break;
1005
1006 case UNOP_CAST:
1007 case UNOP_MEMVAL:
1008 oplen = 3;
1009 args = 1;
1010 break;
1011
1012 case UNOP_ABS:
1013 case UNOP_CAP:
1014 case UNOP_CHR:
1015 case UNOP_FLOAT:
1016 case UNOP_HIGH:
1017 case UNOP_ODD:
1018 case UNOP_ORD:
1019 case UNOP_TRUNC:
1020 oplen = 1;
1021 args = 1;
1022 break;
1023
1024 case STRUCTOP_STRUCT:
1025 case STRUCTOP_PTR:
1026 case OP_LABELED:
1027 args = 1;
1028 /* fall through */
1029 case OP_M2_STRING:
1030 case OP_STRING:
1031 case OP_NAME:
1032 case OP_EXPRSTRING:
1033 oplen = longest_to_int (inexpr->elts[inend - 2].longconst);
1034 oplen = 4 + BYTES_TO_EXP_ELEM (oplen + 1);
1035 break;
1036
1037 case OP_BITSTRING:
1038 oplen = longest_to_int (inexpr->elts[inend - 2].longconst);
1039 oplen = (oplen + HOST_CHAR_BIT - 1) / HOST_CHAR_BIT;
1040 oplen = 4 + BYTES_TO_EXP_ELEM (oplen);
1041 break;
1042
1043 case OP_ARRAY:
1044 oplen = 4;
1045 args = longest_to_int (inexpr->elts[inend - 2].longconst);
1046 args -= longest_to_int (inexpr->elts[inend - 3].longconst);
1047 args += 1;
1048 break;
1049
1050 case TERNOP_COND:
1051 case TERNOP_SLICE:
1052 case TERNOP_SLICE_COUNT:
1053 args = 3;
1054 break;
1055
1056 case BINOP_ASSIGN_MODIFY:
1057 oplen = 3;
1058 args = 2;
1059 break;
1060
1061 /* Modula-2 */
1062 case MULTI_SUBSCRIPT:
1063 oplen = 3;
1064 args = 1 + longest_to_int (inexpr->elts[inend - 2].longconst);
1065 break;
1066
1067 /* C++ */
1068 case OP_THIS:
1069 oplen = 2;
1070 break;
1071
1072 default:
1073 args = 1 + ((int) opcode < (int) BINOP_END);
1074 }
1075
1076 /* Copy the final operator itself, from the end of the input
1077 to the beginning of the output. */
1078 inend -= oplen;
1079 memcpy (&outexpr->elts[outbeg], &inexpr->elts[inend],
1080 EXP_ELEM_TO_BYTES (oplen));
1081 outbeg += oplen;
1082
1083 /* Find the lengths of the arg subexpressions. */
1084 arglens = (int *) alloca (args * sizeof (int));
1085 for (i = args - 1; i >= 0; i--)
1086 {
1087 oplen = length_of_subexp (inexpr, inend);
1088 arglens[i] = oplen;
1089 inend -= oplen;
1090 }
1091
1092 /* Now copy each subexpression, preserving the order of
1093 the subexpressions, but prefixifying each one.
1094 In this loop, inend starts at the beginning of
1095 the expression this level is working on
1096 and marches forward over the arguments.
1097 outbeg does similarly in the output. */
1098 for (i = 0; i < args; i++)
1099 {
1100 oplen = arglens[i];
1101 inend += oplen;
1102 prefixify_subexp (inexpr, outexpr, inend, outbeg);
1103 outbeg += oplen;
1104 }
1105 }
1106 \f
1107 /* This page contains the two entry points to this file. */
1108
1109 /* Read an expression from the string *STRINGPTR points to,
1110 parse it, and return a pointer to a struct expression that we malloc.
1111 Use block BLOCK as the lexical context for variable names;
1112 if BLOCK is zero, use the block of the selected stack frame.
