xref: /netbsd-src/external/gpl3/gdb.old/dist/gdb/rust-parse.c (revision f8cf1a9151c7af1cb0bd8b09c13c66bca599c027)
1 /* Rust expression parsing for GDB, the GNU debugger.
2 
3    Copyright (C) 2016-2023 Free Software Foundation, Inc.
4 
5    This file is part of GDB.
6 
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.
11 
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.
16 
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/>.  */
19 
20 #include "defs.h"
21 
22 #include "block.h"
23 #include "charset.h"
24 #include "cp-support.h"
25 #include "gdbsupport/gdb_obstack.h"
26 #include "gdbsupport/gdb_regex.h"
27 #include "rust-lang.h"
28 #include "parser-defs.h"
29 #include "gdbsupport/selftest.h"
30 #include "value.h"
31 #include "gdbarch.h"
32 #include "rust-exp.h"
33 
34 using namespace expr;
35 
36 /* A regular expression for matching Rust numbers.  This is split up
37    since it is very long and this gives us a way to comment the
38    sections.  */
39 
40 static const char number_regex_text[] =
41   /* subexpression 1: allows use of alternation, otherwise uninteresting */
42   "^("
43   /* First comes floating point.  */
44   /* Recognize number after the decimal point, with optional
45      exponent and optional type suffix.
46      subexpression 2: allows "?", otherwise uninteresting
47      subexpression 3: if present, type suffix
48   */
49   "[0-9][0-9_]*\\.[0-9][0-9_]*([eE][-+]?[0-9][0-9_]*)?(f32|f64)?"
50 #define FLOAT_TYPE1 3
51   "|"
52   /* Recognize exponent without decimal point, with optional type
53      suffix.
54      subexpression 4: if present, type suffix
55   */
56 #define FLOAT_TYPE2 4
57   "[0-9][0-9_]*[eE][-+]?[0-9][0-9_]*(f32|f64)?"
58   "|"
59   /* "23." is a valid floating point number, but "23.e5" and
60      "23.f32" are not.  So, handle the trailing-. case
61      separately.  */
62   "[0-9][0-9_]*\\."
63   "|"
64   /* Finally come integers.
65      subexpression 5: text of integer
66      subexpression 6: if present, type suffix
67      subexpression 7: allows use of alternation, otherwise uninteresting
68   */
69 #define INT_TEXT 5
70 #define INT_TYPE 6
71   "(0x[a-fA-F0-9_]+|0o[0-7_]+|0b[01_]+|[0-9][0-9_]*)"
72   "([iu](size|8|16|32|64))?"
73   ")";
74 /* The number of subexpressions to allocate space for, including the
75    "0th" whole match subexpression.  */
76 #define NUM_SUBEXPRESSIONS 8
77 
78 /* The compiled number-matching regex.  */
79 
80 static regex_t number_regex;
81 
82 /* The kinds of tokens.  Note that single-character tokens are
83    represented by themselves, so for instance '[' is a token.  */
84 enum token_type : int
85 {
86   /* Make sure to start after any ASCII character.  */
87   GDBVAR = 256,
88   IDENT,
89   COMPLETE,
90   INTEGER,
91   DECIMAL_INTEGER,
92   STRING,
93   BYTESTRING,
94   FLOAT,
95   COMPOUND_ASSIGN,
96 
97   /* Keyword tokens.  */
98   KW_AS,
99   KW_IF,
100   KW_TRUE,
101   KW_FALSE,
102   KW_SUPER,
103   KW_SELF,
104   KW_MUT,
105   KW_EXTERN,
106   KW_CONST,
107   KW_FN,
108   KW_SIZEOF,
109 
110   /* Operator tokens.  */
111   DOTDOT,
112   DOTDOTEQ,
113   OROR,
114   ANDAND,
115   EQEQ,
116   NOTEQ,
117   LTEQ,
118   GTEQ,
119   LSH,
120   RSH,
121   COLONCOLON,
122   ARROW,
123 };
124 
125 /* A typed integer constant.  */
126 
127 struct typed_val_int
128 {
129   ULONGEST val;
130   struct type *type;
131 };
132 
133 /* A typed floating point constant.  */
134 
135 struct typed_val_float
136 {
137   float_data val;
138   struct type *type;
139 };
140 
141 /* A struct of this type is used to describe a token.  */
142 
143 struct token_info
144 {
145   const char *name;
146   int value;
147   enum exp_opcode opcode;
148 };
149 
150 /* Identifier tokens.  */
151 
152 static const struct token_info identifier_tokens[] =
153 {
154   { "as", KW_AS, OP_NULL },
155   { "false", KW_FALSE, OP_NULL },
156   { "if", 0, OP_NULL },
157   { "mut", KW_MUT, OP_NULL },
158   { "const", KW_CONST, OP_NULL },
159   { "self", KW_SELF, OP_NULL },
160   { "super", KW_SUPER, OP_NULL },
161   { "true", KW_TRUE, OP_NULL },
162   { "extern", KW_EXTERN, OP_NULL },
163   { "fn", KW_FN, OP_NULL },
164   { "sizeof", KW_SIZEOF, OP_NULL },
165 };
166 
167 /* Operator tokens, sorted longest first.  */
168 
169 static const struct token_info operator_tokens[] =
170 {
171   { ">>=", COMPOUND_ASSIGN, BINOP_RSH },
172   { "<<=", COMPOUND_ASSIGN, BINOP_LSH },
173 
174   { "<<", LSH, OP_NULL },
175   { ">>", RSH, OP_NULL },
176   { "&&", ANDAND, OP_NULL },
177   { "||", OROR, OP_NULL },
178   { "==", EQEQ, OP_NULL },
179   { "!=", NOTEQ, OP_NULL },
180   { "<=", LTEQ, OP_NULL },
181   { ">=", GTEQ, OP_NULL },
182   { "+=", COMPOUND_ASSIGN, BINOP_ADD },
183   { "-=", COMPOUND_ASSIGN, BINOP_SUB },
184   { "*=", COMPOUND_ASSIGN, BINOP_MUL },
185   { "/=", COMPOUND_ASSIGN, BINOP_DIV },
186   { "%=", COMPOUND_ASSIGN, BINOP_REM },
187   { "&=", COMPOUND_ASSIGN, BINOP_BITWISE_AND },
188   { "|=", COMPOUND_ASSIGN, BINOP_BITWISE_IOR },
189   { "^=", COMPOUND_ASSIGN, BINOP_BITWISE_XOR },
190   { "..=", DOTDOTEQ, OP_NULL },
191 
192   { "::", COLONCOLON, OP_NULL },
193   { "..", DOTDOT, OP_NULL },
194   { "->", ARROW, OP_NULL }
195 };
196 
197 /* An instance of this is created before parsing, and destroyed when
198    parsing is finished.  */
199 
200 struct rust_parser
201 {
202   explicit rust_parser (struct parser_state *state)
203     : pstate (state)
204   {
205   }
206 
207   DISABLE_COPY_AND_ASSIGN (rust_parser);
208 
209   /* Return the parser's language.  */
210   const struct language_defn *language () const
211   {
212     return pstate->language ();
213   }
214 
215   /* Return the parser's gdbarch.  */
216   struct gdbarch *arch () const
217   {
218     return pstate->gdbarch ();
219   }
220 
221   /* A helper to look up a Rust type, or fail.  This only works for
222      types defined by rust_language_arch_info.  */
223 
224   struct type *get_type (const char *name)
225   {
226     struct type *type;
227 
228     type = language_lookup_primitive_type (language (), arch (), name);
229     if (type == NULL)
230       error (_("Could not find Rust type %s"), name);
231     return type;
232   }
233 
234   std::string crate_name (const std::string &name);
235   std::string super_name (const std::string &ident, unsigned int n_supers);
236 
237   int lex_character ();
238   int lex_number ();
239   int lex_string ();
240   int lex_identifier ();
241   uint32_t lex_hex (int min, int max);
242   uint32_t lex_escape (int is_byte);
243   int lex_operator ();
244   int lex_one_token ();
245   void push_back (char c);
246 
247   /* The main interface to lexing.  Lexes one token and updates the
248      internal state.  */
249   void lex ()
250   {
251     current_token = lex_one_token ();
252   }
253 
254   /* Assuming the current token is TYPE, lex the next token.  */
255   void assume (int type)
256   {
257     gdb_assert (current_token == type);
258     lex ();
259   }
260 
261   /* Require the single-character token C, and lex the next token; or
262      throw an exception.  */
263   void require (char type)
264   {
265     if (current_token != type)
266       error (_("'%c' expected"), type);
267     lex ();
268   }
269 
270   /* Entry point for all parsing.  */
271   operation_up parse_entry_point ()
272   {
273     lex ();
274     operation_up result = parse_expr ();
275     if (current_token != 0)
276       error (_("Syntax error near '%s'"), pstate->prev_lexptr);
277     return result;
278   }
279 
280   operation_up parse_tuple ();
281   operation_up parse_array ();
282   operation_up name_to_operation (const std::string &name);
283   operation_up parse_struct_expr (struct type *type);
284   operation_up parse_binop (bool required);
285   operation_up parse_range ();
286   operation_up parse_expr ();
287   operation_up parse_sizeof ();
288   operation_up parse_addr ();
289   operation_up parse_field (operation_up &&);
290   operation_up parse_index (operation_up &&);
291   std::vector<operation_up> parse_paren_args ();
292   operation_up parse_call (operation_up &&);
293   std::vector<struct type *> parse_type_list ();
294   std::vector<struct type *> parse_maybe_type_list ();
295   struct type *parse_array_type ();
296   struct type *parse_slice_type ();
297   struct type *parse_pointer_type ();
298   struct type *parse_function_type ();
299   struct type *parse_tuple_type ();
300   struct type *parse_type ();
301   std::string parse_path (bool for_expr);
302   operation_up parse_string ();
303   operation_up parse_tuple_struct (struct type *type);
304   operation_up parse_path_expr ();
305   operation_up parse_atom (bool required);
306 
307   void update_innermost_block (struct block_symbol sym);
308   struct block_symbol lookup_symbol (const char *name,
309 				     const struct block *block,
310 				     const domain_enum domain);
311   struct type *rust_lookup_type (const char *name);
312 
313   /* Clear some state.  This is only used for testing.  */
314 #if GDB_SELF_TEST
315   void reset (const char *input)
316   {
317     pstate->prev_lexptr = nullptr;
318     pstate->lexptr = input;
319     paren_depth = 0;
320     current_token = 0;
321     current_int_val = {};
322     current_float_val = {};
323     current_string_val = {};
324     current_opcode = OP_NULL;
325   }
326 #endif /* GDB_SELF_TEST */
327 
328   /* Return the token's string value as a string.  */
329   std::string get_string () const
330   {
331     return std::string (current_string_val.ptr, current_string_val.length);
332   }
333 
334   /* A pointer to this is installed globally.  */
335   auto_obstack obstack;
336 
337   /* The parser state gdb gave us.  */
338   struct parser_state *pstate;
339 
340   /* Depth of parentheses.  */
341   int paren_depth = 0;
342 
343   /* The current token's type.  */
344   int current_token = 0;
345   /* The current token's payload, if any.  */
346   typed_val_int current_int_val {};
347   typed_val_float current_float_val {};
348   struct stoken current_string_val {};
349   enum exp_opcode current_opcode = OP_NULL;
350 
351   /* When completing, this may be set to the field operation to
352      complete.  */
353   operation_up completion_op;
354 };
355 
356 /* Return an string referring to NAME, but relative to the crate's
357    name.  */
358 
359 std::string
360 rust_parser::crate_name (const std::string &name)
361 {
362   std::string crate = rust_crate_for_block (pstate->expression_context_block);
363 
364   if (crate.empty ())
365     error (_("Could not find crate for current location"));
366   return "::" + crate + "::" + name;
367 }
368 
369 /* Return a string referring to a "super::" qualified name.  IDENT is
370    the base name and N_SUPERS is how many "super::"s were provided.
