From: Tom Tromey Date: Fri, 16 Apr 2021 22:34:07 +0000 (-0600) Subject: Rewrite the Rust expression parser X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=3cbc7ac344cd6b500b54588a9099c037de58e75a;p=binutils-gdb.git Rewrite the Rust expression parser The Rust expression parser was written to construct its own AST, then lower this to GDB expressions. I did this primarily because the old expressions were difficult to work with; after rewriting those, I realized I could remove the AST from the Rust parser. After looking at this, I realized it might be simpler to rewrite the parser. This patch reimplements it as a recursive-descent parser. I kept a fair amount of the existing code -- the lexer is pulled in nearly unchanged. There are several benefits to this approach: * The parser is shorter now (from 2882 LOC to 2351). * The parser is just ordinary C++ code that can be debugged in the usual way. * Memory management in the parser is now straightforward, as parsing methods simply return a unique pointer or vector. This required a couple of minor changes to the test suite, as some errors have changed. While this passes the tests, it's possible there are lurking bugs, particularly around error handling. gdb/ChangeLog 2021-04-16 Tom Tromey * rust-parse.c: New file. * rust-exp.y: Remove. * Makefile.in (COMMON_SFILES): Add rust-parse.c. (SFILES): Remove rust-exp.y. (YYFILES, local-maintainer-clean): Remove rust-exp.c. gdb/testsuite/ChangeLog 2021-04-16 Tom Tromey * gdb.rust/simple.exp: Change error text. * gdb.rust/expr.exp: Change error text. --- diff --git a/gdb/ChangeLog b/gdb/ChangeLog index c8630a43014..d5b6c309084 100644 --- a/gdb/ChangeLog +++ b/gdb/ChangeLog @@ -1,3 +1,11 @@ +2021-04-16 Tom Tromey + + * rust-parse.c: New file. + * rust-exp.y: Remove. + * Makefile.in (COMMON_SFILES): Add rust-parse.c. + (SFILES): Remove rust-exp.y. + (YYFILES, local-maintainer-clean): Remove rust-exp.c. + 2021-04-16 Luis Machado * arch-utils.c (default_floatformat_for_type): Handle bfloat16. diff --git a/gdb/Makefile.in b/gdb/Makefile.in index 3318c1a5215..490419030a3 100644 --- a/gdb/Makefile.in +++ b/gdb/Makefile.in @@ -1134,6 +1134,7 @@ COMMON_SFILES = \ reverse.c \ run-on-main-thread.c \ rust-lang.c \ + rust-parse.c \ sentinel-frame.c \ ser-event.c \ serial.c \ @@ -1200,7 +1201,6 @@ SFILES = \ m2-exp.y \ p-exp.y \ proc-service.list \ - rust-exp.y \ ser-base.c \ ser-unix.c \ sol-thread.c \ @@ -1608,8 +1608,7 @@ YYFILES = \ f-exp.c \ go-exp.c \ m2-exp.c \ - p-exp.c \ - rust-exp.c + p-exp.c # ada-lex.c is included by another file, so it shouldn't wind up as a # .o itself. @@ -1969,7 +1968,7 @@ local-maintainer-clean: rm -f c-exp.c \ cp-name-parser.c \ ada-lex.c ada-exp.c \ - d-exp.c f-exp.c go-exp.c m2-exp.c p-exp.c rust-exp.c + d-exp.c f-exp.c go-exp.c m2-exp.c p-exp.c rm -f TAGS rm -f $(YYFILES) rm -f nm.h config.status diff --git a/gdb/rust-exp.y b/gdb/rust-exp.y deleted file mode 100644 index a0ceb6a42c7..00000000000 --- a/gdb/rust-exp.y +++ /dev/null @@ -1,2884 +0,0 @@ -/* Bison parser for Rust expressions, for GDB. - Copyright (C) 2016-2021 Free Software Foundation, Inc. - - This file is part of GDB. - - This program is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . */ - -/* The Bison manual says that %pure-parser is deprecated, but we use - it anyway because it also works with Byacc. That is also why - this uses %lex-param and %parse-param rather than the simpler - %param -- Byacc does not support the latter. */ -%pure-parser -%lex-param {struct rust_parser *parser} -%parse-param {struct rust_parser *parser} - -/* Removing the last conflict seems difficult. */ -%expect 1 - -%{ - -#include "defs.h" - -#include "block.h" -#include "charset.h" -#include "cp-support.h" -#include "gdb_obstack.h" -#include "gdb_regex.h" -#include "rust-lang.h" -#include "parser-defs.h" -#include "gdbsupport/selftest.h" -#include "value.h" -#include "gdbarch.h" -#include "rust-exp.h" -#include -#include "gdbsupport/hash_enum.h" - -#define GDB_YY_REMAP_PREFIX rust -#include "yy-remap.h" - -#define RUSTSTYPE YYSTYPE - -struct rust_op; -typedef std::vector rust_op_vector; - -/* A typed integer constant. */ - -struct typed_val_int -{ - LONGEST val; - struct type *type; -}; - -/* A typed floating point constant. */ - -struct typed_val_float -{ - gdb_byte val[16]; - struct type *type; -}; - -/* An identifier and an expression. This is used to represent one - element of a struct initializer. */ - -struct set_field -{ - struct stoken name; - const struct rust_op *init; -}; - -typedef std::vector rust_set_vector; - -%} - -%union -{ - /* A typed integer constant. */ - struct typed_val_int typed_val_int; - - /* A typed floating point constant. */ - struct typed_val_float typed_val_float; - - /* An identifier or string. */ - struct stoken sval; - - /* A token representing an opcode, like "==". */ - enum exp_opcode opcode; - - /* A list of expressions; for example, the arguments to a function - call. */ - rust_op_vector *params; - - /* A list of field initializers. */ - rust_set_vector *field_inits; - - /* A single field initializer. */ - struct set_field one_field_init; - - /* An expression. */ - const struct rust_op *op; - - /* A plain integer, for example used to count the number of - "super::" prefixes on a path. */ - unsigned int depth; -} - -%{ - -struct rust_parser; -static int rustyylex (YYSTYPE *, rust_parser *); -static void rustyyerror (rust_parser *parser, const char *msg); - -static struct stoken make_stoken (const char *); - -/* A regular expression for matching Rust numbers. This is split up - since it is very long and this gives us a way to comment the - sections. */ - -static const char number_regex_text[] = - /* subexpression 1: allows use of alternation, otherwise uninteresting */ - "^(" - /* First comes floating point. */ - /* Recognize number after the decimal point, with optional - exponent and optional type suffix. - subexpression 2: allows "?", otherwise uninteresting - subexpression 3: if present, type suffix - */ - "[0-9][0-9_]*\\.[0-9][0-9_]*([eE][-+]?[0-9][0-9_]*)?(f32|f64)?" -#define FLOAT_TYPE1 3 - "|" - /* Recognize exponent without decimal point, with optional type - suffix. - subexpression 4: if present, type suffix - */ -#define FLOAT_TYPE2 4 - "[0-9][0-9_]*[eE][-+]?[0-9][0-9_]*(f32|f64)?" - "|" - /* "23." is a valid floating point number, but "23.e5" and - "23.f32" are not. So, handle the trailing-. case - separately. */ - "[0-9][0-9_]*\\." - "|" - /* Finally come integers. - subexpression 5: text of integer - subexpression 6: if present, type suffix - subexpression 7: allows use of alternation, otherwise uninteresting - */ -#define INT_TEXT 5 -#define INT_TYPE 6 - "(0x[a-fA-F0-9_]+|0o[0-7_]+|0b[01_]+|[0-9][0-9_]*)" - "([iu](size|8|16|32|64))?" - ")"; -/* The number of subexpressions to allocate space for, including the - "0th" whole match subexpression. */ -#define NUM_SUBEXPRESSIONS 8 - -/* The compiled number-matching regex. */ - -static regex_t number_regex; - -/* An instance of this is created before parsing, and destroyed when - parsing is finished. */ - -struct rust_parser -{ - rust_parser (struct parser_state *state) - : rust_ast (nullptr), - pstate (state) - { - } - - ~rust_parser () - { - } - - /* Create a new rust_set_vector. The storage for the new vector is - managed by this class. */ - rust_set_vector *new_set_vector () - { - rust_set_vector *result = new rust_set_vector; - set_vectors.push_back (std::unique_ptr (result)); - return result; - } - - /* Create a new rust_ops_vector. The storage for the new vector is - managed by this class. */ - rust_op_vector *new_op_vector () - { - rust_op_vector *result = new rust_op_vector; - op_vectors.push_back (std::unique_ptr (result)); - return result; - } - - /* Return the parser's language. */ - const struct language_defn *language () const - { - return pstate->language (); - } - - /* Return the parser's gdbarch. */ - struct gdbarch *arch () const - { - return pstate->gdbarch (); - } - - /* A helper to look up a Rust type, or fail. This only works for - types defined by rust_language_arch_info. */ - - struct type *get_type (const char *name) - { - struct type *type; - - type = language_lookup_primitive_type (language (), arch (), name); - if (type == NULL) - error (_("Could not find Rust type %s"), name); - return type; - } - - const char *copy_name (const char *name, int len); - struct stoken concat3 (const char *s1, const char *s2, const char *s3); - const struct rust_op *crate_name (const struct rust_op *name); - const struct rust_op *super_name (const struct rust_op *ident, - unsigned int n_supers); - - int lex_character (YYSTYPE *lvalp); - int lex_number (YYSTYPE *lvalp); - int lex_string (YYSTYPE *lvalp); - int lex_identifier (YYSTYPE *lvalp); - uint32_t lex_hex (int min, int max); - uint32_t lex_escape (int is_byte); - int lex_operator (YYSTYPE *lvalp); - void push_back (char c); - - void update_innermost_block (struct block_symbol sym); - struct block_symbol lookup_symbol (const char *name, - const struct block *block, - const domain_enum domain); - struct type *rust_lookup_type (const char *name, const struct block *block); - std::vector convert_params_to_types (rust_op_vector *params); - struct type *convert_ast_to_type (const struct rust_op *operation); - const char *convert_name (const struct rust_op *operation); - std::vector convert_params_to_expression - (rust_op_vector *params, const struct rust_op *top); - expr::operation_up convert_ast_to_expression (const struct rust_op *opn, - const struct rust_op *top, - bool want_type = false); - - struct rust_op *ast_basic_type (enum type_code typecode); - const struct rust_op *ast_operation (enum exp_opcode opcode, - const struct rust_op *left, - const struct rust_op *right); - const struct rust_op *ast_compound_assignment - (enum exp_opcode opcode, const struct rust_op *left, - const struct rust_op *rust_op); - const struct rust_op *ast_literal (struct typed_val_int val); - const struct rust_op *ast_dliteral (struct typed_val_float val); - const struct rust_op *ast_structop (const struct rust_op *left, - const char *name, - int completing); - const struct rust_op *ast_structop_anonymous - (const struct rust_op *left, struct typed_val_int number); - const struct rust_op *ast_unary (enum exp_opcode opcode, - const struct rust_op *expr); - const struct rust_op *ast_cast (const struct rust_op *expr, - const struct rust_op *type); - const struct rust_op *ast_call_ish (enum exp_opcode opcode, - const struct rust_op *expr, - rust_op_vector *params); - const struct rust_op *ast_path (struct stoken name, - rust_op_vector *params); - const struct rust_op *ast_string (struct stoken str); - const struct rust_op *ast_struct (const struct rust_op *name, - rust_set_vector *fields); - const struct rust_op *ast_range (const struct rust_op *lhs, - const struct rust_op *rhs, - bool inclusive); - const struct rust_op *ast_array_type (const struct rust_op *lhs, - struct typed_val_int val); - const struct rust_op *ast_slice_type (const struct rust_op *type); - const struct rust_op *ast_reference_type (const struct rust_op *type); - const struct rust_op *ast_pointer_type (const struct rust_op *type, - int is_mut); - const struct rust_op *ast_function_type (const struct rust_op *result, - rust_op_vector *params); - const struct rust_op *ast_tuple_type (rust_op_vector *params); - - - /* A pointer to this is installed globally. */ - auto_obstack obstack; - - /* Result of parsing. Points into obstack. */ - const struct rust_op *rust_ast; - - /* This keeps track of the various vectors we allocate. */ - std::vector> set_vectors; - std::vector> op_vectors; - - /* The parser state gdb gave us. */ - struct parser_state *pstate; - - /* Depth of parentheses. */ - int paren_depth = 0; -}; - -/* Rust AST operations. We build a tree of these; then lower them to - gdb expressions when parsing has completed. */ - -struct rust_op -{ - /* The opcode. */ - enum exp_opcode opcode; - /* If OPCODE is OP_TYPE, then this holds information about what type - is described by this node. */ - enum type_code typecode; - /* Indicates whether OPCODE actually represents a compound - assignment. For example, if OPCODE is GTGT and this is false, - then this rust_op represents an ordinary ">>"; but if this is - true, then this rust_op represents ">>=". Unused in other - cases. */ - unsigned int compound_assignment : 1; - /* Only used by a field expression; if set, indicates that the field - name occurred at the end of the expression and is eligible for - completion. */ - unsigned int completing : 1; - /* For OP_RANGE, indicates whether the range is inclusive or - exclusive. */ - unsigned int inclusive : 1; - /* Operands of expression. Which one is used and how depends on the - particular opcode. */ - RUSTSTYPE left; - RUSTSTYPE right; -}; - -%} - -%token GDBVAR -%token IDENT -%token COMPLETE -%token INTEGER -%token DECIMAL_INTEGER -%token STRING -%token BYTESTRING -%token FLOAT -%token COMPOUND_ASSIGN - -/* Keyword tokens. */ -%token KW_AS -%token KW_IF -%token KW_TRUE -%token KW_FALSE -%token KW_SUPER -%token KW_SELF -%token KW_MUT -%token KW_EXTERN -%token KW_CONST -%token KW_FN -%token KW_SIZEOF - -/* Operator tokens. */ -%token DOTDOT -%token DOTDOTEQ -%token OROR -%token ANDAND -%token EQEQ -%token NOTEQ -%token LTEQ -%token GTEQ -%token LSH RSH -%token COLONCOLON -%token ARROW - -%type type -%type path_for_expr -%type identifier_path_for_expr -%type path_for_type -%type identifier_path_for_type -%type just_identifiers_for_type - -%type maybe_type_list -%type type_list - -%type super_path - -%type literal -%type expr -%type field_expr -%type idx_expr -%type unop_expr -%type binop_expr -%type binop_expr_expr -%type type_cast_expr -%type assignment_expr -%type compound_assignment_expr -%type paren_expr -%type call_expr -%type path_expr -%type tuple_expr -%type unit_expr -%type struct_expr -%type array_expr -%type range_expr - -%type expr_list -%type maybe_expr_list -%type paren_expr_list - -%type struct_expr_list -%type struct_expr_tail - -/* Precedence. */ -%nonassoc DOTDOT DOTDOTEQ -%right '=' COMPOUND_ASSIGN -%left OROR -%left ANDAND -%nonassoc EQEQ NOTEQ '<' '>' LTEQ GTEQ -%left '|' -%left '^' -%left '&' -%left LSH RSH -%left '@' -%left '+' '-' -%left '*' '/' '%' -/* These could be %precedence in Bison, but that isn't a yacc - feature. */ -%left KW_AS -%left UNARY -%left '[' '.' '(' - -%% - -start: - expr - { - /* If we are completing and see a valid parse, - rust_ast will already have been set. */ - if (parser->rust_ast == NULL) - parser->rust_ast = $1; - } -; - -/* Note that the Rust grammar includes a method_call_expr, but we - handle this differently, to avoid a shift/reduce conflict with - call_expr. */ -expr: - literal -| path_expr -| tuple_expr -| unit_expr -| struct_expr -| field_expr -| array_expr -| idx_expr -| range_expr -| unop_expr /* Must precede call_expr because of ambiguity with - sizeof. */ -| binop_expr -| paren_expr -| call_expr -; - -tuple_expr: - '(' expr ',' maybe_expr_list ')' - { - $4->push_back ($2); - error (_("Tuple expressions not supported yet")); - } -; - -unit_expr: - '(' ')' - { - struct typed_val_int val; - - val.type - = (language_lookup_primitive_type - (parser->language (), parser->arch (), - "()")); - val.val = 0; - $$ = parser->ast_literal (val); - } -; - -/* To avoid a shift/reduce conflict with call_expr, we don't handle - tuple struct expressions here, but instead when examining the - AST. */ -struct_expr: - path_for_expr '{' struct_expr_list '}' - { $$ = parser->ast_struct ($1, $3); } -; - -struct_expr_tail: - DOTDOT expr - { - struct set_field sf; - - sf.name.ptr = NULL; - sf.name.length = 0; - sf.init = $2; - - $$ = sf; - } -| IDENT ':' expr - { - struct set_field sf; - - sf.name = $1; - sf.init = $3; - $$ = sf; - } -| IDENT - { - struct set_field sf; - - sf.name = $1; - sf.init = parser->ast_path ($1, NULL); - $$ = sf; - } -; - -struct_expr_list: - /* %empty */ - { - $$ = parser->new_set_vector (); - } -| struct_expr_tail - { - rust_set_vector *result = parser->new_set_vector (); - result->push_back ($1); - $$ = result; - } -| IDENT ':' expr ',' struct_expr_list - { - struct set_field sf; - - sf.name = $1; - sf.init = $3; - $5->push_back (sf); - $$ = $5; - } -| IDENT ',' struct_expr_list - { - struct set_field sf; - - sf.name = $1; - sf.init = parser->ast_path ($1, NULL); - $3->push_back (sf); - $$ = $3; - } -; - -array_expr: - '[' KW_MUT expr_list ']' - { $$ = parser->ast_call_ish (OP_ARRAY, NULL, $3); } -| '[' expr_list ']' - { $$ = parser->ast_call_ish (OP_ARRAY, NULL, $2); } -| '[' KW_MUT expr ';' expr ']' - { $$ = parser->ast_operation (OP_RUST_ARRAY, $3, $5); } -| '[' expr ';' expr ']' - { $$ = parser->ast_operation (OP_RUST_ARRAY, $2, $4); } -; - -range_expr: - expr DOTDOT - { $$ = parser->ast_range ($1, NULL, false); } -| expr DOTDOT expr - { $$ = parser->ast_range ($1, $3, false); } -| expr DOTDOTEQ expr - { $$ = parser->ast_range ($1, $3, true); } -| DOTDOT expr - { $$ = parser->ast_range (NULL, $2, false); } -| DOTDOTEQ expr - { $$ = parser->ast_range (NULL, $2, true); } -| DOTDOT - { $$ = parser->ast_range (NULL, NULL, false); } -; - -literal: - INTEGER - { $$ = parser->ast_literal ($1); } -| DECIMAL_INTEGER - { $$ = parser->ast_literal ($1); } -| FLOAT - { $$ = parser->ast_dliteral ($1); } -| STRING - { - struct set_field field; - struct typed_val_int val; - struct stoken token; - - rust_set_vector *fields = parser->new_set_vector (); - - /* Wrap the raw string in the &str struct. */ - field.name.ptr = "data_ptr"; - field.name.length = strlen (field.name.ptr); - field.init = parser->ast_unary (UNOP_ADDR, - parser->ast_string ($1)); - fields->push_back (field); - - val.type = parser->get_type ("usize"); - val.val = $1.length; - - field.name.ptr = "length"; - field.name.length = strlen (field.name.ptr); - field.init = parser->ast_literal (val); - fields->push_back (field); - - token.ptr = "&str"; - token.length = strlen (token.ptr); - $$ = parser->ast_struct (parser->ast_path (token, NULL), - fields); - } -| BYTESTRING - { $$ = parser->ast_string ($1); } -| KW_TRUE - { - struct typed_val_int val; - - val.type = language_bool_type (parser->language (), - parser->arch ()); - val.val = 1; - $$ = parser->ast_literal (val); - } -| KW_FALSE - { - struct typed_val_int val; - - val.type = language_bool_type (parser->language (), - parser->arch ()); - val.val = 0; - $$ = parser->ast_literal (val); - } -; - -field_expr: - expr '.' IDENT - { $$ = parser->ast_structop ($1, $3.ptr, 0); } -| expr '.' COMPLETE - { - $$ = parser->ast_structop ($1, $3.ptr, 1); - parser->rust_ast = $$; - } -| expr '.' DECIMAL_INTEGER - { $$ = parser->ast_structop_anonymous ($1, $3); } -; - -idx_expr: - expr '[' expr ']' - { $$ = parser->ast_operation (BINOP_SUBSCRIPT, $1, $3); } -; - -unop_expr: - '+' expr %prec UNARY - { $$ = parser->ast_unary (UNOP_PLUS, $2); } - -| '-' expr %prec UNARY - { $$ = parser->ast_unary (UNOP_NEG, $2); } - -| '!' expr %prec UNARY - { - /* Note that we provide a Rust-specific evaluator - override for UNOP_COMPLEMENT, so it can do the - right thing for both bool and integral - values. */ - $$ = parser->ast_unary (UNOP_COMPLEMENT, $2); - } - -| '*' expr %prec UNARY - { $$ = parser->ast_unary (UNOP_IND, $2); } - -| '&' expr %prec UNARY - { $$ = parser->ast_unary (UNOP_ADDR, $2); } - -| '&' KW_MUT expr %prec UNARY - { $$ = parser->ast_unary (UNOP_ADDR, $3); } -| KW_SIZEOF '(' expr ')' %prec UNARY - { $$ = parser->ast_unary (UNOP_SIZEOF, $3); } -; - -binop_expr: - binop_expr_expr -| type_cast_expr -| assignment_expr -| compound_assignment_expr -; - -binop_expr_expr: - expr '*' expr - { $$ = parser->ast_operation (BINOP_MUL, $1, $3); } - -| expr '@' expr - { $$ = parser->ast_operation (BINOP_REPEAT, $1, $3); } - -| expr '/' expr - { $$ = parser->ast_operation (BINOP_DIV, $1, $3); } - -| expr '%' expr - { $$ = parser->ast_operation (BINOP_REM, $1, $3); } - -| expr '<' expr - { $$ = parser->ast_operation (BINOP_LESS, $1, $3); } - -| expr '>' expr - { $$ = parser->ast_operation (BINOP_GTR, $1, $3); } - -| expr '&' expr - { $$ = parser->ast_operation (BINOP_BITWISE_AND, $1, $3); } - -| expr '|' expr - { $$ = parser->ast_operation (BINOP_BITWISE_IOR, $1, $3); } - -| expr '^' expr - { $$ = parser->ast_operation (BINOP_BITWISE_XOR, $1, $3); } - -| expr '+' expr - { $$ = parser->ast_operation (BINOP_ADD, $1, $3); } - -| expr '-' expr - { $$ = parser->ast_operation (BINOP_SUB, $1, $3); } - -| expr OROR expr - { $$ = parser->ast_operation (BINOP_LOGICAL_OR, $1, $3); } - -| expr ANDAND expr - { $$ = parser->ast_operation (BINOP_LOGICAL_AND, $1, $3); } - -| expr EQEQ expr - { $$ = parser->ast_operation (BINOP_EQUAL, $1, $3); } - -| expr NOTEQ expr - { $$ = parser->ast_operation (BINOP_NOTEQUAL, $1, $3); } - -| expr LTEQ expr - { $$ = parser->ast_operation (BINOP_LEQ, $1, $3); } - -| expr GTEQ expr - { $$ = parser->ast_operation (BINOP_GEQ, $1, $3); } - -| expr LSH expr - { $$ = parser->ast_operation (BINOP_LSH, $1, $3); } - -| expr RSH expr - { $$ = parser->ast_operation (BINOP_RSH, $1, $3); } -; - -type_cast_expr: - expr KW_AS type - { $$ = parser->ast_cast ($1, $3); } -; - -assignment_expr: - expr '=' expr - { $$ = parser->ast_operation (BINOP_ASSIGN, $1, $3); } -; - -compound_assignment_expr: - expr COMPOUND_ASSIGN expr - { $$ = parser->ast_compound_assignment ($2, $1, $3); } - -; - -paren_expr: - '(' expr ')' - { $$ = $2; } -; - -expr_list: - expr - { - $$ = parser->new_op_vector (); - $$->push_back ($1); - } -| expr_list ',' expr - { - $1->push_back ($3); - $$ = $1; - } -; - -maybe_expr_list: - /* %empty */ - { - /* The result can't be NULL. */ - $$ = parser->new_op_vector (); - } -| expr_list - { $$ = $1; } -; - -paren_expr_list: - '(' maybe_expr_list ')' - { $$ = $2; } -; - -call_expr: - expr paren_expr_list - { $$ = parser->ast_call_ish (OP_FUNCALL, $1, $2); } -; - -maybe_self_path: - /* %empty */ -| KW_SELF COLONCOLON -; - -super_path: - KW_SUPER COLONCOLON - { $$ = 1; } -| super_path KW_SUPER COLONCOLON - { $$ = $1 + 1; } -; - -path_expr: - path_for_expr - { $$ = $1; } -| GDBVAR - { $$ = parser->ast_path ($1, NULL); } -| KW_SELF - { $$ = parser->ast_path (make_stoken ("self"), NULL); } -; - -path_for_expr: - identifier_path_for_expr -| KW_SELF COLONCOLON identifier_path_for_expr - { $$ = parser->super_name ($3, 0); } -| maybe_self_path super_path identifier_path_for_expr - { $$ = parser->super_name ($3, $2); } -| COLONCOLON identifier_path_for_expr - { $$ = parser->crate_name ($2); } -| KW_EXTERN identifier_path_for_expr - { - /* This is a gdb extension to make it possible to - refer to items in other crates. It just bypasses - adding the current crate to the front of the - name. */ - $$ = parser->ast_path (parser->concat3 ("::", - $2->left.sval.ptr, - NULL), - $2->right.params); - } -; - -identifier_path_for_expr: - IDENT - { $$ = parser->ast_path ($1, NULL); } -| identifier_path_for_expr COLONCOLON IDENT - { - $$ = parser->ast_path (parser->concat3 ($1->left.sval.ptr, - "::", $3.ptr), - NULL); - } -| identifier_path_for_expr COLONCOLON '<' type_list '>' - { $$ = parser->ast_path ($1->left.sval, $4); } -| identifier_path_for_expr COLONCOLON '<' type_list RSH - { - $$ = parser->ast_path ($1->left.sval, $4); - parser->push_back ('>'); - } -; - -path_for_type: - identifier_path_for_type -| KW_SELF COLONCOLON identifier_path_for_type - { $$ = parser->super_name ($3, 0); } -| maybe_self_path super_path identifier_path_for_type - { $$ = parser->super_name ($3, $2); } -| COLONCOLON identifier_path_for_type - { $$ = parser->crate_name ($2); } -| KW_EXTERN identifier_path_for_type - { - /* This is a gdb extension to make it possible to - refer to items in other crates. It just bypasses - adding the current crate to the front of the - name. */ - $$ = parser->ast_path (parser->concat3 ("::", - $2->left.sval.ptr, - NULL), - $2->right.params); - } -; - -just_identifiers_for_type: - IDENT - { $$ = parser->ast_path ($1, NULL); } -| just_identifiers_for_type COLONCOLON IDENT - { - $$ = parser->ast_path (parser->concat3 ($1->left.sval.ptr, - "::", $3.ptr), - NULL); - } -; - -identifier_path_for_type: - just_identifiers_for_type -| just_identifiers_for_type '<' type_list '>' - { $$ = parser->ast_path ($1->left.sval, $3); } -| just_identifiers_for_type '<' type_list RSH - { - $$ = parser->ast_path ($1->left.sval, $3); - parser->push_back ('>'); - } -; - -type: - path_for_type -| '[' type ';' INTEGER ']' - { $$ = parser->ast_array_type ($2, $4); } -| '[' type ';' DECIMAL_INTEGER ']' - { $$ = parser->ast_array_type ($2, $4); } -| '&' '[' type ']' - { $$ = parser->ast_slice_type ($3); } -| '&' type - { $$ = parser->ast_reference_type ($2); } -| '*' KW_MUT type - { $$ = parser->ast_pointer_type ($3, 1); } -| '*' KW_CONST type - { $$ = parser->ast_pointer_type ($3, 0); } -| KW_FN '(' maybe_type_list ')' ARROW type - { $$ = parser->ast_function_type ($6, $3); } -| '(' maybe_type_list ')' - { $$ = parser->ast_tuple_type ($2); } -; - -maybe_type_list: - /* %empty */ - { $$ = NULL; } -| type_list - { $$ = $1; } -; - -type_list: - type - { - rust_op_vector *result = parser->new_op_vector (); - result->push_back ($1); - $$ = result; - } -| type_list ',' type - { - $1->push_back ($3); - $$ = $1; - } -; - -%% - -/* A struct of this type is used to describe a token. */ - -struct token_info -{ - const char *name; - int value; - enum exp_opcode opcode; -}; - -/* Identifier tokens. */ - -static const struct token_info identifier_tokens[] = -{ - { "as", KW_AS, OP_NULL }, - { "false", KW_FALSE, OP_NULL }, - { "if", 0, OP_NULL }, - { "mut", KW_MUT, OP_NULL }, - { "const", KW_CONST, OP_NULL }, - { "self", KW_SELF, OP_NULL }, - { "super", KW_SUPER, OP_NULL }, - { "true", KW_TRUE, OP_NULL }, - { "extern", KW_EXTERN, OP_NULL }, - { "fn", KW_FN, OP_NULL }, - { "sizeof", KW_SIZEOF, OP_NULL }, -}; - -/* Operator tokens, sorted longest first. */ - -static const struct token_info operator_tokens[] = -{ - { ">>=", COMPOUND_ASSIGN, BINOP_RSH }, - { "<<=", COMPOUND_ASSIGN, BINOP_LSH }, - - { "<<", LSH, OP_NULL }, - { ">>", RSH, OP_NULL }, - { "&&", ANDAND, OP_NULL }, - { "||", OROR, OP_NULL }, - { "==", EQEQ, OP_NULL }, - { "!=", NOTEQ, OP_NULL }, - { "<=", LTEQ, OP_NULL }, - { ">=", GTEQ, OP_NULL }, - { "+=", COMPOUND_ASSIGN, BINOP_ADD }, - { "-=", COMPOUND_ASSIGN, BINOP_SUB }, - { "*=", COMPOUND_ASSIGN, BINOP_MUL }, - { "/=", COMPOUND_ASSIGN, BINOP_DIV }, - { "%=", COMPOUND_ASSIGN, BINOP_REM }, - { "&=", COMPOUND_ASSIGN, BINOP_BITWISE_AND }, - { "|=", COMPOUND_ASSIGN, BINOP_BITWISE_IOR }, - { "^=", COMPOUND_ASSIGN, BINOP_BITWISE_XOR }, - { "..=", DOTDOTEQ, OP_NULL }, - - { "::", COLONCOLON, OP_NULL }, - { "..", DOTDOT, OP_NULL }, - { "->", ARROW, OP_NULL } -}; - -/* Helper function to copy to the name obstack. */ - -const char * -rust_parser::copy_name (const char *name, int len) -{ - return obstack_strndup (&obstack, name, len); -} - -/* Helper function to make an stoken from a C string. */ - -static struct stoken -make_stoken (const char *p) -{ - struct stoken result; - - result.ptr = p; - result.length = strlen (result.ptr); - return result; -} - -/* Helper function to concatenate three strings on the name - obstack. */ - -struct stoken -rust_parser::concat3 (const char *s1, const char *s2, const char *s3) -{ - return make_stoken (obconcat (&obstack, s1, s2, s3, (char *) NULL)); -} - -/* Return an AST node referring to NAME, but relative to the crate's - name. */ - -const struct rust_op * -rust_parser::crate_name (const struct rust_op *name) -{ - std::string crate = rust_crate_for_block (pstate->expression_context_block); - struct stoken result; - - gdb_assert (name->opcode == OP_VAR_VALUE); - - if (crate.empty ()) - error (_("Could not find crate for current location")); - result = make_stoken (obconcat (&obstack, "::", crate.c_str (), "::", - name->left.sval.ptr, (char *) NULL)); - - return ast_path (result, name->right.params); -} - -/* Create an AST node referring to a "super::" qualified name. IDENT - is the base name and N_SUPERS is how many "super::"s were - provided. N_SUPERS can be zero. */ - -const struct rust_op * -rust_parser::super_name (const struct rust_op *ident, unsigned int n_supers) -{ - const char *scope = block_scope (pstate->expression_context_block); - int offset; - - gdb_assert (ident->opcode == OP_VAR_VALUE); - - if (scope[0] == '\0') - error (_("Couldn't find namespace scope for self::")); - - if (n_supers > 0) - { - int len; - std::vector offsets; - unsigned int current_len; - - current_len = cp_find_first_component (scope); - while (scope[current_len] != '\0') - { - offsets.push_back (current_len); - gdb_assert (scope[current_len] == ':'); - /* The "::". */ - current_len += 2; - current_len += cp_find_first_component (scope - + current_len); - } - - len = offsets.size (); - if (n_supers >= len) - error (_("Too many super:: uses from '%s'"), scope); - - offset = offsets[len - n_supers]; - } - else - offset = strlen (scope); - - obstack_grow (&obstack, "::", 2); - obstack_grow (&obstack, scope, offset); - obstack_grow (&obstack, "::", 2); - obstack_grow0 (&obstack, ident->left.sval.ptr, ident->left.sval.length); - - return ast_path (make_stoken ((const char *) obstack_finish (&obstack)), - ident->right.params); -} - -/* A helper that updates the innermost block as appropriate. */ - -void -rust_parser::update_innermost_block (struct block_symbol sym) -{ - if (symbol_read_needs_frame (sym.symbol)) - pstate->block_tracker->update (sym); -} - -/* Lex a hex number with at least MIN digits and at most MAX - digits. */ - -uint32_t -rust_parser::lex_hex (int min, int max) -{ - uint32_t result = 0; - int len = 0; - /* We only want to stop at MAX if we're lexing a byte escape. */ - int check_max = min == max; - - while ((check_max ? len <= max : 1) - && ((pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'f') - || (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'F') - || (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9'))) - { - result *= 16; - if (pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'f') - result = result + 10 + pstate->lexptr[0] - 'a'; - else if (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'F') - result = result + 10 + pstate->lexptr[0] - 'A'; - else - result = result + pstate->lexptr[0] - '0'; - ++pstate->lexptr; - ++len; - } - - if (len < min) - error (_("Not enough hex digits seen")); - if (len > max) - { - gdb_assert (min != max); - error (_("Overlong hex escape")); - } - - return result; -} - -/* Lex an escape. IS_BYTE is true if we're lexing a byte escape; - otherwise we're lexing a character escape. */ - -uint32_t -rust_parser::lex_escape (int is_byte) -{ - uint32_t result; - - gdb_assert (pstate->lexptr[0] == '\\'); - ++pstate->lexptr; - switch (pstate->lexptr[0]) - { - case 'x': - ++pstate->lexptr; - result = lex_hex (2, 2); - break; - - case 'u': - if (is_byte) - error (_("Unicode escape in byte literal")); - ++pstate->lexptr; - if (pstate->lexptr[0] != '{') - error (_("Missing '{' in Unicode escape")); - ++pstate->lexptr; - result = lex_hex (1, 6); - /* Could do range checks here. */ - if (pstate->lexptr[0] != '}') - error (_("Missing '}' in Unicode escape")); - ++pstate->lexptr; - break; - - case 'n': - result = '\n'; - ++pstate->lexptr; - break; - case 