1113 Meanwhile, advance *STRINGPTR to point after the expression,
1114 at the first nonwhite character that is not part of the expression
1115 (possibly a null character).
1116
1117 If COMMA is nonzero, stop if a comma is reached. */
1118
1119 struct expression *
1120 parse_exp_1 (char **stringptr, struct block *block, int comma)
1121 {
1122 struct cleanup *old_chain;
1123
1124 lexptr = *stringptr;
1125
1126 paren_depth = 0;
1127 type_stack_depth = 0;
1128
1129 comma_terminates = comma;
1130
1131 if (lexptr == 0 || *lexptr == 0)
1132 error_no_arg ("expression to compute");
1133
1134 old_chain = make_cleanup (free_funcalls, 0 /*ignore*/);
1135 funcall_chain = 0;
1136
1137 expression_context_block = block ? block : get_selected_block (0);
1138
1139 namecopy = (char *) alloca (strlen (lexptr) + 1);
1140 expout_size = 10;
1141 expout_ptr = 0;
1142 expout = (struct expression *)
1143 xmalloc (sizeof (struct expression) + EXP_ELEM_TO_BYTES (expout_size));
1144 expout->language_defn = current_language;
1145 make_cleanup (free_current_contents, &expout);
1146
1147 if (current_language->la_parser ())
1148 current_language->la_error (NULL);
1149
1150 discard_cleanups (old_chain);
1151
1152 /* Record the actual number of expression elements, and then
1153 reallocate the expression memory so that we free up any
1154 excess elements. */
1155
1156 expout->nelts = expout_ptr;
1157 expout = (struct expression *)
1158 xrealloc ((char *) expout,
1159 sizeof (struct expression) + EXP_ELEM_TO_BYTES (expout_ptr));;
1160
1161 /* Convert expression from postfix form as generated by yacc
1162 parser, to a prefix form. */
1163
1164 if (expressiondebug)
1165 dump_prefix_expression (expout, gdb_stdlog,
1166 "before conversion to prefix form");
1167
1168 prefixify_expression (expout);
1169
1170 if (expressiondebug)
1171 dump_postfix_expression (expout, gdb_stdlog,
1172 "after conversion to prefix form");
1173
1174 *stringptr = lexptr;
1175 return expout;
1176 }
1177
1178 /* Parse STRING as an expression, and complain if this fails
1179 to use up all of the contents of STRING. */
1180
1181 struct expression *
1182 parse_expression (char *string)
1183 {
1184 register struct expression *exp;
1185 exp = parse_exp_1 (&string, 0, 0);
1186 if (*string)
1187 error ("Junk after end of expression.");
1188 return exp;
1189 }
1190 \f
1191 /* Stuff for maintaining a stack of types. Currently just used by C, but
1192 probably useful for any language which declares its types "backwards". */
1193
1194 static void
1195 check_type_stack_depth (void)
1196 {
1197 if (type_stack_depth == type_stack_size)
1198 {
1199 type_stack_size *= 2;
1200 type_stack = (union type_stack_elt *)
1201 xrealloc ((char *) type_stack, type_stack_size * sizeof (*type_stack));
1202 }
1203 }
1204
1205 void
1206 push_type (enum type_pieces tp)
1207 {
1208 check_type_stack_depth ();
1209 type_stack[type_stack_depth++].piece = tp;
1210 }
1211
1212 void
1213 push_type_int (int n)
1214 {
1215 check_type_stack_depth ();
1216 type_stack[type_stack_depth++].int_val = n;
1217 }
1218
1219 void
1220 push_type_address_space (char *string)