371    N_SUPERS can be zero.  */
372 
373 std::string
374 rust_parser::super_name (const std::string &ident, unsigned int n_supers)
375 {
376   const char *scope = block_scope (pstate->expression_context_block);
377   int offset;
378 
379   if (scope[0] == '\0')
380     error (_("Couldn't find namespace scope for self::"));
381 
382   if (n_supers > 0)
383     {
384       int len;
385       std::vector<int> offsets;
386       unsigned int current_len;
387 
388       current_len = cp_find_first_component (scope);
389       while (scope[current_len] != '\0')
390 	{
391 	  offsets.push_back (current_len);
392 	  gdb_assert (scope[current_len] == ':');
393 	  /* The "::".  */
394 	  current_len += 2;
395 	  current_len += cp_find_first_component (scope
396 						  + current_len);
397 	}
398 
399       len = offsets.size ();
400       if (n_supers >= len)
401 	error (_("Too many super:: uses from '%s'"), scope);
402 
403       offset = offsets[len - n_supers];
404     }
405   else
406     offset = strlen (scope);
407 
408   return "::" + std::string (scope, offset) + "::" + ident;
409 }
410 
411 /* A helper to appropriately munge NAME and BLOCK depending on the
412    presence of a leading "::".  */
413 
414 static void
415 munge_name_and_block (const char **name, const struct block **block)
416 {
417   /* If it is a global reference, skip the current block in favor of
418      the static block.  */
419   if (startswith (*name, "::"))
420     {
421       *name += 2;
422       *block = block_static_block (*block);
423     }
424 }
425 
426 /* Like lookup_symbol, but handles Rust namespace conventions, and
427    doesn't require field_of_this_result.  */
428 
429 struct block_symbol
430 rust_parser::lookup_symbol (const char *name, const struct block *block,
431 			    const domain_enum domain)
432 {
433   struct block_symbol result;
434 
435   munge_name_and_block (&name, &block);
436 
437   result = ::lookup_symbol (name, block, domain, NULL);
438   if (result.symbol != NULL)
439     update_innermost_block (result);
440   return result;
441 }
442 
443 /* Look up a type, following Rust namespace conventions.  */
444 
445 struct type *
446 rust_parser::rust_lookup_type (const char *name)
447 {
448   struct block_symbol result;
449   struct type *type;
450 
451   const struct block *block = pstate->expression_context_block;
452   munge_name_and_block (&name, &block);
453 
454   result = ::lookup_symbol (name, block, STRUCT_DOMAIN, NULL);
455   if (result.symbol != NULL)
456     {
457       update_innermost_block (result);
458       return result.symbol->type ();
459     }
460 
461   type = lookup_typename (language (), name, NULL, 1);
462   if (type != NULL)
463     return type;
464 
465   /* Last chance, try a built-in type.  */
466   return language_lookup_primitive_type (language (), arch (), name);
467 }
468 
469 /* A helper that updates the innermost block as appropriate.  */
470 
471 void
472 rust_parser::update_innermost_block (struct block_symbol sym)
473 {
474   if (symbol_read_needs_frame (sym.symbol))
475     pstate->block_tracker->update (sym);
476 }
477 
478 /* Lex a hex number with at least MIN digits and at most MAX
479    digits.  */
480 
481 uint32_t
482 rust_parser::lex_hex (int min, int max)
483 {
484   uint32_t result = 0;
485   int len = 0;
486   /* We only want to stop at MAX if we're lexing a byte escape.  */
487   int check_max = min == max;
488 
489   while ((check_max ? len <= max : 1)
490 	 && ((pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'f')
491 	     || (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'F')
492 	     || (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9')))
493     {
494       result *= 16;
495       if (pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'f')
496 	result = result + 10 + pstate->lexptr[0] - 'a';
497       else if (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'F')
498 	result = result + 10 + pstate->lexptr[0] - 'A';
499       else
500 	result = result + pstate->lexptr[0] - '0';
501       ++pstate->lexptr;
502       ++len;
503     }
504 
505   if (len < min)
506     error (_("Not enough hex digits seen"));
507   if (len > max)
508     {
509       gdb_assert (min != max);
510       error (_("Overlong hex escape"));
511     }
512 
513   return result;
514 }
515 
516 /* Lex an escape.  IS_BYTE is true if we're lexing a byte escape;
517    otherwise we're lexing a character escape.  */
518 
519 uint32_t
520 rust_parser::lex_escape (int is_byte)
521 {
522   uint32_t result;
523 
524   gdb_assert (pstate->lexptr[0] == '\\');
525   ++pstate->lexptr;
526   switch (pstate->lexptr[0])
527     {
528     case 'x':
529       ++pstate->lexptr;
530       result = lex_hex (2, 2);
531       break;
532 
533     case 'u':
534       if (is_byte)
535 	error (_("Unicode escape in byte literal"));
536       ++pstate->lexptr;
537       if (pstate->lexptr[0] != '{')
538 	error (_("Missing '{' in Unicode escape"));
539       ++pstate->lexptr;
540       result = lex_hex (1, 6);
541       /* Could do range checks here.  */
542       if (pstate->lexptr[0] != '}')
543 	error (_("Missing '}' in Unicode escape"));
544       ++pstate->lexptr;
545       break;
546 
547     case 'n':
548       result = '\n';
549       ++pstate->lexptr;
550       break;
551     case 'r':
552       result = '\r';
553       ++pstate->lexptr;
554       break;
555     case 't':
556       result = '\t';
557       ++pstate->lexptr;
558       break;
559     case '\\':
560       result = '\\';
561       ++pstate->lexptr;
562       break;
563     case '0':
564       result = '\0';
565       ++pstate->lexptr;
566       break;
567     case '\'':
568       result = '\'';
569       ++pstate->lexptr;
570       break;
571     case '"':
572       result = '"';
573       ++pstate->lexptr;
574       break;
575 
576     default:
577       error (_("Invalid escape \\%c in literal"), pstate->lexptr[0]);
578     }
579 
580   return result;
581 }
582 
583 /* A helper for lex_character.  Search forward for the closing single
584    quote, then convert the bytes from the host charset to UTF-32.  */
585 
586 static uint32_t
587 lex_multibyte_char (const char *text, int *len)
588 {
589   /* Only look a maximum of 5 bytes for the closing quote.  This is
590      the maximum for UTF-8.  */
591   int quote;
592   gdb_assert (text[0] != '\'');
593   for (quote = 1; text[quote] != '\0' && text[quote] != '\''; ++quote)
594     ;
595   *len = quote;
596   /* The caller will issue an error.  */
597   if (text[quote] == '\0')
598     return 0;
599 
600   auto_obstack result;
601   convert_between_encodings (host_charset (), HOST_UTF32,
602 			     (const gdb_byte *) text,
603 			     quote, 1, &result, translit_none);
604 
605   int size = obstack_object_size (&result);
606   if (size > 4)
607     error (_("overlong character literal"));
608   uint32_t value;
609   memcpy (&value, obstack_finish (&result), size);
610   return value;
611 }
612 
613 /* Lex a character constant.  */
614 
615 int
616 rust_parser::lex_character ()
617 {
618   int is_byte = 0;
619   uint32_t value;
620 
621   if (pstate->lexptr[0] == 'b')
622     {
623       is_byte = 1;
624       ++pstate->lexptr;
625     }
626   gdb_assert (pstate->lexptr[0] == '\'');
627   ++pstate->lexptr;
628   if (pstate->lexptr[0] == '\'')
629     error (_("empty character literal"));
630   else if (pstate->lexptr[0] == '\\')
631     value = lex_escape (is_byte);
632   else
633     {
634       int len;
635       value = lex_multibyte_char (&pstate->lexptr[0], &len);
636       pstate->lexptr += len;
637     }
638 
639   if (pstate->lexptr[0] != '\'')
640     error (_("Unterminated character literal"));
641   ++pstate->lexptr;
642 
643   current_int_val.val = value;
644   current_int_val.type = get_type (is_byte ? "u8" : "char");
645 
646   return INTEGER;
647 }
648 
649 /* Return the offset of the double quote if STR looks like the start
650    of a raw string, or 0 if STR does not start a raw string.  */
651 
652 static int
653 starts_raw_string (const char *str)
654 {
655   const char *save = str;
656 
657   if (str[0] != 'r')
658     return 0;
659   ++str;
660   while (str[0] == '#')
661     ++str;
662   if (str[0] == '"')
663     return str - save;
