1 /* FLEX lexer for Ada expressions, for GDB. -*- c++ -*-
2 Copyright (C) 1994-2022 Free Software Foundation, Inc.
4 This file is part of GDB.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 3 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 /*----------------------------------------------------------------------*/
21 /* The converted version of this file is to be included in ada-exp.y, */
22 /* the Ada parser for gdb. The function yylex obtains characters from */
23 /* the global pointer lexptr. It returns a syntactic category for */
24 /* each successive token and places a semantic value into yylval */
25 /* (ada-lval), defined by the parser. */
28 NUM10 ({DIG}({DIG}|_)*)
30 NUM16 ({HEXDIG}({HEXDIG}|_)*)
33 ID ({LETTER}({LETTER}|{DIG}|[\x80-\xff])*|"<"{LETTER}({LETTER}|{DIG})*">")
36 GRAPHIC [a-z0-9 #&'()*+,-./:;<>=_|!$%?@\[\]\\^`{}~]
37 OPER ([-+*/=<>&]|"<="|">="|"**"|"/="|"and"|"or"|"xor"|"not"|"mod"|"rem"|"abs")
42 /* This must agree with COMPLETION_CHAR below. See the comment there
43 for the explanation. */
49 #include "diagnostics.h"
51 /* Some old versions of flex generate code that uses the "register" keyword,
52 which clang warns about. This was observed for example with flex 2.5.35,
53 as shipped with macOS 10.12. The same happens with flex 2.5.37 and g++ 11
54 which defaults to ISO C++17, that does not allow register storage class
57 DIAGNOSTIC_IGNORE_DEPRECATED_REGISTER
59 #define NUMERAL_WIDTH 256
60 #define LONGEST_SIGN ((ULONGEST) 1 << (sizeof(LONGEST) * HOST_CHAR_BIT - 1))
62 /* Temporary staging for numeric literals. */
63 static char numbuf[NUMERAL_WIDTH];
64 static void canonicalizeNumeral (char *s1, const char *);
65 static struct stoken processString (const char*, int);
66 static int processInt (struct parser_state *, const char *, const char *,
68 static int processReal (struct parser_state *, const char *);
69 static struct stoken processId (const char *, int);
70 static int processAttribute (const char *);
71 static int find_dot_all (const char *);
72 static void rewind_to_char (int);
75 #define YY_DECL static int yylex ( void )
77 /* Flex generates a static function "input" which is not used.
78 Defining YY_NO_INPUT comments it out. */
81 /* When completing, we'll return a special character at the end of the
82 input, to signal the completion position to the lexer. This is
83 done because flex does not have a generally useful way to detect
84 EOF in a pattern. This variable records whether the special
85 character has been emitted. */
86 static bool returned_complete = false;
88 /* The character we use to represent the completion point. */
89 #define COMPLETE_CHAR '\001'
92 #define YY_INPUT(BUF, RESULT, MAX_SIZE) \
93 if ( *pstate->lexptr == '\000' ) \
95 if (pstate->parse_completion && !returned_complete) \
97 returned_complete = true; \
98 *(BUF) = COMPLETE_CHAR; \
102 (RESULT) = YY_NULL; \
106 *(BUF) = *pstate->lexptr == COMPLETE_CHAR ? ' ' : *pstate->lexptr; \
108 pstate->lexptr += 1; \
111 /* Depth of parentheses. */
112 static int paren_depth;
116 %option case-insensitive interactive nodefault noyywrap
124 "--".* { yyterminate(); }
127 canonicalizeNumeral (numbuf, yytext);
128 char *e_ptr = strrchr (numbuf, 'e');
130 return processInt (pstate, nullptr, numbuf, e_ptr + 1);
134 canonicalizeNumeral (numbuf, yytext);
135 return processInt (pstate, NULL, numbuf, NULL);
138 {NUM10}"#"{HEXDIG}({HEXDIG}|_)*"#"{POSEXP} {
139 canonicalizeNumeral (numbuf, yytext);
140 char *e_ptr = strrchr (numbuf, 'e');