1221 {
1222 push_type_int (address_space_name_to_int (string));
1223 }
1224
1225 enum type_pieces
1226 pop_type (void)
1227 {
1228 if (type_stack_depth)
1229 return type_stack[--type_stack_depth].piece;
1230 return tp_end;
1231 }
1232
1233 int
1234 pop_type_int (void)
1235 {
1236 if (type_stack_depth)
1237 return type_stack[--type_stack_depth].int_val;
1238 /* "Can't happen". */
1239 return 0;
1240 }
1241
1242 /* Pop the type stack and return the type which corresponds to FOLLOW_TYPE
1243 as modified by all the stuff on the stack. */
1244 struct type *
1245 follow_types (struct type *follow_type)
1246 {
1247 int done = 0;
1248 int make_const = 0;
1249 int make_volatile = 0;
1250 int make_addr_space = 0;
1251 int array_size;
1252 struct type *range_type;
1253
1254 while (!done)
1255 switch (pop_type ())
1256 {
1257 case tp_end:
1258 done = 1;
1259 if (make_const)
1260 follow_type = make_cv_type (make_const,
1261 TYPE_VOLATILE (follow_type),
1262 follow_type, 0);
1263 if (make_volatile)
1264 follow_type = make_cv_type (TYPE_CONST (follow_type),
1265 make_volatile,
1266 follow_type, 0);
1267 if (make_addr_space)
1268 follow_type = make_type_with_address_space (follow_type,
1269 make_addr_space);
1270 make_const = make_volatile = 0;
1271 make_addr_space = 0;
1272 break;
1273 case tp_const:
1274 make_const = 1;
1275 break;
1276 case tp_volatile:
1277 make_volatile = 1;
1278 break;
1279 case tp_space_identifier:
1280 make_addr_space = pop_type_int ();
1281 break;
1282 case tp_pointer:
1283 follow_type = lookup_pointer_type (follow_type);
1284 if (make_const)
1285 follow_type = make_cv_type (make_const,
1286 TYPE_VOLATILE (follow_type),
1287 follow_type, 0);
1288 if (make_volatile)
1289 follow_type = make_cv_type (TYPE_CONST (follow_type),
1290 make_volatile,
1291 follow_type, 0);
1292 if (make_addr_space)
1293 follow_type = make_type_with_address_space (follow_type,
1294 make_addr_space);
1295 make_const = make_volatile = 0;
1296 make_addr_space = 0;
1297 break;
1298 case tp_reference:
1299 follow_type = lookup_reference_type (follow_type);
1300 if (make_const)
1301 follow_type = make_cv_type (make_const,
1302 TYPE_VOLATILE (follow_type),
1303 follow_type, 0);
1304 if (make_volatile)
1305 follow_type = make_cv_type (TYPE_CONST (follow_type),
1306 make_volatile,
1307 follow_type, 0);
1308 if (make_addr_space)
1309 follow_type = make_type_with_address_space (follow_type,
1310 make_addr_space);
1311 make_const = make_volatile = 0;
1312 make_addr_space = 0;
1313 break;
1314 case tp_array:
1315 array_size = pop_type_int ();
1316 /* FIXME-type-allocation: need a way to free this type when we are
1317 done with it. */
1318 range_type =
1319 create_range_type ((struct type *) NULL,
1320 builtin_type_int, 0,
1321 array_size >= 0 ? array_size - 1 : 0);
1322 follow_type =
1323 create_array_type ((struct type *) NULL,
1324 follow_type, range_type);
1325 if (array_size < 0)
1326 TYPE_ARRAY_UPPER_BOUND_TYPE (follow_type)
1327 = BOUND_CANNOT_BE_DETERMINED;
1328 break;
1329 case tp_function:
1330 /* FIXME-type-allocation: need a way to free this type when we are