664   return 0;
665 }
666 
667 /* Return true if STR looks like the end of a raw string that had N
668    hashes at the start.  */
669 
670 static bool
671 ends_raw_string (const char *str, int n)
672 {
673   int i;
674 
675   gdb_assert (str[0] == '"');
676   for (i = 0; i < n; ++i)
677     if (str[i + 1] != '#')
678       return false;
679   return true;
680 }
681 
682 /* Lex a string constant.  */
683 
684 int
685 rust_parser::lex_string ()
686 {
687   int is_byte = pstate->lexptr[0] == 'b';
688   int raw_length;
689 
690   if (is_byte)
691     ++pstate->lexptr;
692   raw_length = starts_raw_string (pstate->lexptr);
693   pstate->lexptr += raw_length;
694   gdb_assert (pstate->lexptr[0] == '"');
695   ++pstate->lexptr;
696 
697   while (1)
698     {
699       uint32_t value;
700 
701       if (raw_length > 0)
702 	{
703 	  if (pstate->lexptr[0] == '"' && ends_raw_string (pstate->lexptr,
704 							   raw_length - 1))
705 	    {
706 	      /* Exit with lexptr pointing after the final "#".  */
707 	      pstate->lexptr += raw_length;
708 	      break;
709 	    }
710 	  else if (pstate->lexptr[0] == '\0')
711 	    error (_("Unexpected EOF in string"));
712 
713 	  value = pstate->lexptr[0] & 0xff;
714 	  if (is_byte && value > 127)
715 	    error (_("Non-ASCII value in raw byte string"));
716 	  obstack_1grow (&obstack, value);
717 
718 	  ++pstate->lexptr;
719 	}
720       else if (pstate->lexptr[0] == '"')
721 	{
722 	  /* Make sure to skip the quote.  */
723 	  ++pstate->lexptr;
724 	  break;
725 	}
726       else if (pstate->lexptr[0] == '\\')
727 	{
728 	  value = lex_escape (is_byte);
729 
730 	  if (is_byte)
731 	    obstack_1grow (&obstack, value);
732 	  else
733 	    convert_between_encodings (HOST_UTF32, "UTF-8",
734 				       (gdb_byte *) &value,
735 				       sizeof (value), sizeof (value),
736 				       &obstack, translit_none);
737 	}
738       else if (pstate->lexptr[0] == '\0')
739 	error (_("Unexpected EOF in string"));
740       else
741 	{
742 	  value = pstate->lexptr[0] & 0xff;
743 	  if (is_byte && value > 127)
744 	    error (_("Non-ASCII value in byte string"));
745 	  obstack_1grow (&obstack, value);
746 	  ++pstate->lexptr;
747 	}
748     }
749 
750   current_string_val.length = obstack_object_size (&obstack);
751   current_string_val.ptr = (const char *) obstack_finish (&obstack);
752   return is_byte ? BYTESTRING : STRING;
753 }
754 
755 /* Return true if STRING starts with whitespace followed by a digit.  */
756 
757 static bool
758 space_then_number (const char *string)
759 {
760   const char *p = string;
761 
762   while (p[0] == ' ' || p[0] == '\t')
763     ++p;
764   if (p == string)
765     return false;
766 
767   return *p >= '0' && *p <= '9';
768 }
769 
770 /* Return true if C can start an identifier.  */
771 
772 static bool
773 rust_identifier_start_p (char c)
774 {
775   return ((c >= 'a' && c <= 'z')
776 	  || (c >= 'A' && c <= 'Z')
777 	  || c == '_'
778 	  || c == '$'
779 	  /* Allow any non-ASCII character as an identifier.  There
780 	     doesn't seem to be a need to be picky about this.  */
781 	  || (c & 0x80) != 0);
782 }
783 
784 /* Lex an identifier.  */
785 
786 int
787 rust_parser::lex_identifier ()
788 {
789   unsigned int length;
790   const struct token_info *token;
791   int is_gdb_var = pstate->lexptr[0] == '$';
792 
793   bool is_raw = false;
794   if (pstate->lexptr[0] == 'r'
795       && pstate->lexptr[1] == '#'
796       && rust_identifier_start_p (pstate->lexptr[2]))
797     {
798       is_raw = true;
799       pstate->lexptr += 2;
800     }
801 
802   const char *start = pstate->lexptr;
803   gdb_assert (rust_identifier_start_p (pstate->lexptr[0]));
804 
805   ++pstate->lexptr;
806 
807   /* Allow any non-ASCII character here.  This "handles" UTF-8 by
808      passing it through.  */
809   while ((pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'z')
810 	 || (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'Z')
811 	 || pstate->lexptr[0] == '_'
812 	 || (is_gdb_var && pstate->lexptr[0] == '$')
813 	 || (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9')
814 	 || (pstate->lexptr[0] & 0x80) != 0)
815     ++pstate->lexptr;
816 
817 
818   length = pstate->lexptr - start;
819   token = NULL;
820   if (!is_raw)
821     {
822       for (const auto &candidate : identifier_tokens)
823 	{
824 	  if (length == strlen (candidate.name)
825 	      && strncmp (candidate.name, start, length) == 0)
826 	    {
827 	      token = &candidate;
828 	      break;
829 	    }
830 	}
831     }
832 
833   if (token != NULL)
834     {
835       if (token->value == 0)
836 	{
837 	  /* Leave the terminating token alone.  */
838 	  pstate->lexptr = start;
839 	  return 0;
840 	}
841     }
842   else if (token == NULL
843 	   && !is_raw
844 	   && (strncmp (start, "thread", length) == 0
845 	       || strncmp (start, "task", length) == 0)
846 	   && space_then_number (pstate->lexptr))
847     {
848       /* "task" or "thread" followed by a number terminates the
849 	 parse, per gdb rules.  */
850       pstate->lexptr = start;
851       return 0;
852     }
853 
854   if (token == NULL || (pstate->parse_completion && pstate->lexptr[0] == '\0'))
855     {
856       current_string_val.length = length;
857       current_string_val.ptr = start;
858     }
859 
860   if (pstate->parse_completion && pstate->lexptr[0] == '\0')
861     {
862       /* Prevent rustyylex from returning two COMPLETE tokens.  */
863       pstate->prev_lexptr = pstate->lexptr;
864       return COMPLETE;
865     }
866 
867   if (token != NULL)
868     return token->value;
869   if (is_gdb_var)
870     return GDBVAR;
871   return IDENT;
872 }
873 
874 /* Lex an operator.  */
875 
876 int
877 rust_parser::lex_operator ()
878 {
879   const struct token_info *token = NULL;
880 
881   for (const auto &candidate : operator_tokens)
882     {
883       if (strncmp (candidate.name, pstate->lexptr,
884 		   strlen (candidate.name)) == 0)
885 	{
886 	  pstate->lexptr += strlen (candidate.name);
887 	  token = &candidate;
888 	  break;
889 	}
890     }
891 
892   if (token != NULL)
893     {
894       current_opcode = token->opcode;
895       return token->value;
896     }
897 
898   return *pstate->lexptr++;
899 }
900 
901 /* Lex a number.  */
902 
903 int
904 rust_parser::lex_number ()
905 {
906   regmatch_t subexps[NUM_SUBEXPRESSIONS];
907   int match;
908   int is_integer = 0;
909   int could_be_decimal = 1;
910   int implicit_i32 = 0;
911   const char *type_name = NULL;
912   struct type *type;
913   int end_index;
914   int type_index = -1;
915   int i;
916 
917   match = regexec (&number_regex, pstate->lexptr, ARRAY_SIZE (subexps),
918 		   subexps, 0);
919   /* Failure means the regexp is broken.  */
920   gdb_assert (match == 0);
921 
922   if (subexps[INT_TEXT].rm_so != -1)
923     {
924       /* Integer part matched.  */
925       is_integer = 1;
926       end_index = subexps[INT_TEXT].rm_eo;
927       if (subexps[INT_TYPE].rm_so == -1)
928 	{
929 	  type_name = "i32";
930 	  implicit_i32 = 1;
931 	}
932       else
933 	{
934 	  type_index = INT_TYPE;
935 	  could_be_decimal = 0;
936 	}
937     }
938   else if (subexps[FLOAT_TYPE1].rm_so != -1)
939     {
940       /* Found floating point type suffix.  */
941       end_index = subexps[FLOAT_TYPE1].rm_so;
942       type_index = FLOAT_TYPE1;
943     }
944   else if (subexps[FLOAT_TYPE2].rm_so != -1)
945     {
946       /* Found floating point type suffix.  */
947       end_index = subexps[FLOAT_TYPE2].rm_so;
948       type_index = FLOAT_TYPE2;
949     }
950   else
951     {
952       /* Any other floating point match.  */
953       end_index = subexps[0].rm_eo;
954       type_name = "f64";
955     }
956 
957   /* We need a special case if the final character is ".".  In this
958      case we might need to parse an integer.  For example, "23.f()" is
959      a request for a trait method call, not a syntax error involving
960      the floating point number "23.".  */
961   gdb_assert (subexps[0].rm_eo > 0);
962   if (pstate->lexptr[subexps[0].rm_eo - 1] == '.')
963     {
964       const char *next = skip_spaces (&pstate->lexptr[subexps[0].rm_eo]);
965 
966       if (rust_identifier_start_p (*next) || *next == '.')