'r': - result = '\r'; - ++pstate->lexptr; - break; - case 't': - result = '\t'; - ++pstate->lexptr; - break; - case '\\': - result = '\\'; - ++pstate->lexptr; - break; - case '0': - result = '\0'; - ++pstate->lexptr; - break; - case '\'': - result = '\''; - ++pstate->lexptr; - break; - case '"': - result = '"'; - ++pstate->lexptr; - break; - - default: - error (_("Invalid escape \\%c in literal"), pstate->lexptr[0]); - } - - return result; -} - -/* Lex a character constant. */ - -int -rust_parser::lex_character (YYSTYPE *lvalp) -{ - int is_byte = 0; - uint32_t value; - - if (pstate->lexptr[0] == 'b') - { - is_byte = 1; - ++pstate->lexptr; - } - gdb_assert (pstate->lexptr[0] == '\''); - ++pstate->lexptr; - /* This should handle UTF-8 here. */ - if (pstate->lexptr[0] == '\\') - value = lex_escape (is_byte); - else - { - value = pstate->lexptr[0] & 0xff; - ++pstate->lexptr; - } - - if (pstate->lexptr[0] != '\'') - error (_("Unterminated character literal")); - ++pstate->lexptr; - - lvalp->typed_val_int.val = value; - lvalp->typed_val_int.type = get_type (is_byte ? "u8" : "char"); - - return INTEGER; -} - -/* Return the offset of the double quote if STR looks like the start - of a raw string, or 0 if STR does not start a raw string. */ - -static int -starts_raw_string (const char *str) -{ - const char *save = str; - - if (str[0] != 'r') - return 0; - ++str; - while (str[0] == '#') - ++str; - if (str[0] == '"') - return str - save; - return 0; -} - -/* Return true if STR looks like the end of a raw string that had N - hashes at the start. */ - -static bool -ends_raw_string (const char *str, int n) -{ - int i; - - gdb_assert (str[0] == '"'); - for (i = 0; i < n; ++i) - if (str[i + 1] != '#') - return false; - return true; -} - -/* Lex a string constant. */ - -int -rust_parser::lex_string (YYSTYPE *lvalp) -{ - int is_byte = pstate->lexptr[0] == 'b'; - int raw_length; - - if (is_byte) - ++pstate->lexptr; - raw_length = starts_raw_string (pstate->lexptr); - pstate->lexptr += raw_length; - gdb_assert (pstate->lexptr[0] == '"'); - ++pstate->lexptr; - - while (1) - { - uint32_t value; - - if (raw_length > 0) - { - if (pstate->lexptr[0] == '"' && ends_raw_string (pstate->lexptr, - raw_length - 1)) - { - /* Exit with lexptr pointing after the final "#". */ - pstate->lexptr += raw_length; - break; - } - else if (pstate->lexptr[0] == '\0') - error (_("Unexpected EOF in string")); - - value = pstate->lexptr[0] & 0xff; - if (is_byte && value > 127) - error (_("Non-ASCII value in raw byte string")); - obstack_1grow (&obstack, value); - - ++pstate->lexptr; - } - else if (pstate->lexptr[0] == '"') - { - /* Make sure to skip the quote. */ - ++pstate->lexptr; - break; - } - else if (pstate->lexptr[0] == '\\') - { - value = lex_escape (is_byte); - - if (is_byte) - obstack_1grow (&obstack, value); - else - convert_between_encodings ("UTF-32", "UTF-8", (gdb_byte *) &value, - sizeof (value), sizeof (value), - &obstack, translit_none); - } - else if (pstate->lexptr[0] == '\0') - error (_("Unexpected EOF in string")); - else - { - value = pstate->lexptr[0] & 0xff; - if (is_byte && value > 127) - error (_("Non-ASCII value in byte string")); - obstack_1grow (&obstack, value); - ++pstate->lexptr; - } - } - - lvalp->sval.length = obstack_object_size (&obstack); - lvalp->sval.ptr = (const char *) obstack_finish (&obstack); - return is_byte ? BYTESTRING : STRING; -} - -/* Return true if STRING starts with whitespace followed by a digit. */ - -static bool -space_then_number (const char *string) -{ - const char *p = string; - - while (p[0] == ' ' || p[0] == '\t') - ++p; - if (p == string) - return false; - - return *p >= '0' && *p <= '9'; -} - -/* Return true if C can start an identifier. */ - -static bool -rust_identifier_start_p (char c) -{ - return ((c >= 'a' && c <= 'z') - || (c >= 'A' && c <= 'Z') - || c == '_' - || c == '$'); -} - -/* Lex an identifier. */ - -int -rust_parser::lex_identifier (YYSTYPE *lvalp) -{ - const char *start = pstate->lexptr; - unsigned int length; - const struct token_info *token; - int i; - int is_gdb_var = pstate->lexptr[0] == '$'; - - gdb_assert (rust_identifier_start_p (pstate->lexptr[0])); - - ++pstate->lexptr; - - /* For the time being this doesn't handle Unicode rules. Non-ASCII - identifiers are gated anyway. */ - while ((pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'z') - || (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'Z') - || pstate->lexptr[0] == '_' - || (is_gdb_var && pstate->lexptr[0] == '$') - || (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9')) - ++pstate->lexptr; - - - length = pstate->lexptr - start; - token = NULL; - for (i = 0; i < ARRAY_SIZE (identifier_tokens); ++i) - { - if (length == strlen (identifier_tokens[i].name) - && strncmp (identifier_tokens[i].name, start, length) == 0) - { - token = &identifier_tokens[i]; - break; - } - } - - if (token != NULL) - { - if (token->value == 0) - { - /* Leave the terminating token alone. */ - pstate->lexptr = start; - return 0; - } - } - else if (token == NULL - && (strncmp (start, "thread", length) == 0 - || strncmp (start, "task", length) == 0) - && space_then_number (pstate->lexptr)) - { - /* "task" or "thread" followed by a number terminates the - parse, per gdb rules. */ - pstate->lexptr = start; - return 0; - } - - if (token == NULL || (pstate->parse_completion && pstate->lexptr[0] == '\0')) - lvalp->sval = make_stoken (copy_name (start, length)); - - if (pstate->parse_completion && pstate->lexptr[0] == '\0') - { - /* Prevent rustyylex from returning two COMPLETE tokens. */ - pstate->prev_lexptr = pstate->lexptr; - return COMPLETE; - } - - if (token != NULL) - return token->value; - if (is_gdb_var) - return GDBVAR; - return IDENT; -} - -/* Lex an operator. */ - -int -rust_parser::lex_operator (YYSTYPE *lvalp) -{ - const struct token_info *token = NULL; - int i; - - for (i = 0; i < ARRAY_SIZE (operator_tokens); ++i) - { - if (strncmp (operator_tokens[i].name, pstate->lexptr, - strlen (operator_tokens[i].name)) == 0) - { - pstate->lexptr += strlen (operator_tokens[i].name); - token = &operator_tokens[i]; - break; - } - } - - if (token != NULL) - { - lvalp->opcode = token->opcode; - return token->value; - } - - return *pstate->lexptr++; -} - -/* Lex a number. */ - -int -rust_parser::lex_number (YYSTYPE *lvalp) -{ - regmatch_t subexps[NUM_SUBEXPRESSIONS]; - int match; - int is_integer = 0; - int could_be_decimal = 1; - int implicit_i32 = 0; - const char *type_name = NULL; - struct type *type; - int end_index; - int type_index = -1; - int i; - - match = regexec (&number_regex, pstate->lexptr, ARRAY_SIZE (subexps), - subexps, 0); - /* Failure means the regexp is broken. */ - gdb_assert (match == 0); - - if (subexps[INT_TEXT].rm_so != -1) - { - /* Integer part matched. */ - is_integer = 1; - end_index = subexps[INT_TEXT].rm_eo; - if (subexps[INT_TYPE].rm_so == -1) - { - type_name = "i32"; - implicit_i32 = 1; - } - else - { - type_index = INT_TYPE; - could_be_decimal = 0; - } - } - else if (subexps[FLOAT_TYPE1].rm_so != -1) - { - /* Found floating point type suffix. */ - end_index = subexps[FLOAT_TYPE1].rm_so; - type_index = FLOAT_TYPE1; - } - else if (subexps[FLOAT_TYPE2].rm_so != -1) - { - /* Found floating point type suffix. */ - end_index = subexps[FLOAT_TYPE2].rm_so; - type_index = FLOAT_TYPE2; - } - else - { - /* Any other floating point match. */ - end_index = subexps[0].rm_eo; - type_name = "f64"; - } - - /* We need a special case if the final character is ".". In this - case we might need to parse an integer. For example, "23.f()" is - a request for a trait method call, not a syntax error involving - the floating point number "23.". */ - gdb_assert (subexps[0].rm_eo > 0); - if (pstate->lexptr[subexps[0].rm_eo - 1] == '.') - { - const char *next = skip_spaces (&pstate->lexptr[subexps[0].rm_eo]); - - if (rust_identifier_start_p (*next) || *next == '.') - { - --subexps[0].rm_eo; - is_integer = 1; - end_index = subexps[0].rm_eo; - type_name = "i32"; - could_be_decimal = 1; - implicit_i32 = 1; - } - } - - /* Compute the type name if we haven't already. */ - std::string type_name_holder; - if (type_name == NULL) - { - gdb_assert (type_index != -1); - type_name_holder = std::string ((pstate->lexptr - + subexps[type_index].rm_so), - (subexps[type_index].rm_eo - - subexps[type_index].rm_so)); - type_name = type_name_holder.c_str (); - } - - /* Look up the type. */ - type = get_type (type_name); - - /* Copy the text of the number and remove the "_"s. */ - std::string number; - for (i = 0; i < end_index && pstate->lexptr[i]; ++i) - { - if (pstate->lexptr[i] == '_') - could_be_decimal = 0; - else - number.push_back (pstate->lexptr[i]); - } - - /* Advance past the match. */ - pstate->lexptr += subexps[0].rm_eo; - - /* Parse the number. */ - if (is_integer) - { - uint64_t value; - int radix = 10; - int offset = 0; - - if (number[0] == '0') - { - if (number[1] == 'x') - radix = 16; - else if (number[1] == 'o') - radix = 8; - else if (number[1] == 'b') - radix = 2; - if (radix != 10) - { - offset = 2; - could_be_decimal = 0; - } - } - - value = strtoulst (number.c_str () + offset, NULL, radix); - if (implicit_i32 && value >= ((uint64_t) 1) << 31) - type = get_type ("i64"); - - lvalp->typed_val_int.val = value; - lvalp->typed_val_int.type = type; - } - else - { - lvalp->typed_val_float.type = type; - bool parsed = parse_float (number.c_str (), number.length (), - lvalp->typed_val_float.type, - lvalp->typed_val_float.val); - gdb_assert (parsed); - } - - return is_integer ? (could_be_decimal ? DECIMAL_INTEGER : INTEGER) : FLOAT; -} - -/* The lexer. */ - -static int -rustyylex (YYSTYPE *lvalp, rust_parser *parser) -{ - struct parser_state *pstate = parser->pstate; - - /* Skip all leading whitespace. */ - while (pstate->lexptr[0] == ' ' - || pstate->lexptr[0] == '\t' - || pstate->lexptr[0] == '\r' - || pstate->lexptr[0] == '\n') - ++pstate->lexptr; - - /* If we hit EOF and we're completing, then return COMPLETE -- maybe - we're completing an empty string at the end of a field_expr. - But, we don't want to return two COMPLETE tokens in a row. */ - if (pstate->lexptr[0] == '\0' && pstate->lexptr == pstate->prev_lexptr) - return 0; - pstate->prev_lexptr = pstate->lexptr; - if (pstate->lexptr[0] == '\0') - { - if (pstate->parse_completion) - { - lvalp->sval = make_stoken (""); - return COMPLETE; - } - return 0; - } - - if (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9') - return