1331 done with it. */
1332 follow_type = lookup_function_type (follow_type);
1333 break;
1334 }
1335 return follow_type;
1336 }
1337 \f
1338 static void build_parse (void);
1339 static void
1340 build_parse (void)
1341 {
1342 int i;
1343
1344 msym_text_symbol_type =
1345 init_type (TYPE_CODE_FUNC, 1, 0, "<text variable, no debug info>", NULL);
1346 TYPE_TARGET_TYPE (msym_text_symbol_type) = builtin_type_int;
1347 msym_data_symbol_type =
1348 init_type (TYPE_CODE_INT, TARGET_INT_BIT / HOST_CHAR_BIT, 0,
1349 "<data variable, no debug info>", NULL);
1350 msym_unknown_symbol_type =
1351 init_type (TYPE_CODE_INT, 1, 0,
1352 "<variable (not text or data), no debug info>",
1353 NULL);
1354
1355 /* create the std_regs table */
1356
1357 num_std_regs = 0;
1358 #ifdef PC_REGNUM
1359 if (PC_REGNUM >= 0)
1360 num_std_regs++;
1361 #endif
1362 #ifdef FP_REGNUM
1363 if (FP_REGNUM >= 0)
1364 num_std_regs++;
1365 #endif
1366 #ifdef SP_REGNUM
1367 if (SP_REGNUM >= 0)
1368 num_std_regs++;
1369 #endif
1370 #ifdef PS_REGNUM
1371 if (PS_REGNUM >= 0)
1372 num_std_regs++;
1373 #endif
1374 /* create an empty table */
1375 std_regs = xmalloc ((num_std_regs + 1) * sizeof *std_regs);
1376 i = 0;
1377 /* fill it in */
1378 #ifdef PC_REGNUM
1379 if (PC_REGNUM >= 0)
1380 {
1381 std_regs[i].name = "pc";
1382 std_regs[i].regnum = PC_REGNUM;
1383 i++;
1384 }
1385 #endif
1386 #ifdef FP_REGNUM
1387 if (FP_REGNUM >= 0)
1388 {
1389 std_regs[i].name = "fp";
1390 std_regs[i].regnum = FP_REGNUM;
1391 i++;
1392 }
1393 #endif
1394 #ifdef SP_REGNUM
1395 if (SP_REGNUM >= 0)
1396 {
1397 std_regs[i].name = "sp";
1398 std_regs[i].regnum = SP_REGNUM;
1399 i++;
1400 }
1401 #endif
1402 #ifdef PS_REGNUM
1403 if (PS_REGNUM >= 0)
1404 {
1405 std_regs[i].name = "ps";
1406 std_regs[i].regnum = PS_REGNUM;
1407 i++;
1408 }
1409 #endif
1410 memset (&std_regs[i], 0, sizeof (std_regs[i]));
1411 }
1412
1413 void
1414 _initialize_parse (void)
1415 {
1416 type_stack_size = 80;
1417 type_stack_depth = 0;
1418 type_stack = (union type_stack_elt *)
1419 xmalloc (type_stack_size * sizeof (*type_stack));
1420
1421 build_parse ();
1422
1423 /* FIXME - For the moment, handle types by swapping them in and out.
1424 Should be using the per-architecture data-pointer and a large
1425 struct. */
1426 register_gdbarch_swap (&msym_text_symbol_type, sizeof (msym_text_symbol_type), NULL);
1427 register_gdbarch_swap (&msym_data_symbol_type, sizeof (msym_data_symbol_type), NULL);
1428 register_gdbarch_swap (&msym_unknown_symbol_type, sizeof (msym_unknown_symbol_type), NULL);
1429
1430 register_gdbarch_swap (&num_std_regs, sizeof (std_regs), NULL);
1431 register_gdbarch_swap (&std_regs, sizeof (std_regs), NULL);
1432 register_gdbarch_swap (NULL, 0, build_parse);
1433
1434 add_show_from_set (
1435 add_set_cmd ("expression", class_maintenance, var_zinteger,
1436 (char *) &expressiondebug,
1437 "Set expression debugging.\n\
1438 When non-zero, the internal representation of expressions will be printed.",
1439 &setdebuglist),
1440 &showdebuglist);
1441 }
This page took 0.081276 seconds and 4 git commands to generate.