967 	{
968 	  --subexps[0].rm_eo;
969 	  is_integer = 1;
970 	  end_index = subexps[0].rm_eo;
971 	  type_name = "i32";
972 	  could_be_decimal = 1;
973 	  implicit_i32 = 1;
974 	}
975     }
976 
977   /* Compute the type name if we haven't already.  */
978   std::string type_name_holder;
979   if (type_name == NULL)
980     {
981       gdb_assert (type_index != -1);
982       type_name_holder = std::string ((pstate->lexptr
983 				       + subexps[type_index].rm_so),
984 				      (subexps[type_index].rm_eo
985 				       - subexps[type_index].rm_so));
986       type_name = type_name_holder.c_str ();
987     }
988 
989   /* Look up the type.  */
990   type = get_type (type_name);
991 
992   /* Copy the text of the number and remove the "_"s.  */
993   std::string number;
994   for (i = 0; i < end_index && pstate->lexptr[i]; ++i)
995     {
996       if (pstate->lexptr[i] == '_')
997 	could_be_decimal = 0;
998       else
999 	number.push_back (pstate->lexptr[i]);
1000     }
1001 
1002   /* Advance past the match.  */
1003   pstate->lexptr += subexps[0].rm_eo;
1004 
1005   /* Parse the number.  */
1006   if (is_integer)
1007     {
1008       uint64_t value;
1009       int radix = 10;
1010       int offset = 0;
1011 
1012       if (number[0] == '0')
1013 	{
1014 	  if (number[1] == 'x')
1015 	    radix = 16;
1016 	  else if (number[1] == 'o')
1017 	    radix = 8;
1018 	  else if (number[1] == 'b')
1019 	    radix = 2;
1020 	  if (radix != 10)
1021 	    {
1022 	      offset = 2;
1023 	      could_be_decimal = 0;
1024 	    }
1025 	}
1026 
1027       const char *trailer;
1028       value = strtoulst (number.c_str () + offset, &trailer, radix);
1029       if (*trailer != '\0')
1030 	error (_("Integer literal is too large"));
1031       if (implicit_i32 && value >= ((uint64_t) 1) << 31)
1032 	type = get_type ("i64");
1033 
1034       current_int_val.val = value;
1035       current_int_val.type = type;
1036     }
1037   else
1038     {
1039       current_float_val.type = type;
1040       bool parsed = parse_float (number.c_str (), number.length (),
1041 				 current_float_val.type,
1042 				 current_float_val.val.data ());
1043       gdb_assert (parsed);
1044     }
1045 
1046   return is_integer ? (could_be_decimal ? DECIMAL_INTEGER : INTEGER) : FLOAT;
1047 }
1048 
1049 /* The lexer.  */
1050 
1051 int
1052 rust_parser::lex_one_token ()
1053 {
1054   /* Skip all leading whitespace.  */
1055   while (pstate->lexptr[0] == ' '
1056 	 || pstate->lexptr[0] == '\t'
1057 	 || pstate->lexptr[0] == '\r'
1058 	 || pstate->lexptr[0] == '\n')
1059     ++pstate->lexptr;
1060 
1061   /* If we hit EOF and we're completing, then return COMPLETE -- maybe
1062      we're completing an empty string at the end of a field_expr.
1063      But, we don't want to return two COMPLETE tokens in a row.  */
1064   if (pstate->lexptr[0] == '\0' && pstate->lexptr == pstate->prev_lexptr)
1065     return 0;
1066   pstate->prev_lexptr = pstate->lexptr;
1067   if (pstate->lexptr[0] == '\0')
1068     {
1069       if (pstate->parse_completion)
1070 	{
1071 	  current_string_val.length =0;
1072 	  current_string_val.ptr = "";
1073 	  return COMPLETE;
1074 	}
1075       return 0;
1076     }
1077 
1078   if (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9')
1079     return lex_number ();
1080   else if (pstate->lexptr[0] == 'b' && pstate->lexptr[1] == '\'')
1081     return lex_character ();
1082   else if (pstate->lexptr[0] == 'b' && pstate->lexptr[1] == '"')
1083     return lex_string ();
1084   else if (pstate->lexptr[0] == 'b' && starts_raw_string (pstate->lexptr + 1))
1085     return lex_string ();
1086   else if (starts_raw_string (pstate->lexptr))
1087     return lex_string ();
1088   else if (rust_identifier_start_p (pstate->lexptr[0]))
1089     return lex_identifier ();
1090   else if (pstate->lexptr[0] == '"')
1091     return lex_string ();
1092   else if (pstate->lexptr[0] == '\'')
1093     return lex_character ();
1094   else if (pstate->lexptr[0] == '}' || pstate->lexptr[0] == ']')
1095     {
1096       /* Falls through to lex_operator.  */
1097       --paren_depth;
1098     }
1099   else if (pstate->lexptr[0] == '(' || pstate->lexptr[0] == '{')
1100     {
1101       /* Falls through to lex_operator.  */
1102       ++paren_depth;
1103     }
1104   else if (pstate->lexptr[0] == ',' && pstate->comma_terminates
1105 	   && paren_depth == 0)
1106     return 0;
1107 
1108   return lex_operator ();
1109 }
1110 
1111 /* Push back a single character to be re-lexed.  */
1112 
1113 void
1114 rust_parser::push_back (char c)
1115 {
1116   /* Can't be called before any lexing.  */
1117   gdb_assert (pstate->prev_lexptr != NULL);
1118 
1119   --pstate->lexptr;
1120   gdb_assert (*pstate->lexptr == c);
1121 }
1122 
1123 
1124 
1125 /* Parse a tuple or paren expression.  */
1126 
1127 operation_up
1128 rust_parser::parse_tuple ()
1129 {
1130   assume ('(');
1131 
1132   if (current_token == ')')
1133     {
1134       lex ();
1135       struct type *unit = get_type ("()");
1136       return make_operation<long_const_operation> (unit, 0);
1137     }
1138 
1139   operation_up expr = parse_expr ();
1140   if (current_token == ')')
1141     {
1142       /* Parenthesized expression.  */
1143       lex ();
1144       return make_operation<rust_parenthesized_operation> (std::move (expr));
1145     }
1146 
1147   std::vector<operation_up> ops;
1148   ops.push_back (std::move (expr));
1149   while (current_token != ')')
1150     {
1151       if (current_token != ',')
1152 	error (_("',' or ')' expected"));
1153       lex ();
1154 
1155       /* A trailing "," is ok.  */
1156       if (current_token != ')')
1157 	ops.push_back (parse_expr ());
1158     }
1159 
1160   assume (')');
1161 
1162   error (_("Tuple expressions not supported yet"));
1163 }
1164 
1165 /* Parse an array expression.  */
1166 
1167 operation_up
1168 rust_parser::parse_array ()
1169 {
1170   assume ('[');
1171 
1172   if (current_token == KW_MUT)
1173     lex ();
1174 
1175   operation_up result;
1176   operation_up expr = parse_expr ();
1177   if (current_token == ';')
1178     {
1179       lex ();
1180       operation_up rhs = parse_expr ();
1181       result = make_operation<rust_array_operation> (std::move (expr),
1182 						     std::move (rhs));
1183     }
1184   else if (current_token == ',')
1185     {
1186       std::vector<operation_up> ops;
1187       ops.push_back (std::move (expr));
1188       while (current_token != ']')
1189 	{
1190 	  if (current_token != ',')
1191 	    error (_("',' or ']' expected"));
1192 	  lex ();
1193 	  ops.push_back (parse_expr ());
1194 	}
1195       ops.shrink_to_fit ();
1196       int len = ops.size () - 1;
1197       result = make_operation<array_operation> (0, len, std::move (ops));
1198     }
1199   else if (current_token != ']')
1200     error (_("',', ';', or ']' expected"));
1201 
1202   require (']');
1203 
1204   return result;
1205 }
1206 
1207 /* Turn a name into an operation.  */
1208 
1209 operation_up
1210 rust_parser::name_to_operation (const std::string &name)
1211 {
1212   struct block_symbol sym = lookup_symbol (name.c_str (),
1213 					   pstate->expression_context_block,
1214 					   VAR_DOMAIN);
1215   if (sym.symbol != nullptr && sym.symbol->aclass () != LOC_TYPEDEF)
1216     return make_operation<var_value_operation> (sym);
1217 
1218   struct type *type = nullptr;
1219 
1220   if (sym.symbol != nullptr)
1221     {
1222       gdb_assert (sym.symbol->aclass () == LOC_TYPEDEF);
1223       type = sym.symbol->type ();
1224     }
1225   if (type == nullptr)
1226     type = rust_lookup_type (name.c_str ());
1227   if (type == nullptr)
1228     error (_("No symbol '%s' in current context"), name.c_str ());
1229 
1230   if (type->code () == TYPE_CODE_STRUCT && type->num_fields () == 0)
1231     {
1232       /* A unit-like struct.  */
1233       operation_up result (new rust_aggregate_operation (type, {}, {}));
1234       return result;
1235     }
1236   else
1237     return make_operation<type_operation> (type);
1238 }
1239 
1240 /* Parse a struct expression.  */
1241 
1242 operation_up
1243 rust_parser::parse_struct_expr (struct type *type)
1244 {
1245   assume ('{');
1246 
1247   if (type->code () != TYPE_CODE_STRUCT
1248       || rust_tuple_type_p (type)
1249       || rust_tuple_struct_type_p (type))
1250     error (_("Struct expression applied to non-struct type"));
1251 
1252   std::vector<std::pair<std::string, operation_up>> field_v;
1253   while (current_token != '}' && current_token != DOTDOT)
1254     {
1255       if (current_token != IDENT)
1256 	error (_("'}', '..', or identifier expected"));
1257 
1258       std::string name = get_string ();
1259       lex ();
1260 
1261       operation_up expr;
1262       if (current_token == ',' || current_token == '}'
1263 	  || current_token == DOTDOT)
1264 	expr = name_to_operation (name);
1265       else
1266 	{
1267 	  require (':');
1268 	  expr = parse_expr ();
1269 	}
1270       field_v.emplace_back (std::move (name), std::move (expr));
1271 
1272       /* A trailing "," is ok.  */
1273       if (current_token == ',')
1274 	lex ();
1275     }
1276 
1277   operation_up others;
1278   if (current_token == DOTDOT)
1279     {
1280       lex ();
1281       others = parse_expr ();
1282     }
1283 
1284   require ('}');
1285 
1286   return make_operation<rust_aggregate_operation> (type,
1287 						   std::move (others),
1288 						   std::move (field_v));
1289 }
1290 
1291 /* Used by the operator precedence parser.  */
1292 struct rustop_item
1293 {
1294   rustop_item (int token_, int precedence_, enum exp_opcode opcode_,
1295 	       operation_up &&op_)
1296     : token (token_),
1297       precedence (precedence_),
1298       opcode (opcode_),
1299       op (std::move (op_))
1300   {
1301   }
1302 
1303   /* The token value.  */
1304   int token;
1305   /* Precedence of this operator.  */
1306   int precedence;
1307   /* This is used only for assign-modify.  */
1308   enum exp_opcode opcode;
1309   /* The right hand side of this operation.  */
1310   operation_up op;
1311 };
1312 
1313 /* An operator precedence parser for binary operations, including
1314    "as".  */
1315 
1316 operation_up
1317 rust_parser::parse_binop (bool required)