142 return processInt (pstate, numbuf,
143 strchr (numbuf, '#') + 1,
147 /* The "llf" is a gdb extension to allow a floating-point
148 constant to be written in some other base. The
149 floating-point number is formed by reinterpreting the
150 bytes, allowing direct control over the bits. */
151 {NUM10}(l{0,2}f)?"#"{HEXDIG}({HEXDIG}|_)*"#" {
152 canonicalizeNumeral (numbuf, yytext);
153 return processInt (pstate, numbuf, strchr (numbuf, '#') + 1,
158 canonicalizeNumeral (numbuf, yytext+2);
159 return processInt (pstate, "16#", numbuf, NULL);
163 {NUM10}"."{NUM10}{EXP} {
164 canonicalizeNumeral (numbuf, yytext);
165 return processReal (pstate, numbuf);
169 canonicalizeNumeral (numbuf, yytext);
170 return processReal (pstate, numbuf);
173 {NUM10}"#"{NUM16}"."{NUM16}"#"{EXP} {
174 error (_("Based real literals not implemented yet."));
177 {NUM10}"#"{NUM16}"."{NUM16}"#" {
178 error (_("Based real literals not implemented yet."));
181 <INITIAL>"'"({GRAPHIC}|\")"'" {
182 yylval.typed_val.val = yytext[1];
183 yylval.typed_val.type = type_for_char (pstate, yytext[1]);
187 <INITIAL>"'[\""{HEXDIG}{2,}"\"]'" {
188 ULONGEST v = strtoulst (yytext+3, nullptr, 16);
189 yylval.typed_val.val = v;
190 yylval.typed_val.type = type_for_char (pstate, v);
194 /* Note that we don't handle bracket sequences of more than 2
195 digits here. Currently there's no support for wide or
196 wide-wide strings. */
197 \"({GRAPHIC}|"[\""({HEXDIG}{2,}|\")"\"]")*\" {
198 yylval.sval = processString (yytext+1, yyleng-2);
203 error (_("ill-formed or non-terminated string literal"));
208 rewind_to_char ('i');
213 rewind_to_char ('t');
217 thread{WHITE}+{DIG} {
218 /* This keyword signals the end of the expression and
219 will be processed separately. */
220 rewind_to_char ('t');
227 and { return _AND_; }
228 else { return ELSE; }
233 null { return NULL_PTR; }
235 others { return OTHERS; }
237 then { return THEN; }
240 /* BOOLEAN "KEYWORDS" */
242 /* True and False are not keywords in Ada, but rather enumeration constants.
243 However, the boolean type is no longer represented as an enum, so True
244 and False are no longer defined in symbol tables. We compromise by
245 making them keywords (when bare). */
247 true { return TRUEKEYWORD; }
248 false { return FALSEKEYWORD; }
252 {TICK}([a-z][a-z_]*)?{COMPLETE}? { BEGIN INITIAL; return processAttribute (yytext); }
256 "=>" { return ARROW; }
257 ".." { return DOTDOT; }
258 "**" { return STARSTAR; }
259 ":=" { return ASSIGN; }
260 "/=" { return NOTEQUAL; }
264 <BEFORE_QUAL_QUOTE>"'"/{NOT_COMPLETE} { BEGIN INITIAL; return '\''; }
266 [-&*+{}@/:<>=|;\[\]] { return yytext[0]; }
268 "," { if (paren_depth == 0 && pstate->comma_terminates)
270 rewind_to_char (',');
277 "(" { paren_depth += 1; return '('; }
278 ")" { if (paren_depth == 0)
280 rewind_to_char (')');
290 "."{WHITE}*{ID}{COMPLETE}? {
291 yylval.sval = processId (yytext+1, yyleng-1);
292 if (yytext[yyleng - 1] == COMPLETE_CHAR)
297 "."{WHITE}*{COMPLETE} {
298 yylval.sval.ptr = "";
299 yylval.sval.length = 0;
303 {ID}({WHITE}*"."{WHITE}*({ID}|\"{OPER}\"))*(" "*"'"|{COMPLETE})? {
304 int all_posn = find_dot_all (yytext);
306 if (all_posn == -1 && yytext[yyleng-1] == '\'')
308 BEGIN BEFORE_QUAL_QUOTE;
311 else if (all_posn >= 0)
313 bool is_completion = yytext[yyleng - 1] == COMPLETE_CHAR;
314 yylval.sval = processId (yytext, yyleng);
315 return is_completion ? NAME_COMPLETE : NAME;
319 /* GDB EXPRESSION CONSTRUCTS */
321 "'"[^']+"'"{WHITE}*:: {