1318 {
1319   /* All the binary  operators.  Each one is of the form
1320      OPERATION(TOKEN, PRECEDENCE, TYPE)
1321      TOKEN is the corresponding operator token.
1322      PRECEDENCE is a value indicating relative precedence.
1323      TYPE is the operation type corresponding to the operator.
1324      Assignment operations are handled specially, not via this
1325      table; they have precedence 0.  */
1326 #define ALL_OPS					\
1327   OPERATION ('*', 10, mul_operation)		\
1328   OPERATION ('/', 10, div_operation)		\
1329   OPERATION ('%', 10, rem_operation)		\
1330   OPERATION ('@', 9, repeat_operation)		\
1331   OPERATION ('+', 8, add_operation)		\
1332   OPERATION ('-', 8, sub_operation)		\
1333   OPERATION (LSH, 7, lsh_operation)		\
1334   OPERATION (RSH, 7, rsh_operation)		\
1335   OPERATION ('&', 6, bitwise_and_operation)	\
1336   OPERATION ('^', 5, bitwise_xor_operation)	\
1337   OPERATION ('|', 4, bitwise_ior_operation)	\
1338   OPERATION (EQEQ, 3, equal_operation)		\
1339   OPERATION (NOTEQ, 3, notequal_operation)	\
1340   OPERATION ('<', 3, less_operation)		\
1341   OPERATION (LTEQ, 3, leq_operation)		\
1342   OPERATION ('>', 3, gtr_operation)		\
1343   OPERATION (GTEQ, 3, geq_operation)		\
1344   OPERATION (ANDAND, 2, logical_and_operation)	\
1345   OPERATION (OROR, 1, logical_or_operation)
1346 
1347 #define ASSIGN_PREC 0
1348 
1349   operation_up start = parse_atom (required);
1350   if (start == nullptr)
1351     {
1352       gdb_assert (!required);
1353       return start;
1354     }
1355 
1356   std::vector<rustop_item> operator_stack;
1357   operator_stack.emplace_back (0, -1, OP_NULL, std::move (start));
1358 
1359   while (true)
1360     {
1361       int this_token = current_token;
1362       enum exp_opcode compound_assign_op = OP_NULL;
1363       int precedence = -2;
1364 
1365       switch (this_token)
1366 	{
1367 #define OPERATION(TOKEN, PRECEDENCE, TYPE)		\
1368 	  case TOKEN:				\
1369 	    precedence = PRECEDENCE;		\
1370 	    lex ();				\
1371 	    break;
1372 
1373 	  ALL_OPS
1374 
1375 #undef OPERATION
1376 
1377 	case COMPOUND_ASSIGN:
1378 	  compound_assign_op = current_opcode;
1379 	  /* FALLTHROUGH */
1380 	case '=':
1381 	  precedence = ASSIGN_PREC;
1382 	  lex ();
1383 	  break;
1384 
1385 	  /* "as" must be handled specially.  */
1386 	case KW_AS:
1387 	  {
1388 	    lex ();
1389 	    rustop_item &lhs = operator_stack.back ();
1390 	    struct type *type = parse_type ();
1391 	    lhs.op = make_operation<unop_cast_operation> (std::move (lhs.op),
1392 							  type);
1393 	  }
1394 	  /* Bypass the rest of the loop.  */
1395 	  continue;
1396 
1397 	default:
1398 	  /* Arrange to pop the entire stack.  */
1399 	  precedence = -2;
1400 	  break;
1401         }
1402 
1403       /* Make sure that assignments are right-associative while other
1404 	 operations are left-associative.  */
1405       while ((precedence == ASSIGN_PREC
1406 	      ? precedence < operator_stack.back ().precedence
1407 	      : precedence <= operator_stack.back ().precedence)
1408 	     && operator_stack.size () > 1)
1409 	{
1410 	  rustop_item rhs = std::move (operator_stack.back ());
1411 	  operator_stack.pop_back ();
1412 
1413 	  rustop_item &lhs = operator_stack.back ();
1414 
1415 	  switch (rhs.token)
1416 	    {
1417 #define OPERATION(TOKEN, PRECEDENCE, TYPE)			\
1418 	  case TOKEN:						\
1419 	    lhs.op = make_operation<TYPE> (std::move (lhs.op),	\
1420 					   std::move (rhs.op)); \
1421 	    break;
1422 
1423 	      ALL_OPS
1424 
1425 #undef OPERATION
1426 
1427 	    case '=':
1428 	    case COMPOUND_ASSIGN:
1429 	      {
1430 		if (rhs.token == '=')
1431 		  lhs.op = (make_operation<assign_operation>
1432 			    (std::move (lhs.op), std::move (rhs.op)));
1433 		else
1434 		  lhs.op = (make_operation<assign_modify_operation>
1435 			    (rhs.opcode, std::move (lhs.op),
1436 			     std::move (rhs.op)));
1437 
1438 		struct type *unit_type = get_type ("()");
1439 
1440 		operation_up nil (new long_const_operation (unit_type, 0));
1441 		lhs.op = (make_operation<comma_operation>
1442 			  (std::move (lhs.op), std::move (nil)));
1443 	      }
1444 	      break;
1445 
1446 	    default:
1447 	      gdb_assert_not_reached ("bad binary operator");
1448 	    }
1449 	}
1450 
1451       if (precedence == -2)
1452 	break;
1453 
1454       operator_stack.emplace_back (this_token, precedence, compound_assign_op,
1455 				   parse_atom (true));
1456     }
1457 
1458   gdb_assert (operator_stack.size () == 1);
1459   return std::move (operator_stack[0].op);
1460 #undef ALL_OPS
1461 }
1462 
1463 /* Parse a range expression.  */
1464 
1465 operation_up
1466 rust_parser::parse_range ()
1467 {
1468   enum range_flag kind = (RANGE_HIGH_BOUND_DEFAULT
1469 			  | RANGE_LOW_BOUND_DEFAULT);
1470 
1471   operation_up lhs;
1472   if (current_token != DOTDOT && current_token != DOTDOTEQ)
1473     {
1474       lhs = parse_binop (true);
1475       kind &= ~RANGE_LOW_BOUND_DEFAULT;
1476     }
1477 
1478   if (current_token == DOTDOT)
1479     kind |= RANGE_HIGH_BOUND_EXCLUSIVE;
1480   else if (current_token != DOTDOTEQ)
1481     return lhs;
1482   lex ();
1483 
1484   /* A "..=" range requires a high bound, but otherwise it is
1485      optional.  */
1486   operation_up rhs = parse_binop ((kind & RANGE_HIGH_BOUND_EXCLUSIVE) == 0);
1487   if (rhs != nullptr)
1488     kind &= ~RANGE_HIGH_BOUND_DEFAULT;
1489 
1490   return make_operation<rust_range_operation> (kind,
1491 					       std::move (lhs),
1492 					       std::move (rhs));
1493 }
1494 
1495 /* Parse an expression.  */
1496 
1497 operation_up
1498 rust_parser::parse_expr ()
1499 {
1500   return parse_range ();
1501 }
1502 
1503 /* Parse a sizeof expression.  */
1504 
1505 operation_up
1506 rust_parser::parse_sizeof ()
1507 {
1508   assume (KW_SIZEOF);
1509 
1510   require ('(');
1511   operation_up result = make_operation<unop_sizeof_operation> (parse_expr ());
1512   require (')');
1513   return result;
1514 }
1515 
1516 /* Parse an address-of operation.  */
1517 
1518 operation_up
1519 rust_parser::parse_addr ()
1520 {
1521   assume ('&');
1522 
1523   if (current_token == KW_MUT)
1524     lex ();
1525 
1526   return make_operation<rust_unop_addr_operation> (parse_atom (true));
1527 }
1528 
1529 /* Parse a field expression.  */
1530 
1531 operation_up
1532 rust_parser::parse_field (operation_up &&lhs)
1533 {
1534   assume ('.');
1535 
1536   operation_up result;
1537   switch (current_token)
1538     {
1539     case IDENT:
1540     case COMPLETE:
1541       {
1542 	bool is_complete = current_token == COMPLETE;
1543 	auto struct_op = new rust_structop (std::move (lhs), get_string ());
1544 	lex ();
1545 	if (is_complete)
1546 	  {
1547 	    completion_op.reset (struct_op);
1548 	    pstate->mark_struct_expression (struct_op);
1549 	    /* Throw to the outermost level of the parser.  */
1550 	    error (_("not really an error"));
1551 	  }
1552 	result.reset (struct_op);
1553       }
1554       break;
1555 
1556     case DECIMAL_INTEGER:
1557       result = make_operation<rust_struct_anon> (current_int_val.val,
1558 						 std::move (lhs));
1559       lex ();
1560       break;
1561 
1562     case INTEGER:
1563       error (_("'_' not allowed in integers in anonymous field references"));
1564 
1565     default:
1566       error (_("field name expected"));
1567     }
1568 
1569   return result;
1570 }
1571 
1572 /* Parse an index expression.  */
1573 
1574 operation_up
1575 rust_parser::parse_index (operation_up &&lhs)
1576 {
1577   assume ('[');
1578   operation_up rhs = parse_expr ();
1579   require (']');
1580 
1581   return make_operation<rust_subscript_operation> (std::move (lhs),
1582 						   std::move (rhs));
1583 }
1584 
1585 /* Parse a sequence of comma-separated expressions in parens.  */
1586 
1587 std::vector<operation_up>
1588 rust_parser::parse_paren_args ()
1589 {
1590   assume ('(');
1591 
1592   std::vector<operation_up> args;
1593   while (current_token != ')')
1594     {
1595       if (!args.empty ())
1596 	{
1597 	  if (current_token != ',')
1598 	    error (_("',' or ')' expected"));
1599 	  lex ();
1600 	}
1601 
1602       args.push_back (parse_expr ());
1603     }
1604 
1605   assume (')');
1606 
1607   return args;
1608 }
1609 
1610 /* Parse the parenthesized part of a function call.  */
1611 
1612 operation_up
1613 rust_parser::parse_call (operation_up &&lhs)
1614 {
1615   std::vector<operation_up> args = parse_paren_args ();
1616 
1617   return make_operation<funcall_operation> (std::move (lhs),
1618 					    std::move (args));
1619 }
1620 
1621 /* Parse a list of types.  */
1622 
1623 std::vector<struct type *>
1624 rust_parser::parse_type_list ()
1625 {
1626   std::vector<struct type *> result;
1627   result.push_back (parse_type ());
1628   while (current_token == ',')
1629     {
1630       lex ();
1631       result.push_back (parse_type ());
1632     }
1633   return result;
1634 }
1635 
1636 /* Parse a possibly-empty list of types, surrounded in parens.  */
1637 
1638 std::vector<struct type *>
1639 rust_parser::parse_maybe_type_list ()
1640 {
1641   assume ('(');
1642   std::vector<struct type *> types;
1643   if (current_token != ')')
1644     types = parse_type_list ();
1645   require (')');
1646   return types;
1647 }
1648 
1649 /* Parse an array type.  */
1650 
1651 struct type *
1652 rust_parser::parse_array_type ()
1653 {
1654   assume ('[');
1655   struct type *elt_type = parse_type ();
1656   require (';');
1657 
1658   if (current_token != INTEGER && current_token != DECIMAL_INTEGER)
1659     error (_("integer expected"));
1660   ULONGEST val = current_int_val.val;
1661   lex ();
1662   require (']');
1663 
1664   return lookup_array_range_type (elt_type, 0, val - 1);
1665 }
1666 
1667 /* Parse a slice type.  */
1668 
1669 struct type *
1670 rust_parser::parse_slice_type ()
1671 {
1672   assume ('&');
1673 
1674   bool is_slice = current_token == '[';
1675   if (is_slice)
1676     lex ();
1677 
1678   struct type *target = parse_type ();
1679 
1680   if (is_slice)
1681     {
1682       require (']');
1683       return rust_slice_type ("&[*gdb*]", target, get_type ("usize"));
1684     }
1685 
1686   /* For now we treat &x and *x identically.  */
1687   return lookup_pointer_type (target);
1688 }
1689 
1690 /* Parse a pointer type.  */
1691 
1692 struct type *
1693 rust_parser::parse_pointer_type ()
1694 {
1695   assume ('*');
1696 
1697   if (current_token == KW_MUT || current_token == KW_CONST)
1698     lex ();
1699 
1700   struct type *target = parse_type ();
1701   /* For the time being we ignore mut/const.  */
1702   return lookup_pointer_type (target);
1703 }
1704 
1705 /* Parse a function type.  */
1706 
1707 struct type *
1708 rust_parser::parse_function_type ()
1709 {
1710   assume (KW_FN);
1711 
1712   if (current_token != '(')
1713     error (_("'(' expected"));
1714 
1715   std::vector<struct type *> types = parse_maybe_type_list ();
1716 
1717   if (current_token != ARROW)
1718     error (_("'->' expected"));
1719   lex ();
1720 
1721   struct type *result_type = parse_type ();
1722 
1723   struct type **argtypes = nullptr;
1724   if (!types.empty ())
1725     argtypes = types.data ();
1726 
1727   result_type = lookup_function_type_with_arguments (result_type,
1728 						     types.size (),
1729 						     argtypes);
1730   return lookup_pointer_type (result_type);
1731 }
1732 
1733 /* Parse a tuple type.  */
1734 
1735 struct type *
1736 rust_parser::parse_tuple_type ()
1737 {
1738   std::vector<struct type *> types = parse_maybe_type_list ();
1739 
1740   auto_obstack obstack;
1741   obstack_1grow (&obstack, '(');
1742   for (int i = 0; i < types.size (); ++i)
1743     {
1744       std::string type_name = type_to_string (types[i]);
1745 
1746       if (i > 0)
1747 	obstack_1grow (&obstack, ',');
1748       obstack_grow_str (&obstack, type_name.c_str ());
1749     }
1750 
1751   obstack_grow_str0 (&obstack, ")");
1752   const char *name = (const char *) obstack_finish (&obstack);
1753 
1754   /* We don't allow creating new tuple types (yet), but we do allow
1755      looking up existing tuple types.  */
1756   struct type *result = rust_lookup_type (name);
1757   if (result == nullptr)
1758     error (_("could not find tuple type '%s'"), name);
1759 
1760   return result;
1761 }
1762 
1763 /* Parse a type.  */
1764 
1765 struct type *
1766 rust_parser::parse_type ()
1767 {
1768   switch (current_token)
1769     {
1770     case '[':
1771       return parse_array_type ();
1772     case '&':
1773       return parse_slice_type ();
1774     case '*':
1775       return parse_pointer_type ();
1776     case KW_FN:
1777       return parse_function_type ();
1778     case '(':
1779       return parse_tuple_type ();
1780     case KW_SELF:
1781     case KW_SUPER:
1782     case COLONCOLON:
1783     case KW_EXTERN:
1784     case IDENT:
1785       {
1786 	std::string path = parse_path (false);
1787 	struct type *result = rust_lookup_type (path.c_str ());
1788 	if (result == nullptr)
1789 	  error (_("No type name '%s' in current context"), path.c_str ());
1790 	return result;
1791       }
1792     default:
1793       error (_("type expected"));
1794     }
1795 }
1796 
1797 /* Parse a path.  */
1798 
1799 std::string
1800 rust_parser::parse_path (bool for_expr)
1801 {
1802   unsigned n_supers = 0;
1803   int first_token = current_token;
1804 
1805   switch (current_token)
1806     {
1807     case KW_SELF:
1808       lex ();
1809       if (current_token != COLONCOLON)
1810 	return "self";
1811       lex ();
1812       /* FALLTHROUGH */
1813     case KW_SUPER:
1814       while (current_token == KW_SUPER)
1815 	{
1816 	  ++n_supers;
1817 	  lex ();
1818 	  if (current_token != COLONCOLON)
1819 	    error (_("'::' expected"));
1820 	  lex ();
1821 	}
1822       break;
1823 
1824     case COLONCOLON:
1825       lex ();
1826       break;
1827 
1828     case KW_EXTERN:
1829       /* This is a gdb extension to make it possible to refer to items
1830 	 in other crates.  It just bypasses adding the current crate
1831 	 to the front of the name.  */
1832       lex ();
1833       break;
1834     }
1835 
1836   if (current_token != IDENT)
1837     error (_("identifier expected"));
1838   std::string path = get_string ();
1839   bool saw_ident = true;
1840   lex ();
1841 
1842   /* The condition here lets us enter the loop even if we see
1843      "ident<...>".  */
1844   while (current_token == COLONCOLON || current_token == '<')
1845     {
1846       if (current_token == COLONCOLON)
1847 	{
1848 	  lex ();
1849 	  saw_ident = false;
1850 
1851 	  if (current_token == IDENT)
1852 	    {
1853 	      path = path + "::" + get_string ();
1854 	      lex ();
1855 	      saw_ident = true;
1856 	    }
1857 	  else if (current_token == COLONCOLON)
1858 	    {
1859 	      /* The code below won't detect this scenario.  */
1860 	      error (_("unexpected '::'"));
1861 	    }
1862 	}
1863 
1864       if (current_token != '<')
1865 	continue;
1866 
1867       /* Expression use name::<...>, whereas types use name<...>.  */
1868       if (for_expr)
1869 	{
1870 	  /* Expressions use "name::<...>", so if we saw an identifier
1871 	     after the "::", we ignore the "<" here.  */
1872 	  if (saw_ident)
1873 	    break;
1874 	}
1875       else
1876 	{
1877 	  /* Types use "name<...>", so we need to have seen the
1878 	     identifier.  */
1879 	  if (!saw_ident)
1880 	    break;
1881 	}
1882 
1883       lex ();
1884       std::vector<struct type *> types = parse_type_list ();
1885       if (current_token == '>')
1886 	lex ();
1887       else if (current_token == RSH)
1888 	{
1889 	  push_back ('>');
1890 	  lex ();
1891 	}
1892       else
1893 	error (_("'>' expected"));
1894 
1895       path += "<";
1896       for (int i = 0; i < types.size (); ++i)
1897 	{
1898 	  if (i > 0)
1899 	    path += ",";
1900 	  path += type_to_string (types[i]);
1901 	}
1902       path += ">";
1903       break;
1904     }
1905 
1906   switch (first_token)
1907     {
1908     case KW_SELF:
1909     case KW_SUPER:
1910       return super_name (path, n_supers);
1911 
1912     case COLONCOLON:
1913       return crate_name (path);
1914 
1915     case KW_EXTERN:
1916       return "::" + path;
1917 
1918     case IDENT:
1919       return path;