323 yylval.sval = processId (yytext, yyleng);
327 "::" { return COLONCOLON; }
329 /* REGISTERS AND GDB CONVENIENCE VARIABLES */
331 "$"({LETTER}|{DIG}|"$")* {
332 yylval.sval.ptr = yytext;
333 yylval.sval.length = yyleng;
334 return DOLLAR_VARIABLE;
337 /* CATCH-ALL ERROR CASE */
339 . { error (_("Invalid character '%s' in expression."), yytext); }
343 /* Initialize the lexer for processing new expression. */
346 lexer_init (FILE *inp)
350 returned_complete = false;
355 /* Copy S2 to S1, removing all underscores, and downcasing all letters. */
358 canonicalizeNumeral (char *s1, const char *s2)
360 for (; *s2 != '\000'; s2 += 1)
371 /* Interprets the prefix of NUM that consists of digits of the given BASE
372 as an integer of that BASE, with the string EXP as an exponent.
373 Puts value in yylval, and returns INT, if the string is valid. Causes
374 an error if the number is improperly formated. BASE, if NULL, defaults
375 to "10", and EXP to "1". The EXP does not contain a leading 'e' or 'E'.
379 processInt (struct parser_state *par_state, const char *base0,
380 const char *num0, const char *exp0)
384 /* For the based literal with an "f" prefix, we'll return a
385 floating-point number. This counts the the number of "l"s seen,
386 to decide the width of the floating-point number to return. -1
388 int floating_point_l_count = -1;
395 base = strtol (base0, &end_of_base, 10);
396 if (base < 2 || base > 16)
397 error (_("Invalid base: %d."), base);
398 while (*end_of_base == 'l')
400 ++floating_point_l_count;
403 /* This assertion is ensured by the pattern. */
404 gdb_assert (floating_point_l_count == -1 || *end_of_base == 'f');
405 if (*end_of_base == 'f')
408 ++floating_point_l_count;
410 /* This assertion is ensured by the pattern. */
411 gdb_assert (*end_of_base == '#');
417 exp = strtol(exp0, (char **) NULL, 10);
420 while (isxdigit (*num0))
422 int dig = fromhex (*num0);
424 error (_("Invalid digit `%c' in based literal"), *num0);
425 mpz_mul_ui (result.val, result.val, base);
426 mpz_add_ui (result.val, result.val, dig);
432 mpz_mul_ui (result.val, result.val, base);
436 if (floating_point_l_count > -1)
438 struct type *fp_type;
439 if (floating_point_l_count == 0)
440 fp_type = language_lookup_primitive_type (par_state->language (),
441 par_state->gdbarch (),
443 else if (floating_point_l_count == 1)
444 fp_type = language_lookup_primitive_type (par_state->language (),
445 par_state->gdbarch (),
449 /* This assertion is ensured by the pattern. */
450 gdb_assert (floating_point_l_count == 2);
451 fp_type = language_lookup_primitive_type (par_state->language (),
452 par_state->gdbarch (),
456 yylval.typed_val_float.type = fp_type;
457 result.write (gdb::make_array_view (yylval.typed_val_float.val,
458 TYPE_LENGTH (fp_type)),
459 type_byte_order (fp_type),
465 gdb_mpz maxval (ULONGEST_MAX);
466 if (mpz_cmp (result.val, maxval.val) > 0)
467 error (_("Integer literal out of range"));
469 int int_bits = gdbarch_int_bit (par_state->gdbarch ());
470 int long_bits = gdbarch_long_bit (par_state->gdbarch ());
471 int long_long_bits = gdbarch_long_long_bit (par_state->gdbarch ());
473 ULONGEST value = result.as_integer<ULONGEST> ();
474 if (fits_in_type (1, value, int_bits, true))
475 yylval.typed_val.type = type_int (par_state);
476 else if (fits_in_type (1, value, long_bits, true))
477 yylval.typed_val.type = type_long (par_state);
478 else if (fits_in_type (1, value, long_bits, false))
480 /* We have a number representable as an unsigned integer quantity.