1920 
1921     default:
1922       gdb_assert_not_reached ("missing case in path parsing");
1923     }
1924 }
1925 
1926 /* Handle the parsing for a string expression.  */
1927 
1928 operation_up
1929 rust_parser::parse_string ()
1930 {
1931   gdb_assert (current_token == STRING);
1932 
1933   /* Wrap the raw string in the &str struct.  */
1934   struct type *type = rust_lookup_type ("&str");
1935   if (type == nullptr)
1936     error (_("Could not find type '&str'"));
1937 
1938   std::vector<std::pair<std::string, operation_up>> field_v;
1939 
1940   size_t len = current_string_val.length;
1941   operation_up str = make_operation<string_operation> (get_string ());
1942   operation_up addr
1943     = make_operation<rust_unop_addr_operation> (std::move (str));
1944   field_v.emplace_back ("data_ptr", std::move (addr));
1945 
1946   struct type *valtype = get_type ("usize");
1947   operation_up lenop = make_operation<long_const_operation> (valtype, len);
1948   field_v.emplace_back ("length", std::move (lenop));
1949 
1950   return make_operation<rust_aggregate_operation> (type,
1951 						   operation_up (),
1952 						   std::move (field_v));
1953 }
1954 
1955 /* Parse a tuple struct expression.  */
1956 
1957 operation_up
1958 rust_parser::parse_tuple_struct (struct type *type)
1959 {
1960   std::vector<operation_up> args = parse_paren_args ();
1961 
1962   std::vector<std::pair<std::string, operation_up>> field_v (args.size ());
1963   for (int i = 0; i < args.size (); ++i)
1964     field_v[i] = { string_printf ("__%d", i), std::move (args[i]) };
1965 
1966   return (make_operation<rust_aggregate_operation>
1967 	  (type, operation_up (), std::move (field_v)));
1968 }
1969 
1970 /* Parse a path expression.  */
1971 
1972 operation_up
1973 rust_parser::parse_path_expr ()
1974 {
1975   std::string path = parse_path (true);
1976 
1977   if (current_token == '{')
1978     {
1979       struct type *type = rust_lookup_type (path.c_str ());
1980       if (type == nullptr)
1981 	error (_("Could not find type '%s'"), path.c_str ());
1982 
1983       return parse_struct_expr (type);
1984     }
1985   else if (current_token == '(')
1986     {
1987       struct type *type = rust_lookup_type (path.c_str ());
1988       /* If this is actually a tuple struct expression, handle it
1989 	 here.  If it is a call, it will be handled elsewhere.  */
1990       if (type != nullptr)
1991 	{
1992 	  if (!rust_tuple_struct_type_p (type))
1993 	    error (_("Type %s is not a tuple struct"), path.c_str ());
1994 	  return parse_tuple_struct (type);
1995 	}
1996     }
1997 
1998   return name_to_operation (path);
1999 }
2000 
2001 /* Parse an atom.  "Atom" isn't a Rust term, but this refers to a
2002    single unitary item in the grammar; but here including some unary
2003    prefix and postfix expressions.  */
2004 
2005 operation_up
2006 rust_parser::parse_atom (bool required)
2007 {
2008   operation_up result;
2009 
2010   switch (current_token)
2011     {
2012     case '(':
2013       result = parse_tuple ();
2014       break;
2015 
2016     case '[':
2017       result = parse_array ();
2018       break;
2019 
2020     case INTEGER:
2021     case DECIMAL_INTEGER:
2022       result = make_operation<long_const_operation> (current_int_val.type,
2023 						     current_int_val.val);
2024       lex ();
2025       break;
2026 
2027     case FLOAT:
2028       result = make_operation<float_const_operation> (current_float_val.type,
2029 						      current_float_val.val);
2030       lex ();
2031       break;
2032 
2033     case STRING:
2034       result = parse_string ();
2035       lex ();
2036       break;
2037 
2038     case BYTESTRING:
2039       result = make_operation<string_operation> (get_string ());
2040       lex ();
2041       break;
2042 
2043     case KW_TRUE:
2044     case KW_FALSE:
2045       result = make_operation<bool_operation> (current_token == KW_TRUE);
2046       lex ();
2047       break;
2048 
2049     case GDBVAR:
2050       /* This is kind of a hacky approach.  */
2051       {
2052 	pstate->push_dollar (current_string_val);
2053 	result = pstate->pop ();
2054 	lex ();
2055       }
2056       break;
2057 
2058     case KW_SELF:
2059     case KW_SUPER:
2060     case COLONCOLON:
2061     case KW_EXTERN:
2062     case IDENT:
2063       result = parse_path_expr ();
2064       break;
2065 
2066     case '*':
2067       lex ();
2068       result = make_operation<rust_unop_ind_operation> (parse_atom (true));
2069       break;
2070     case '+':
2071       lex ();
2072       result = make_operation<unary_plus_operation> (parse_atom (true));
2073       break;
2074     case '-':
2075       lex ();
2076       result = make_operation<unary_neg_operation> (parse_atom (true));
2077       break;
2078     case '!':
2079       lex ();
2080       result = make_operation<rust_unop_compl_operation> (parse_atom (true));
2081       break;
2082     case KW_SIZEOF:
2083       result = parse_sizeof ();
2084       break;
2085     case '&':
2086       result = parse_addr ();
2087       break;
2088 
2089     default:
2090       if (!required)
2091 	return {};
2092       error (_("unexpected token"));
2093     }
2094 
2095   /* Now parse suffixes.  */
2096   while (true)
2097     {
2098       switch (current_token)
2099 	{
2100 	case '.':
2101 	  result = parse_field (std::move (result));
2102 	  break;
2103 
2104 	case '[':
2105 	  result = parse_index (std::move (result));
2106 	  break;
2107 
2108 	case '(':
2109 	  result = parse_call (std::move (result));
2110 	  break;
2111 
2112 	default:
2113 	  return result;
2114 	}
2115     }
2116 }
2117 
2118 
2119 
2120 /* The parser as exposed to gdb.  */
2121 
2122 int
2123 rust_language::parser (struct parser_state *state) const
2124 {
2125   rust_parser parser (state);
2126 
2127   operation_up result;
2128   try
2129     {
2130       result = parser.parse_entry_point ();
2131     }
2132   catch (const gdb_exception &exc)
2133     {
2134       if (state->parse_completion)
2135 	{
2136 	  result = std::move (parser.completion_op);
2137 	  if (result == nullptr)
2138 	    throw;
2139 	}
2140       else
2141 	throw;
2142     }
2143 
2144   state->set_operation (std::move (result));
2145 
2146   return 0;
2147 }
2148 
2149 
2150 
2151 #if GDB_SELF_TEST
2152 
2153 /* A test helper that lexes a string, expecting a single token.  */
2154 
2155 static void
2156 rust_lex_test_one (rust_parser *parser, const char *input, int expected)
2157 {
2158   int token;
2159 
2160   parser->reset (input);
2161 
2162   token = parser->lex_one_token ();
2163   SELF_CHECK (token == expected);
2164 
2165   if (token)
2166     {
2167       token = parser->lex_one_token ();
2168       SELF_CHECK (token == 0);
2169     }
2170 }
2171 
2172 /* Test that INPUT lexes as the integer VALUE.  */
2173 
2174 static void
2175 rust_lex_int_test (rust_parser *parser, const char *input,
2176 		   ULONGEST value, int kind)
2177 {
2178   rust_lex_test_one (parser, input, kind);
2179   SELF_CHECK (parser->current_int_val.val == value);
2180 }
2181 
2182 /* Test that INPUT throws an exception with text ERR.  */
2183 
2184 static void
2185 rust_lex_exception_test (rust_parser *parser, const char *input,
2186 			 const char *err)
2187 {
2188   try
2189     {
2190       /* The "kind" doesn't matter.  */
2191       rust_lex_test_one (parser, input, DECIMAL_INTEGER);
2192       SELF_CHECK (0);
2193     }
2194   catch (const gdb_exception_error &except)
2195     {
2196       SELF_CHECK (strcmp (except.what (), err) == 0);
2197     }
2198 }
2199 
2200 /* Test that INPUT lexes as the identifier, string, or byte-string
2201    VALUE.  KIND holds the expected token kind.  */
2202 
2203 static void
2204 rust_lex_stringish_test (rust_parser *parser, const char *input,
2205 			 const char *value, int kind)
2206 {
2207   rust_lex_test_one (parser, input, kind);
2208   SELF_CHECK (parser->get_string () == value);
2209 }
2210 
2211 /* Helper to test that a string parses as a given token sequence.  */
2212 
2213 static void
2214 rust_lex_test_sequence (rust_parser *parser, const char *input, int len,
2215 			const int expected[])
2216 {
2217   int i;
2218 
2219   parser->reset (input);
2220 
2221   for (i = 0; i < len; ++i)
2222     {
2223       int token = parser->lex_one_token ();
2224       SELF_CHECK (token == expected[i]);
2225     }
2226 }
2227 
2228 /* Tests for an integer-parsing corner case.  */
2229 
2230 static void
2231 rust_lex_test_trailing_dot (rust_parser *parser)
2232 {
2233   const int expected1[] = { DECIMAL_INTEGER, '.', IDENT, '(', ')', 0 };
2234   const int expected2[] = { INTEGER, '.', IDENT, '(', ')', 0 };
2235   const int expected3[] = { FLOAT, EQEQ, '(', ')', 0 };
2236   const int expected4[] = { DECIMAL_INTEGER, DOTDOT, DECIMAL_INTEGER, 0 };
2237 
2238   rust_lex_test_sequence (parser, "23.g()", ARRAY_SIZE (expected1), expected1);
2239   rust_lex_test_sequence (parser, "23_0.g()", ARRAY_SIZE (expected2),
2240 			  expected2);
2241   rust_lex_test_sequence (parser, "23.==()", ARRAY_SIZE (expected3),
2242 			  expected3);
2243   rust_lex_test_sequence (parser, "23..25", ARRAY_SIZE (expected4), expected4);
2244 }
2245 
2246 /* Tests of completion.  */
2247 
2248 static void
2249 rust_lex_test_completion (rust_parser *parser)
2250 {
2251   const int expected[] = { IDENT, '.', COMPLETE, 0 };
2252 
2253   parser->pstate->parse_completion = 1;
2254 
2255   rust_lex_test_sequence (parser, "something.wha", ARRAY_SIZE (expected),
2256 			  expected);
2257   rust_lex_test_sequence (parser, "something.", ARRAY_SIZE (expected),
2258 			  expected);
2259 
2260   parser->pstate->parse_completion = 0;
2261 }
2262 
2263 /* Test pushback.  */
2264 
2265 static void
2266 rust_lex_test_push_back (rust_parser *parser)
2267 {
2268   int token;
2269 
2270   parser->reset (">>=");
2271 
2272   token = parser->lex_one_token ();
2273   SELF_CHECK (token == COMPOUND_ASSIGN);
2274   SELF_CHECK (parser->current_opcode == BINOP_RSH);
2275 
2276   parser->push_back ('=');
2277 
2278   token = parser->lex_one_token ();
2279   SELF_CHECK (token == '=');
2280 
2281   token = parser->lex_one_token ();
2282   SELF_CHECK (token == 0);
2283 }
2284 
2285 /* Unit test the lexer.  */
2286 
2287 static void
2288 rust_lex_tests (void)
2289 {
2290   /* Set up dummy "parser", so that rust_type works.  */
2291   struct parser_state ps (language_def (language_rust), target_gdbarch (),
2292 			  nullptr, 0, 0, nullptr, 0, nullptr, false);
2293   rust_parser parser (&ps);
2294 
2295   rust_lex_test_one (&parser, "", 0);
2296   rust_lex_test_one (&parser, "    \t  \n \r  ", 0);
2297   rust_lex_test_one (&parser, "thread 23", 0);
2298   rust_lex_test_one (&parser, "task 23", 0);
2299   rust_lex_test_one (&parser, "th 104", 0);
2300   rust_lex_test_one (&parser, "ta 97", 0);
2301 
2302   rust_lex_int_test (&parser, "'z'", 'z', INTEGER);
2303   rust_lex_int_test (&parser, "'\\xff'", 0xff, INTEGER);
2304   rust_lex_int_test (&parser, "'\\u{1016f}'", 0x1016f, INTEGER);
2305   rust_lex_int_test (&parser, "b'z'", 'z', INTEGER);
2306   rust_lex_int_test (&parser, "b'\\xfe'", 0xfe, INTEGER);
2307   rust_lex_int_test (&parser, "b'\\xFE'", 0xfe, INTEGER);
2308   rust_lex_int_test (&parser, "b'\\xfE'", 0xfe, INTEGER);
2309 
2310   /* Test all escapes in both modes.  */
2311   rust_lex_int_test (&parser, "'\\n'", '\n', INTEGER);
2312   rust_lex_int_test (&parser, "'\\r'", '\r', INTEGER);
2313   rust_lex_int_test (&parser, "'\\t'", '\t', INTEGER);
2314   rust_lex_int_test (&parser, "'\\\\'", '\\', INTEGER);
2315   rust_lex_int_test (&parser, "'\\0'", '\0', INTEGER);
2316   rust_lex_int_test (&parser, "'\\''", '\'', INTEGER);
2317   rust_lex_int_test (&parser, "'\\\"'", '"', INTEGER);
2318 
2319   rust_lex_int_test (&parser, "b'\\n'", '\n', INTEGER);
2320   rust_lex_int_test (&parser, "b'\\r'", '\r', INTEGER);
2321   rust_lex_int_test (&parser, "b'\\t'", '\t', INTEGER);
2322   rust_lex_int_test (&parser, "b'\\\\'", '\\', INTEGER);
2323   rust_lex_int_test (&parser, "b'\\0'", '\0', INTEGER);
2324   rust_lex_int_test (&parser, "b'\\''", '\'', INTEGER);
2325   rust_lex_int_test (&parser, "b'\\\"'", '"', INTEGER);
2326 
2327   rust_lex_exception_test (&parser, "'z", "Unterminated character literal");
2328   rust_lex_exception_test (&parser, "b'\\x0'", "Not enough hex digits seen");
2329   rust_lex_exception_test (&parser, "b'\\u{0}'",
2330 			   "Unicode escape in byte literal");
2331   rust_lex_exception_test (&parser, "'\\x0'", "Not enough hex digits seen");
2332   rust_lex_exception_test (&parser, "'\\u0'", "Missing '{' in Unicode escape");
2333   rust_lex_exception_test (&parser, "'\\u{0", "Missing '}' in Unicode escape");
2334   rust_lex_exception_test (&parser, "'\\u{0000007}", "Overlong hex escape");
2335   rust_lex_exception_test (&parser, "'\\u{}", "Not enough hex digits seen");
2336   rust_lex_exception_test (&parser, "'\\Q'", "Invalid escape \\Q in literal");
2337   rust_lex_exception_test (&parser, "b'\\Q'", "Invalid escape \\Q in literal");
2338 
2339   rust_lex_int_test (&parser, "23", 23, DECIMAL_INTEGER);
2340   rust_lex_int_test (&parser, "2_344__29", 234429, INTEGER);
2341   rust_lex_int_test (&parser, "0x1f", 0x1f, INTEGER);
2342   rust_lex_int_test (&parser, "23usize", 23, INTEGER);
2343   rust_lex_int_test (&parser, "23i32", 23, INTEGER);
2344   rust_lex_int_test (&parser, "0x1_f", 0x1f, INTEGER);
2345   rust_lex_int_test (&parser, "0b1_101011__", 0x6b, INTEGER);
2346   rust_lex_int_test (&parser, "0o001177i64", 639, INTEGER);
2347   rust_lex_int_test (&parser, "0x123456789u64", 0x123456789ull, INTEGER);
2348 
2349   rust_lex_test_trailing_dot (&parser);
2350 
2351   rust_lex_test_one (&parser, "23.", FLOAT);
2352   rust_lex_test_one (&parser, "23.99f32", FLOAT);
2353   rust_lex_test_one (&parser, "23e7", FLOAT);
2354   rust_lex_test_one (&parser, "23E-7", FLOAT);
2355   rust_lex_test_one (&parser, "23e+7", FLOAT);
2356   rust_lex_test_one (&parser, "23.99e+7f64", FLOAT);
2357   rust_lex_test_one (&parser, "23.82f32", FLOAT);
2358 
2359   rust_lex_stringish_test (&parser, "hibob", "hibob", IDENT);
2360   rust_lex_stringish_test (&parser, "hibob__93", "hibob__93", IDENT);
2361   rust_lex_stringish_test (&parser, "thread", "thread", IDENT);
2362   rust_lex_stringish_test (&parser, "r#true", "true", IDENT);
2363 
2364   const int expected1[] = { IDENT, DECIMAL_INTEGER, 0 };
2365   rust_lex_test_sequence (&parser, "r#thread 23", ARRAY_SIZE (expected1),
2366 			  expected1);
2367   const int expected2[] = { IDENT, '#', 0 };
2368   rust_lex_test_sequence (&parser, "r#", ARRAY_SIZE (expected2), expected2);
2369 
2370   rust_lex_stringish_test (&parser, "\"string\"", "string", STRING);
2371   rust_lex_stringish_test (&parser, "\"str\\ting\"", "str\ting", STRING);
2372   rust_lex_stringish_test (&parser, "\"str\\\"ing\"", "str\"ing", STRING);
2373   rust_lex_stringish_test (&parser, "r\"str\\ing\"", "str\\ing", STRING);
2374   rust_lex_stringish_test (&parser, "r#\"str\\ting\"#", "str\\ting", STRING);
2375   rust_lex_stringish_test (&parser, "r###\"str\\\"ing\"###", "str\\\"ing",
2376 			   STRING);
2377 
2378   rust_lex_stringish_test (&parser, "b\"string\"", "string", BYTESTRING);
2379   rust_lex_stringish_test (&parser, "b\"\x73tring\"", "string", BYTESTRING);
2380   rust_lex_stringish_test (&parser, "b\"str\\\"ing\"", "str\"ing", BYTESTRING);
2381   rust_lex_stringish_test (&parser, "br####\"\\x73tring\"####", "\\x73tring",
2382 			   BYTESTRING);
2383 
2384   for (const auto &candidate : identifier_tokens)
2385     rust_lex_test_one (&parser, candidate.name, candidate.value);
2386 
2387   for (const auto &candidate : operator_tokens)
2388     rust_lex_test_one (&parser, candidate.name, candidate.value);
2389 
2390   rust_lex_test_completion (&parser);
2391   rust_lex_test_push_back (&parser);
2392 }
2393 
2394 #endif /* GDB_SELF_TEST */
2395 
2396 
2397 
2398 void _initialize_rust_exp ();
2399 void
2400 _initialize_rust_exp ()
2401 {
2402   int code = regcomp (&number_regex, number_regex_text, REG_EXTENDED);
2403   /* If the regular expression was incorrect, it was a programming
2404      error.  */
2405   gdb_assert (code == 0);
2406 
2407 #if GDB_SELF_TEST
2408   selftests::register_test ("rust-lex", rust_lex_tests);
2409 #endif
2410 }
2411