481 For consistency with the C treatment, we will treat it as an
482 anonymous modular (unsigned) quantity. Alas, the types are such
483 that we need to store .val as a signed quantity. Sorry
484 for the mess, but C doesn't officially guarantee that a simple
485 assignment does the trick (no, it doesn't; read the reference manual).
487 yylval.typed_val.type
488 = builtin_type (par_state->gdbarch ())->builtin_unsigned_long;
489 if (value & LONGEST_SIGN)
490 yylval.typed_val.val =
491 (LONGEST) (value & ~LONGEST_SIGN)
492 - (LONGEST_SIGN>>1) - (LONGEST_SIGN>>1);
494 yylval.typed_val.val = (LONGEST) value;
497 else if (fits_in_type (1, value, long_long_bits, true))
498 yylval.typed_val.type = type_long_long (par_state);
499 else if (fits_in_type (1, value, long_long_bits, false))
501 /* Note: Interprets ULLONG_MAX as -1. */
502 yylval.typed_val.type = type_long_long (par_state);
503 /* See unsigned long case above. */
504 if (value & LONGEST_SIGN)
505 yylval.typed_val.val =
506 (LONGEST) (value & ~LONGEST_SIGN)
507 - (LONGEST_SIGN>>1) - (LONGEST_SIGN>>1);
509 yylval.typed_val.val = (LONGEST) value;
513 error (_("Integer literal out of range"));
515 yylval.typed_val.val = value;
520 processReal (struct parser_state *par_state, const char *num0)
522 yylval.typed_val_float.type = type_long_double (par_state);
524 bool parsed = parse_float (num0, strlen (num0),
525 yylval.typed_val_float.type,
526 yylval.typed_val_float.val);
532 /* Store a canonicalized version of NAME0[0..LEN-1] in yylval.ssym. The
533 resulting string is valid until the next call to ada_parse. If
534 NAME0 contains the substring "___", it is assumed to be already
535 encoded and the resulting name is equal to it. Similarly, if the name
536 starts with '<', it is copied verbatim. Otherwise, it differs
538 + Characters between '...' are transfered verbatim to yylval.ssym.
539 + Trailing "'" characters in quoted sequences are removed (a leading quote is
540 preserved to indicate that the name is not to be GNAT-encoded).
541 + Unquoted whitespace is removed.
542 + Unquoted alphabetic characters are mapped to lower case.
543 Result is returned as a struct stoken, but for convenience, the string
544 is also null-terminated. Result string valid until the next call of
548 processId (const char *name0, int len)
550 char *name = (char *) obstack_alloc (&temp_parse_space, len + 11);
552 struct stoken result;
555 while (len > 0 && isspace (name0[len-1]))
558 if (name0[0] == '<' || strstr (name0, "___") != NULL)
560 strncpy (name, name0, len);
566 bool in_quotes = false;
570 if (name0[i0] == COMPLETE_CHAR)
576 name[i++] = name0[i0++];
577 else if (isalnum (name0[i0]))
579 name[i] = tolower (name0[i0]);
582 else if (isspace (name0[i0]))
584 else if (name0[i0] == '\'')
586 /* Copy the starting quote, but not the ending quote. */
588 name[i++] = name0[i0++];
589 in_quotes = !in_quotes;
592 name[i++] = name0[i0++];
600 /* Return TEXT[0..LEN-1], a string literal without surrounding quotes,
601 with special hex character notations replaced with characters.
602 Result valid until the next call to ada_parse. */
605 processString (const char *text, int len)
609 const char *lim = text + len;
610 struct stoken result;
612 q = (char *) obstack_alloc (&temp_parse_space, len);
617 if (p[0] == '[' && p[1] == '"' && p+2 < lim)
619 if (p[2] == '"') /* "...["""]... */
627 ULONGEST chr = strtoulst (p + 2, &end, 16);
629 error (_("wide strings are not yet supported"));
639 result.length = q - result.ptr;
643 /* Returns the position within STR of the '.' in a
644 '.{WHITE}*all' component of a dotted name, or -1 if there is none.
645 Note: we actually don't need this routine, since 'all' can never be an
646 Ada identifier. Thus, looking up foo.all or foo.all.x as a name
647 must fail, and will eventually be interpreted as (foo).all or
648 (foo).all.x. However, this does avoid an extraneous lookup. */
651 find_dot_all (const char *str)
655 for (i = 0; str[i] != '\000'; i++)
662 while (isspace (str[i]));
664 if (strncasecmp (str + i, "all", 3) == 0
665 && !isalnum (str[i + 3]) && str[i + 3] != '_')
671 /* Returns non-zero iff string SUBSEQ matches a subsequence of STR, ignoring
675 subseqMatch (const char *subseq, const char *str)
677 if (subseq[0] == '\0')
679 else if (str[0] == '\0')
681 else if (tolower (subseq[0]) == tolower (str[0]))
682 return subseqMatch (subseq+1, str+1) || subseqMatch (subseq, str+1);
684 return subseqMatch (subseq, str+1);
688 static struct { const char *name; int code; }
690 { "address", TICK_ADDRESS },
691 { "unchecked_access", TICK_ACCESS },
692 { "unrestricted_access", TICK_ACCESS },
693 { "access", TICK_ACCESS },
694 { "first", TICK_FIRST },
695 { "last", TICK_LAST },
696 { "length", TICK_LENGTH },
699 { "modulus", TICK_MODULUS },
701 { "range", TICK_RANGE },
702 { "size", TICK_SIZE },
707 /* Return the syntactic code corresponding to the attribute name or
711 processAttribute (const char *str)
713 gdb_assert (*str == '\'');
715 while (isspace (*str))
718 int len = strlen (str);
719 if (len > 0 && str[len - 1] == COMPLETE_CHAR)
721 /* This is enforced by YY_INPUT. */
722 gdb_assert (pstate->parse_completion);
723 yylval.sval.ptr = obstack_strndup (&temp_parse_space, str, len - 1);
724 yylval.sval.length = len - 1;
725 return TICK_COMPLETE;
728 for (const auto &item : attributes)
729 if (strcasecmp (str, item.name) == 0)
732 gdb::optional<int> found;
733 for (const auto &item : attributes)
734 if (subseqMatch (str, item.name))
736 if (!found.has_value ())
739 error (_("ambiguous attribute name: `%s'"), str);
741 if (!found.has_value ())
742 error (_("unrecognized attribute: `%s'"), str);
748 ada_tick_completer::complete (struct expression *exp,
749 completion_tracker &tracker)
751 completion_list output;
752 for (const auto &item : attributes)
754 if (strncasecmp (item.name, m_name.c_str (), m_name.length ()) == 0)
755 output.emplace_back (xstrdup (item.name));
757 tracker.add_completions (std::move (output));
761 /* Back up lexptr by yyleng and then to the rightmost occurrence of
762 character CH, case-folded (there must be one). WARNING: since
763 lexptr points to the next input character that Flex has not yet
764 transferred to its internal buffer, the use of this function
765 depends on the assumption that Flex calls YY_INPUT only when it is
766 logically necessary to do so (thus, there is no reading ahead
767 farther than needed to identify the next token.) */
770 rewind_to_char (int ch)
772 pstate->lexptr -= yyleng;
773 while (toupper (*pstate->lexptr) != toupper (ch))
778 /* Dummy definition to suppress warnings about unused static definitions. */
779 typedef void (*dummy_function) ();
780 dummy_function ada_flex_use[] =
782 (dummy_function) yyunput