parser->lex_number (lvalp); - else if (pstate->lexptr[0] == 'b' && pstate->lexptr[1] == '\'') - return parser->lex_character (lvalp); - else if (pstate->lexptr[0] == 'b' && pstate->lexptr[1] == '"') - return parser->lex_string (lvalp); - else if (pstate->lexptr[0] == 'b' && starts_raw_string (pstate->lexptr + 1)) - return parser->lex_string (lvalp); - else if (starts_raw_string (pstate->lexptr)) - return parser->lex_string (lvalp); - else if (rust_identifier_start_p (pstate->lexptr[0])) - return parser->lex_identifier (lvalp); - else if (pstate->lexptr[0] == '"') - return parser->lex_string (lvalp); - else if (pstate->lexptr[0] == '\'') - return parser->lex_character (lvalp); - else if (pstate->lexptr[0] == '}' || pstate->lexptr[0] == ']') - { - /* Falls through to lex_operator. */ - --parser->paren_depth; - } - else if (pstate->lexptr[0] == '(' || pstate->lexptr[0] == '{') - { - /* Falls through to lex_operator. */ - ++parser->paren_depth; - } - else if (pstate->lexptr[0] == ',' && pstate->comma_terminates - && parser->paren_depth == 0) - return 0; - - return parser->lex_operator (lvalp); -} - -/* Push back a single character to be re-lexed. */ - -void -rust_parser::push_back (char c) -{ - /* Can't be called before any lexing. */ - gdb_assert (pstate->prev_lexptr != NULL); - - --pstate->lexptr; - gdb_assert (*pstate->lexptr == c); -} - - - -/* Make an arbitrary operation and fill in the fields. */ - -const struct rust_op * -rust_parser::ast_operation (enum exp_opcode opcode, const struct rust_op *left, - const struct rust_op *right) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = opcode; - result->left.op = left; - result->right.op = right; - - return result; -} - -/* Make a compound assignment operation. */ - -const struct rust_op * -rust_parser::ast_compound_assignment (enum exp_opcode opcode, - const struct rust_op *left, - const struct rust_op *right) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = opcode; - result->compound_assignment = 1; - result->left.op = left; - result->right.op = right; - - return result; -} - -/* Make a typed integer literal operation. */ - -const struct rust_op * -rust_parser::ast_literal (struct typed_val_int val) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_LONG; - result->left.typed_val_int = val; - - return result; -} - -/* Make a typed floating point literal operation. */ - -const struct rust_op * -rust_parser::ast_dliteral (struct typed_val_float val) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_FLOAT; - result->left.typed_val_float = val; - - return result; -} - -/* Make a unary operation. */ - -const struct rust_op * -rust_parser::ast_unary (enum exp_opcode opcode, const struct rust_op *expr) -{ - return ast_operation (opcode, expr, NULL); -} - -/* Make a cast operation. */ - -const struct rust_op * -rust_parser::ast_cast (const struct rust_op *expr, const struct rust_op *type) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = UNOP_CAST; - result->left.op = expr; - result->right.op = type; - - return result; -} - -/* Make a call-like operation. This is nominally a function call, but - when lowering we may discover that it actually represents the - creation of a tuple struct. */ - -const struct rust_op * -rust_parser::ast_call_ish (enum exp_opcode opcode, const struct rust_op *expr, - rust_op_vector *params) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = opcode; - result->left.op = expr; - result->right.params = params; - - return result; -} - -/* Make a structure creation operation. */ - -const struct rust_op * -rust_parser::ast_struct (const struct rust_op *name, rust_set_vector *fields) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_AGGREGATE; - result->left.op = name; - result->right.field_inits = fields; - - return result; -} - -/* Make an identifier path. */ - -const struct rust_op * -rust_parser::ast_path (struct stoken path, rust_op_vector *params) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_VAR_VALUE; - result->left.sval = path; - result->right.params = params; - - return result; -} - -/* Make a string constant operation. */ - -const struct rust_op * -rust_parser::ast_string (struct stoken str) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_STRING; - result->left.sval = str; - - return result; -} - -/* Make a field expression. */ - -const struct rust_op * -rust_parser::ast_structop (const struct rust_op *left, const char *name, - int completing) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = STRUCTOP_STRUCT; - result->completing = completing; - result->left.op = left; - result->right.sval = make_stoken (name); - - return result; -} - -/* Make an anonymous struct operation, like 'x.0'. */ - -const struct rust_op * -rust_parser::ast_structop_anonymous (const struct rust_op *left, - struct typed_val_int number) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = STRUCTOP_ANONYMOUS; - result->left.op = left; - result->right.typed_val_int = number; - - return result; -} - -/* Make a range operation. */ - -const struct rust_op * -rust_parser::ast_range (const struct rust_op *lhs, const struct rust_op *rhs, - bool inclusive) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_RANGE; - result->inclusive = inclusive; - result->left.op = lhs; - result->right.op = rhs; - - return result; -} - -/* A helper function to make a type-related AST node. */ - -struct rust_op * -rust_parser::ast_basic_type (enum type_code typecode) -{ - struct rust_op *result = OBSTACK_ZALLOC (&obstack, struct rust_op); - - result->opcode = OP_TYPE; - result->typecode = typecode; - return result; -} - -/* Create an AST node describing an array type. */ - -const struct rust_op * -rust_parser::ast_array_type (const struct rust_op *lhs, - struct typed_val_int val) -{ - struct rust_op *result = ast_basic_type (TYPE_CODE_ARRAY); - - result->left.op = lhs; - result->right.typed_val_int = val; - return result; -} - -/* Create an AST node describing a reference type. */ - -const struct rust_op * -rust_parser::ast_slice_type (const struct rust_op *type) -{ - /* Use TYPE_CODE_COMPLEX just because it is handy. */ - struct rust_op *result = ast_basic_type (TYPE_CODE_COMPLEX); - - result->left.op = type; - return result; -} - -/* Create an AST node describing a reference type. */ - -const struct rust_op * -rust_parser::ast_reference_type (const struct rust_op *type) -{ - struct rust_op *result = ast_basic_type (TYPE_CODE_REF); - - result->left.op = type; - return result; -} - -/* Create an AST node describing a pointer type. */ - -const struct rust_op * -rust_parser::ast_pointer_type (const struct rust_op *type, int is_mut) -{ - struct rust_op *result = ast_basic_type (TYPE_CODE_PTR); - - result->left.op = type; - /* For the time being we ignore is_mut. */ - return result; -} - -/* Create an AST node describing a function type. */ - -const struct rust_op * -rust_parser::ast_function_type (const struct rust_op *rtype, - rust_op_vector *params) -{ - struct rust_op *result = ast_basic_type (TYPE_CODE_FUNC); - - result->left.op = rtype; - result->right.params = params; - return result; -} - -/* Create an AST node describing a tuple type. */ - -const struct rust_op * -rust_parser::ast_tuple_type (rust_op_vector *params) -{ - struct rust_op *result = ast_basic_type (TYPE_CODE_STRUCT); - - result->left.params = params; - return result; -} - -/* A helper to appropriately munge NAME and BLOCK depending on the - presence of a leading "::". */ - -static void -munge_name_and_block (const char **name, const struct block **block) -{ - /* If it is a global reference, skip the current block in favor of - the static block. */ - if (startswith (*name, "::")) - { - *name += 2; - *block = block_static_block (*block); - } -} - -/* Like lookup_symbol, but handles Rust namespace conventions, and - doesn't require field_of_this_result. */ - -struct block_symbol -rust_parser::lookup_symbol (const char *name, const struct block *block, - const domain_enum domain) -{ - struct block_symbol result; - - munge_name_and_block (&name, &block); - - result = ::lookup_symbol (name, block, domain, NULL); - if (result.symbol != NULL) - update_innermost_block (result); - return result; -} - -/* Look up a type, following Rust namespace conventions. */ - -struct type * -rust_parser::rust_lookup_type (const char *name, const struct block *block) -{ - struct block_symbol result; - struct type *type; - - munge_name_and_block (&name, &block); - - result = ::lookup_symbol (name, block, STRUCT_DOMAIN, NULL); - if (result.symbol != NULL) - { - update_innermost_block (result); - return SYMBOL_TYPE (result.symbol); - } - - type = lookup_typename (language (), name, NULL, 1); - if (type != NULL) - return type; - - /* Last chance, try a built-in type. */ - return language_lookup_primitive_type (language (), arch (), name); -} - -/* Convert a vector of rust_ops representing types to a vector of - types. */ - -std::vector -rust_parser::convert_params_to_types (rust_op_vector *params) -{ - std::vector result; - - if (params != nullptr) - { - for (const rust_op *op : *params) - result.push_back (convert_ast_to_type (op)); - } - - return result; -} - -/* Convert a rust_op representing a type to a struct type *. */ - -struct type * -rust_parser::convert_ast_to_type (const struct rust_op *operation) -{ - struct type *type, *result = NULL; - - if (operation->opcode == OP_VAR_VALUE) - { - const char *varname = convert_name (operation); - - result = rust_lookup_type (varname, pstate->expression_context_block); - if (result == NULL) - error (_("No typed name '%s' in current context"), varname); - return result; - } - - gdb_assert (operation->opcode == OP_TYPE); - - switch (operation->typecode) - { - case TYPE_CODE_ARRAY: - type = convert_ast_to_type (operation->left.op); - if (operation->right.typed_val_int.val < 0) - error (_("Negative array length")); - result = lookup_array_range_type (type, 0, - operation->right.typed_val_int.val - 1); - break; - - case TYPE_CODE_COMPLEX: - { - struct type *usize = get_type ("usize"); - - type = convert_ast_to_type (operation->left.op); - result = rust_slice_type ("&[*gdb*]", type, usize); - } - break; - - case TYPE_CODE_REF: - case TYPE_CODE_PTR: - /* For now we treat &x and *x identically. */ - type = convert_ast_to_type (operation->left.op); - result = lookup_pointer_type (type); - break; - - case TYPE_CODE_FUNC: - { - std::vector args - (convert_params_to_types (operation->right.params)); - struct type **argtypes = NULL; - - type = convert_ast_to_type (operation->left.op); - if (!args.empty ()) - argtypes = args.data (); - - result - = lookup_function_type_with_arguments (type, args.size (), - argtypes); - result = lookup_pointer_type (result); - } - break; - - case TYPE_CODE_STRUCT: - { - std::vector args - (convert_params_to_types (operation->left.params)); - int i; - const char *name; - - obstack_1grow (&obstack, '('); - for (i = 0; i < args.size (); ++i) - { - std::string type_name = type_to_string (args[i]); - - if (i > 0) - obstack_1grow (&obstack, ','); - obstack_grow_str (&obstack, type_name.c_str ()); - } - - obstack_grow_str0 (&obstack, ")"); - name = (const char *) obstack_finish (&obstack); - - /* We don't allow creating new tuple types (yet), but we do - allow looking up existing tuple types. */ - result = rust_lookup_type (name, pstate->expression_context_block); - if (result == NULL) - error (_("could not find tuple type '%s'"), name); - } - break; - - default: - gdb_assert_not_reached ("unhandled opcode in convert_ast_to_type"); - } - - gdb_assert (result != NULL); - return result; -} - -/* A helper function to turn a rust_op representing a name into a full - name. This applies generic arguments as needed. The returned name - is allocated on the work obstack. */ - -const char * -rust_parser::convert_name (const struct rust_op *operation) -{ - int i; - - gdb_assert (operation->opcode == OP_VAR_VALUE); - - if (operation->right.params == NULL) - return operation->left.sval.ptr; - - std::vector types - (convert_params_to_types (operation->right.params)); - - obstack_grow_str (&obstack, operation->left.sval.ptr); - obstack_1grow (&obstack, '<'); - for (i = 0; i < types.size (); ++i) - { - std::string type_name = type_to_string (types[i]); - - if (i > 0) - obstack_1grow (&obstack, ','); - - obstack_grow_str (&obstack, type_name.c_str ()); - } - obstack_grow_str0 (&obstack, ">"); - - return (const char *) obstack_finish (&obstack); -} - -/* A helper function that converts a vec of rust_ops to a gdb - expression. */ - -std::vector -rust_parser::convert_params_to_expression (rust_op_vector *params, - const struct rust_op *top) -{ - std::vector result; - for (const rust_op *elem : *params) - result.push_back (convert_ast_to_expression (elem, top)); - result.shrink_to_fit (); - return result; -} - -typedef expr::operation_up binop_maker_ftype (expr::operation_up &&, - expr::operation_up &&); - -/* Map from an expression opcode to a function that will create an - instance of the appropriate operation subclass. Only binary - operations are handled this way. */ -static std::unordered_map> maker_map; - -/* Lower a rust_op to a gdb expression. STATE is the parser state. - OPERATION is the operation to lower. TOP is a pointer to the - top-most operation; it is used to handle the special case where the - top-most expression is an identifier and can be optionally lowered - to OP_TYPE. WANT_TYPE is a flag indicating that, if the expression - is the name of a type, then emit an OP_TYPE for it (rather than - erroring). If WANT_TYPE is set, then the similar TOP handling is - not done. */ - -expr::operation_up -rust_parser::convert_ast_to_expression (const struct rust_op *operation, - const struct rust_op *top, - bool want_type) -{ - using namespace expr; - - switch (operation->opcode) - { - case OP_LONG: - return operation_up - (new long_const_operation (operation->left.typed_val_int.type, - operation->left.typed_val_int.val)); - - case OP_FLOAT: - { - float_data data; - memcpy (data.data (), operation->left.typed_val_float.val, - sizeof (operation->left.typed_val_float.val)); - return operation_up - (new float_const_operation (operation->left.typed_val_float.type, - data)); - } - - case STRUCTOP_STRUCT: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - auto result = new rust_structop (std::move (lhs), - operation->right.sval.ptr); - if (operation->completing) - pstate->mark_struct_expression (result); - return operation_up (result); - } - - case STRUCTOP_ANONYMOUS: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - - return operation_up - (new rust_struct_anon (operation->right.typed_val_int.val, - std::move (lhs))); - } - - case UNOP_SIZEOF: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top, - true); - return operation_up - (new unop_sizeof_operation (std::move (lhs))); - } - - case UNOP_PLUS: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - return operation_up - (new unary_plus_operation (std::move (lhs))); - } - case UNOP_NEG: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - return operation_up - (new unary_neg_operation (std::move (lhs))); - } - case UNOP_COMPLEMENT: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - return operation_up - (new rust_unop_compl_operation (std::move (lhs))); - } - case UNOP_IND: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - return operation_up - (new rust_unop_ind_operation (std::move (lhs))); - } - case UNOP_ADDR: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - return operation_up - (new rust_unop_addr_operation (std::move (lhs))); - } - - case BINOP_SUBSCRIPT: - case BINOP_MUL: - case BINOP_REPEAT: - case BINOP_DIV: - case BINOP_REM: - case BINOP_LESS: - case BINOP_GTR: - case BINOP_BITWISE_AND: - case BINOP_BITWISE_IOR: - case BINOP_BITWISE_XOR: - case BINOP_ADD: - case BINOP_SUB: - case BINOP_LOGICAL_OR: - case BINOP_LOGICAL_AND: - case BINOP_EQUAL: - case BINOP_NOTEQUAL: - case BINOP_LEQ: - case BINOP_GEQ: - case BINOP_LSH: - case BINOP_RSH: - case BINOP_ASSIGN: - case OP_RUST_ARRAY: - { - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - operation_up rhs = convert_ast_to_expression (operation->right.op, - top); - operation_up result; - if (operation->compound_assignment) - result = (operation_up - (new assign_modify_operation (operation->opcode, - std::move (lhs), - std::move (rhs)))); - else - { - auto iter = maker_map.find (operation->opcode); - gdb_assert (iter != maker_map.end ()); - result = iter->second (std::move (lhs), std::move (rhs)); - } - - if (operation->compound_assignment - || operation->opcode == BINOP_ASSIGN) - { - struct type *type - = language_lookup_primitive_type (pstate->language (), - pstate->gdbarch (), - "()"); - - operation_up nil (new long_const_operation (type, 0)); - result.reset (new comma_operation (std::move (result), - std::move (nil))); - } - - return result; - } - - case UNOP_CAST: - { - struct type *type = convert_ast_to_type (operation->right.op); - operation_up lhs = convert_ast_to_expression (operation->left.op, top); - return operation_up (new unop_cast_operation (std::move (lhs), type)); - } - - case OP_FUNCALL: - { - if (operation->left.op->opcode == OP_VAR_VALUE) - { - struct type *type; - const char *varname = convert_name (operation->left.op); - - type = rust_lookup_type (varname, - pstate->expression_context_block); - if (type != NULL) - { - /* This is actually a tuple struct expression, not a - call expression. */ - rust_op_vector *params = operation->right.params; - - if (type->code () != TYPE_CODE_NAMESPACE) - { - if (!rust_tuple_struct_type_p (type)) - error (_("Type %s is not a tuple struct"), varname); - - std::vector> args - (params->size ()); - for (int i = 0; i < params->size (); ++i) - { - char *cell = get_print_cell (); - - operation_up op - = convert_ast_to_expression ((*params)[i], top); - xsnprintf (cell, PRINT_CELL_SIZE, "__%d", i); - args[i] = { cell, std::move (op) }; - } - - return make_operation - (type, operation_up (), std::move (args)); - } - } - } - operation_up callee = convert_ast_to_expression (operation->left.op, - top); - std::vector args - = convert_params_to_expression (operation->right.params, top); - return make_operation (std::move (callee), - std::move (args)); - } - - case OP_ARRAY: - { - gdb_assert (operation->left.op == NULL); - std::vector subexps - = convert_params_to_expression (operation->right.params, top); - return make_operation - (0, operation->right.params->size () - 1, std::move (subexps)); - } - - case OP_VAR_VALUE: - { - struct block_symbol sym; - const char *varname; - - if (operation->left.sval.ptr[0] == '$') - { - pstate->push_dollar (operation->left.sval); - return pstate->pop (); - } - - varname = convert_name (operation); - sym = lookup_symbol (varname, pstate->expression_context_block, - VAR_DOMAIN); - operation_up result; - if (sym.symbol != NULL && SYMBOL_CLASS (sym.symbol) != LOC_TYPEDEF) - result.reset (new var_value_operation (sym)); - else - { - struct type *type = NULL; - - if (sym.symbol != NULL) - { - gdb_assert (SYMBOL_CLASS (sym.symbol) == LOC_TYPEDEF); - type = SYMBOL_TYPE (sym.symbol); - } - if (type == NULL) - type = rust_lookup_type (varname, - pstate->expression_context_block); - if (type == NULL) - error (_("No symbol '%s' in current context"), varname); - - if (!want_type - && type->code () == TYPE_CODE_STRUCT - && type->num_fields () == 0) - { - /* A unit-like struct. */ - result.reset (new rust_aggregate_operation (type, {}, {})); - } - else if (want_type || operation == top) - result.reset (new type_operation (type)); - else - error (_("Found type '%s', which can't be " - "evaluated in this context"), - varname); - } - - return result; - } - - case OP_AGGREGATE: - { - rust_set_vector *fields = operation->right.field_inits; - struct type *type; - const char *name; - - operation_up others; - std::vector> field_v; - for (const set_field &init : *fields) - { - operation_up expr = convert_ast_to_expression (init.init, top); - - if (init.name.ptr == NULL) - others = std::move (expr); - else - field_v.emplace_back (init.name.ptr, std::move (expr)); - } - - name = convert_name (operation->left.op); - type = rust_lookup_type (name, pstate->expression_context_block); - if (type == NULL) - error (_("Could not find type '%s'"), operation->left.sval.ptr); - - if (type->code () != TYPE_CODE_STRUCT - || rust_tuple_type_p (type) - || rust_tuple_struct_type_p (type)) - error (_("Struct expression applied to non-struct type")); - - return operation_up - (new rust_aggregate_operation (type, std::move (others), - std::move (field_v))); - } - - case OP_STRING: - return (operation_up - (new string_operation (::copy_name (operation->left.sval)))); - - case OP_RANGE: - { - enum range_flag kind = (RANGE_HIGH_BOUND_DEFAULT - | RANGE_LOW_BOUND_DEFAULT); - operation_up lhs, rhs; - - if (operation->left.op != NULL) - { - lhs = convert_ast_to_expression (operation->left.op, top); - kind &= ~RANGE_LOW_BOUND_DEFAULT; - } - if (operation->right.op != NULL) - { - rhs = convert_ast_to_expression (operation->right.op, top); - if (kind == (RANGE_HIGH_BOUND_DEFAULT | RANGE_LOW_BOUND_DEFAULT)) - { - kind = RANGE_LOW_BOUND_DEFAULT; - if (!operation->inclusive) - kind |= RANGE_HIGH_BOUND_EXCLUSIVE; - } - else - { - gdb_assert (kind == RANGE_HIGH_BOUND_DEFAULT); - kind = RANGE_STANDARD; - if (!operation->inclusive) - kind |= RANGE_HIGH_BOUND_EXCLUSIVE; - } - } - else - { - /* Nothing should make an inclusive range without an upper - bound. */ - gdb_assert (!operation->inclusive); - } - - return operation_up (new rust_range_operation (kind, - std::move (lhs), - std::move (rhs))); - } - - default: - gdb_assert_not_reached ("unhandled opcode in convert_ast_to_expression"); - } -} - - - -/* The parser as exposed to gdb. */ - -int -rust_language::parser (struct parser_state *state) const -{ - int result; - - /* This sets various globals and also clears them on - destruction. */ - rust_parser parser (state); - - result = rustyyparse (&parser); - - if (!result || (state->parse_completion && parser.rust_ast != NULL)) - { - expr::operation_up op - = parser.convert_ast_to_expression (parser.rust_ast, parser.rust_ast); - state->set_operation (std::move (op)); - } - - return result; -} - -/* The parser error handler. */ - -static void -rustyyerror (rust_parser *parser, const char *msg) -{ - const char *where = (parser->pstate->prev_lexptr - ? parser->pstate->prev_lexptr - : parser->pstate->lexptr); - error (_("%s in expression, near `%s'."), msg, where); -} - - - -#if GDB_SELF_TEST - -/* Initialize the lexer for testing. */ - -static void -rust_lex_test_init (rust_parser *parser, const char *input) -{ - parser->pstate->prev_lexptr = NULL; - parser->pstate->lexptr = input; - parser->paren_depth = 0; -} - -/* A test helper that lexes a string, expecting a single token. It - returns the lexer data for this token. */ - -static RUSTSTYPE -rust_lex_test_one (rust_parser *parser, const char *input, int expected) -{ - int token; - RUSTSTYPE result; - - rust_lex_test_init (parser, input); - - token = rustyylex (&result, parser); - SELF_CHECK (token == expected); - - if (token) - { - RUSTSTYPE ignore; - token = rustyylex (&ignore, parser); - SELF_CHECK (token == 0); - } - - return result; -} - -/* Test that INPUT lexes as the integer VALUE. */ - -static void -rust_lex_int_test (rust_parser *parser, const char *input, - LONGEST value, int kind) -{ - RUSTSTYPE result = rust_lex_test_one (parser, input, kind); - SELF_CHECK (result.typed_val_int.val == value); -} - -/* Test that INPUT throws an exception with text ERR. */ - -static void -rust_lex_exception_test (rust_parser *parser, const char *input, - const char *err) -{ - try - { - /* The "kind" doesn't matter. */ - rust_lex_test_one (parser, input, DECIMAL_INTEGER); - SELF_CHECK (0); - } - catch (const gdb_exception_error &except) - { - SELF_CHECK (strcmp (except.what (), err) == 0); - } -} - -/* Test that INPUT lexes as the identifier, string, or byte-string - VALUE. KIND holds the expected token kind. */ - -static void -rust_lex_stringish_test (rust_parser *parser, const char *input, - const char *value, int kind) -{ - RUSTSTYPE result = rust_lex_test_one (parser, input, kind); - SELF_CHECK (result.sval.length == strlen (value)); - SELF_CHECK (strncmp (result.sval.ptr, value, result.sval.length) == 0); -} - -/* Helper to test that a string parses as a given token sequence. */ - -static void -rust_lex_test_sequence (rust_parser *parser, const char *input, int len, - const int expected[]) -{ - int i; - - parser->pstate->lexptr = input; - parser->paren_depth = 0; - - for (i = 0; i < len; ++i) - { - RUSTSTYPE ignore; - int token = rustyylex (&ignore, parser); - - SELF_CHECK (token == expected[i]); - } -} - -/* Tests for an integer-parsing corner case. */ - -static void -rust_lex_test_trailing_dot (rust_parser *parser) -{ - const int expected1[] = { DECIMAL_INTEGER, '.', IDENT, '(', ')', 0 }; - const int expected2[] = { INTEGER, '.', IDENT, '(', ')', 0 }; - const int expected3[] = { FLOAT, EQEQ, '(', ')', 0 }; - const int expected4[] = { DECIMAL_INTEGER, DOTDOT, DECIMAL_INTEGER, 0 }; - - rust_lex_test_sequence (parser, "23.g()", ARRAY_SIZE (expected1), expected1); - rust_lex_test_sequence (parser, "23_0.g()", ARRAY_SIZE (expected2), - expected2); - rust_lex_test_sequence (parser, "23.==()", ARRAY_SIZE (expected3), - expected3); - rust_lex_test_sequence (parser, "23..25", ARRAY_SIZE (expected4), expected4); -} - -/* Tests of completion. */ - -static void -rust_lex_test_completion (rust_parser *parser) -{ - const int expected[] = { IDENT, '.', COMPLETE, 0 }; - - parser->pstate->parse_completion = 1; - - rust_lex_test_sequence (parser, "something.wha", ARRAY_SIZE (expected), - expected); - rust_lex_test_sequence (parser, "something.", ARRAY_SIZE (expected), - expected); - - parser->pstate->parse_completion = 0; -} - -/* Test pushback. */ - -static void -rust_lex_test_push_back (rust_parser *parser) -{ - int token; - RUSTSTYPE lval; - - rust_lex_test_init (parser, ">>="); - - token = rustyylex (&lval, parser); - SELF_CHECK (token == COMPOUND_ASSIGN); - SELF_CHECK (lval.opcode == BINOP_RSH); - - parser->push_back ('='); - - token = rustyylex (&lval, parser); - SELF_CHECK (token == '='); - - token = rustyylex (&lval, parser); - SELF_CHECK (token == 0); -} - -/* Unit test the lexer. */ - -static void -rust_lex_tests (void) -{ - int i; - - /* Set up dummy "parser", so that rust_type works. */ - struct parser_state ps (language_def (language_rust), target_gdbarch (), - nullptr, 0, 0, nullptr, 0, nullptr, false); - rust_parser parser (&ps); - - rust_lex_test_one (&parser, "", 0); - rust_lex_test_one (&parser, " \t \n \r ", 0); - rust_lex_test_one (&parser, "thread 23", 0); - rust_lex_test_one (&parser, "task 23", 0); - rust_lex_test_one (&parser, "th 104", 0); - rust_lex_test_one (&parser, "ta 97", 0); - - rust_lex_int_test (&parser, "'z'", 'z', INTEGER); - rust_lex_int_test (&parser, "'\\xff'", 0xff, INTEGER); - rust_lex_int_test (&parser, "'\\u{1016f}'", 0x1016f, INTEGER); - rust_lex_int_test (&parser, "b'z'", 'z', INTEGER); - rust_lex_int_test (&parser, "b'\\xfe'", 0xfe, INTEGER); - rust_lex_int_test (&parser, "b'\\xFE'", 0xfe, INTEGER); - rust_lex_int_test (&parser, "b'\\xfE'", 0xfe, INTEGER); - - /* Test all escapes in both modes. */ - rust_lex_int_test (&parser, "'\\n'", '\n', INTEGER); - rust_lex_int_test (&parser, "'\\r'", '\r', INTEGER); - rust_lex_int_test (&parser, "'\\t'", '\t', INTEGER); - rust_lex_int_test (&parser, "'\\\\'", '\\', INTEGER); - rust_lex_int_test (&parser, "'\\0'", '\0', INTEGER); - rust_lex_int_test (&parser, "'\\''", '\'', INTEGER); - rust_lex_int_test (&parser, "'\\\"'", '"', INTEGER); - - rust_lex_int_test (&parser, "b'\\n'", '\n', INTEGER); - rust_lex_int_test (&parser, "b'\\r'", '\r', INTEGER); - rust_lex_int_test (&parser, "b'\\t'", '\t', INTEGER); - rust_lex_int_test (&parser, "b'\\\\'", '\\', INTEGER); - rust_lex_int_test (&parser, "b'\\0'", '\0', INTEGER); - rust_lex_int_test (&parser, "b'\\''", '\'', INTEGER); - rust_lex_int_test (&parser, "b'\\\"'", '"', INTEGER); - - rust_lex_exception_test (&parser, "'z", "Unterminated character literal"); - rust_lex_exception_test (&parser, "b'\\x0'", "Not enough hex digits seen"); - rust_lex_exception_test (&parser, "b'\\u{0}'", - "Unicode escape in byte literal"); - rust_lex_exception_test (&parser, "'\\x0'", "Not enough hex digits seen"); - rust_lex_exception_test (&parser, "'\\u0'", "Missing '{' in Unicode escape"); - rust_lex_exception_test (&parser, "'\\u{0", "Missing '}' in Unicode escape"); - rust_lex_exception_test (&parser, "'\\u{0000007}", "Overlong hex escape"); - rust_lex_exception_test (&parser, "'\\u{}", "Not enough hex digits seen"); - rust_lex_exception_test (&parser, "'\\Q'", "Invalid escape \\Q in literal"); - rust_lex_exception_test (&parser, "b'\\Q'", "Invalid escape \\Q in literal"); - - rust_lex_int_test (&parser, "23", 23, DECIMAL_INTEGER); - rust_lex_int_test (&parser, "2_344__29", 234429, INTEGER); - rust_lex_int_test (&parser, "0x1f", 0x1f, INTEGER); - rust_lex_int_test (&parser, "23usize", 23, INTEGER); - rust_lex_int_test (&parser, "23i32", 23, INTEGER); - rust_lex_int_test (&parser, "0x1_f", 0x1f, INTEGER); - rust_lex_int_test (&parser, "0b1_101011__", 0x6b, INTEGER); - rust_lex_int_test (&parser, "0o001177i64", 639, INTEGER); - rust_lex_int_test (&parser, "0x123456789u64", 0x123456789ull, INTEGER); - - rust_lex_test_trailing_dot (&parser); - - rust_lex_test_one (&parser, "23.", FLOAT); - rust_lex_test_one (&parser, "23.99f32", FLOAT); - rust_lex_test_one (&parser, "23e7", FLOAT); - rust_lex_test_one (&parser, "23E-7", FLOAT); - rust_lex_test_one (&parser, "23e+7", FLOAT); - rust_lex_test_one (&parser, "23.99e+7f64", FLOAT); - rust_lex_test_one (&parser, "23.82f32", FLOAT); - - rust_lex_stringish_test (&parser, "hibob", "hibob", IDENT); - rust_lex_stringish_test (&parser, "hibob__93", "hibob__93", IDENT); - rust_lex_stringish_test (&parser, "thread", "thread", IDENT); - - rust_lex_stringish_test (&parser, "\"string\"", "string", STRING); - rust_lex_stringish_test (&parser, "\"str\\ting\"", "str\ting", STRING); - rust_lex_stringish_test (&parser, "\"str\\\"ing\"", "str\"ing", STRING); - rust_lex_stringish_test (&parser, "r\"str\\ing\"", "str\\ing", STRING); - rust_lex_stringish_test (&parser, "r#\"str\\ting\"#", "str\\ting", STRING); - rust_lex_stringish_test (&parser, "r###\"str\\\"ing\"###", "str\\\"ing", - STRING); - - rust_lex_stringish_test (&parser, "b\"string\"", "string", BYTESTRING); - rust_lex_stringish_test (&parser, "b\"\x73tring\"", "string", BYTESTRING); - rust_lex_stringish_test (&parser, "b\"str\\\"ing\"", "str\"ing", BYTESTRING); - rust_lex_stringish_test (&parser, "br####\"\\x73tring\"####", "\\x73tring", - BYTESTRING); - - for (i = 0; i < ARRAY_SIZE (identifier_tokens); ++i) - rust_lex_test_one (&parser, identifier_tokens[i].name, - identifier_tokens[i].value); - - for (i = 0; i < ARRAY_SIZE (operator_tokens); ++i) - rust_lex_test_one (&parser, operator_tokens[i].name, - operator_tokens[i].value); - - rust_lex_test_completion (&parser); - rust_lex_test_push_back (&parser); -} - -#endif /* GDB_SELF_TEST */ - -void _initialize_rust_exp (); -void -_initialize_rust_exp () -{ - int code = regcomp (&number_regex, number_regex_text, REG_EXTENDED); - /* If the regular expression was incorrect, it was a programming - error. */ - gdb_assert (code == 0); - - using namespace expr; - maker_map[BINOP_SUBSCRIPT] = make_operation; - maker_map[BINOP_MUL] = make_operation; - maker_map[BINOP_REPEAT] = make_operation; - maker_map[BINOP_DIV] = make_operation; - maker_map[BINOP_REM] = make_operation; - maker_map[BINOP_LESS] = make_operation; - maker_map[BINOP_GTR] = make_operation; - maker_map[BINOP_BITWISE_AND] = make_operation; - maker_map[BINOP_BITWISE_IOR] = make_operation; - maker_map[BINOP_BITWISE_XOR] = make_operation; - maker_map[BINOP_ADD] = make_operation; - maker_map[BINOP_SUB] = make_operation; - maker_map[BINOP_LOGICAL_OR] = make_operation; - maker_map[BINOP_LOGICAL_AND] = make_operation; - maker_map[BINOP_EQUAL] = make_operation; - maker_map[BINOP_NOTEQUAL] = make_operation; - maker_map[BINOP_LEQ] = make_operation; - maker_map[BINOP_GEQ] = make_operation; - maker_map[BINOP_LSH] = make_operation; - maker_map[BINOP_RSH] = make_operation; - maker_map[BINOP_ASSIGN] = make_operation; - maker_map[OP_RUST_ARRAY] = make_operation; - -#if GDB_SELF_TEST - selftests::register_test ("rust-lex", rust_lex_tests); -#endif -} diff --git a/gdb/rust-parse.c b/gdb/rust-parse.c new file mode 100644 index 00000000000..bb31782811c --- /dev/null +++ b/gdb/rust-parse.c @@ -0,0 +1,2351 @@ +/* Rust expression parsing for GDB, the GNU debugger. + + Copyright (C) 2016-2021 Free Software Foundation, Inc. + + This file is part of GDB. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . */ + +#include "defs.h" + +#include "block.h" +#include "charset.h" +#include "cp-support.h" +#include "gdb_obstack.h" +#include "gdb_regex.h" +#include "rust-lang.h" +#include "parser-defs.h" +#include "gdbsupport/selftest.h" +#include "value.h" +#include "gdbarch.h" +#include "rust-exp.h" + +using namespace expr; + +/* A regular expression for matching Rust numbers. This is split up + since it is very long and this gives us a way to comment the + sections. */ + +static const char number_regex_text[] = + /* subexpression 1: allows use of alternation, otherwise uninteresting */ + "^(" + /* First comes floating point. */ + /* Recognize number after the decimal point, with optional + exponent and optional type suffix. + subexpression 2: allows "?", otherwise uninteresting + subexpression 3: if present, type suffix + */ + "[0-9][0-9_]*\\.[0-9][0-9_]*([eE][-+]?[0-9][0-9_]*)?(f32|f64)?" +#define FLOAT_TYPE1 3 + "|" + /* Recognize exponent without decimal point, with optional type + suffix. + subexpression 4: if present, type suffix + */ +#define FLOAT_TYPE2 4 + "[0-9][0-9_]*[eE][-+]?[0-9][0-9_]*(f32|f64)?" + "|" + /* "23." is a valid floating point number, but "23.e5" and + "23.f32" are not. So, handle the trailing-. case + separately. */ + "[0-9][0-9_]*\\." + "|" + /* Finally come integers. + subexpression 5: text of integer + subexpression 6: if present, type suffix + subexpression 7: allows use of alternation, otherwise uninteresting + */ +#define INT_TEXT 5 +#define INT_TYPE 6 + "(0x[a-fA-F0-9_]+|0o[0-7_]+|0b[01_]+|[0-9][0-9_]*)" + "([iu](size|8|16|32|64))?" + ")"; +/* The number of subexpressions to allocate space for, including the + "0th" whole match subexpression. */ +#define NUM_SUBEXPRESSIONS 8 + +/* The compiled number-matching regex. */ + +static regex_t number_regex; + +/* The kinds of tokens. Note that single-character tokens are + represented by themselves, so for instance '[' is a token. */ +enum token_type : int +{ + /* Make sure to start after any ASCII character. */ + GDBVAR = 256, + IDENT, + COMPLETE, + INTEGER, + DECIMAL_INTEGER, + STRING, + BYTESTRING, + FLOAT, + COMPOUND_ASSIGN, + + /* Keyword tokens. */ + KW_AS, + KW_IF, + KW_TRUE, + KW_FALSE, + KW_SUPER, + KW_SELF, + KW_MUT, + KW_EXTERN, + KW_CONST, + KW_FN, + KW_SIZEOF, + + /* Operator tokens. */ + DOTDOT, + DOTDOTEQ, + OROR, + ANDAND, + EQEQ, + NOTEQ, + LTEQ, + GTEQ, + LSH, + RSH, + COLONCOLON, + ARROW, +}; + +/* A typed integer constant. */ + +struct typed_val_int +{ + LONGEST val; + struct type *type; +}; + +/* A typed floating point constant. */ + +struct typed_val_float +{ + float_data val; + struct type *type; +}; + +/* A struct of this type is used to describe a token. */ + +struct token_info +{ + const char *name; + int value; + enum exp_opcode opcode; +}; + +/* Identifier tokens. */ + +static const struct token_info identifier_tokens[] = +{ + { "as", KW_AS, OP_NULL }, + { "false", KW_FALSE, OP_NULL }, + { "if", 0, OP_NULL }, + { "mut", KW_MUT, OP_NULL }, + { "const", KW_CONST, OP_NULL }, + { "self", KW_SELF, OP_NULL }, + { "super", KW_SUPER, OP_NULL }, + { "true", KW_TRUE, OP_NULL }, + { "extern", KW_EXTERN, OP_NULL }, + { "fn", KW_FN, OP_NULL }, + { "sizeof", KW_SIZEOF, OP_NULL }, +}; + +/* Operator tokens, sorted longest first. */ + +static const struct token_info operator_tokens[] = +{ + { ">>=", COMPOUND_ASSIGN, BINOP_RSH }, + { "<<=", COMPOUND_ASSIGN, BINOP_LSH }, + + { "<<", LSH, OP_NULL }, + { ">>", RSH, OP_NULL }, + { "&&", ANDAND, OP_NULL }, + { "||", OROR, OP_NULL }, + { "==", EQEQ, OP_NULL }, + { "!=", NOTEQ, OP_NULL }, + { "<=", LTEQ, OP_NULL }, + { ">=", GTEQ, OP_NULL }, + { "+=", COMPOUND_ASSIGN, BINOP_ADD }, + { "-=", COMPOUND_ASSIGN, BINOP_SUB }, + { "*=", COMPOUND_ASSIGN, BINOP_MUL }, + { "/=", COMPOUND_ASSIGN, BINOP_DIV }, + { "%=", COMPOUND_ASSIGN, BINOP_REM }, + { "&=", COMPOUND_ASSIGN, BINOP_BITWISE_AND }, + { "|=", COMPOUND_ASSIGN, BINOP_BITWISE_IOR }, + { "^=", COMPOUND_ASSIGN, BINOP_BITWISE_XOR }, + { "..=", DOTDOTEQ, OP_NULL }, + + { "::", COLONCOLON, OP_NULL }, + { "..", DOTDOT, OP_NULL }, + { "->", ARROW, OP_NULL } +}; + +/* An instance of this is created before parsing, and destroyed when + parsing is finished. */ + +struct rust_parser +{ + explicit rust_parser (struct parser_state *state) + : pstate (state) + { + } + + DISABLE_COPY_AND_ASSIGN (rust_parser); + + /* Return the parser's language. */ + const struct language_defn *language () const + { + return pstate->language (); + } + + /* Return the parser's gdbarch. */ + struct gdbarch *arch () const + { + return pstate->gdbarch (); + } + + /* A helper to look up a Rust type, or fail. This only works for + types defined by rust_language_arch_info. */ + + struct type *get_type (const char *name) + { + struct type *type; + + type = language_lookup_primitive_type (language (), arch (), name); + if (type == NULL) + error (_("Could not find Rust type %s"), name); + return type; + } + + std::string crate_name (const std::string &name); + std::string super_name (const std::string &ident, unsigned int n_supers); + + int lex_character (); + int lex_number (); + int lex_string (); + int lex_identifier (); + uint32_t lex_hex (int min, int max); + uint32_t lex_escape (int is_byte); + int lex_operator (); + int lex_one_token (); + void push_back (char c); + + /* The main interface to lexing. Lexes one token and updates the + internal state. */ + void lex () + { + current_token = lex_one_token (); + } + + /* Assuming the current token is TYPE, lex the next token. */ + void assume (int type) + { + gdb_assert (current_token == type); + lex (); + } + + /* Require the single-character token C, and lex the next token; or + throw an exception. */ + void require (char type) + { + if (current_token != type) + error (_("'%c' expected"), type); + lex (); + } + + /* Entry point for all parsing. */ + operation_up parse_entry_point () + { + lex (); + return parse_expr (); + } + + operation_up parse_tuple (); + operation_up parse_array (); + operation_up name_to_operation (const std::string &name); + operation_up parse_struct_expr (struct type *type); + operation_up parse_binop (bool required); + operation_up parse_range (); + operation_up parse_expr (); + operation_up parse_sizeof (); + operation_up parse_addr (); + operation_up parse_field (operation_up &&); + operation_up parse_index (operation_up &&); + std::vector parse_paren_args (); + operation_up parse_call (operation_up &&); + std::vector parse_type_list (); + std::vector parse_maybe_type_list (); + struct type *parse_array_type (); + struct type *parse_slice_type (); + struct type *parse_pointer_type (); + struct type *parse_function_type (); + struct type *parse_tuple_type (); + struct type *parse_type (); + std::string parse_path (bool for_expr); + operation_up parse_string (); + operation_up parse_tuple_struct (struct type *type); + operation_up parse_path_expr (); + operation_up parse_atom (bool required); + + void update_innermost_block (struct block_symbol sym); + struct block_symbol lookup_symbol (const char *name, + const struct block *block, + const domain_enum domain); + struct type *rust_lookup_type (const char *name); + + /* Clear some state. This is only used for testing. */ +#if GDB_SELF_TEST + void reset (const char *input) + { + pstate->prev_lexptr = nullptr; + pstate->lexptr = input; + paren_depth = 0; + current_token = 0; + current_int_val = {}; + current_float_val = {}; + current_string_val = {}; + current_opcode = OP_NULL; + } +#endif /* GDB_SELF_TEST */ + + /* Return the token's string value as a string. */ + std::string get_string () const + { + return std::string (current_string_val.ptr, current_string_val.length); + } + + /* A pointer to this is installed globally. */ + auto_obstack obstack; + + /* The parser state gdb gave us. */ + struct parser_state *pstate; + + /* Depth of parentheses. */ + int paren_depth = 0; + + /* The current token's type. */ + int current_token = 0; + /* The current token's payload, if any. */ + typed_val_int current_int_val {}; + typed_val_float current_float_val {}; + struct stoken current_string_val {}; + enum exp_opcode current_opcode = OP_NULL; + + /* When completing, this may be set to the field operation to + complete. */ + operation_up completion_op; +}; + +/* Return an string referring to NAME, but relative to the crate's + name. */ + +std::string +rust_parser::crate_name (const std::string &name) +{ + std::string crate = rust_crate_for_block (pstate->expression_context_block); + + if (crate.empty ()) + error (_("Could not find crate for current location")); + return "::" + crate + "::" + name; +} + +/* Return a string referring to a "super::" qualified name. IDENT is + the base name and N_SUPERS is how many "super::"s were provided. + N_SUPERS can be zero. */ + +std::string +rust_parser::super_name (const std::string &ident, unsigned int n_supers) +{ + const char *scope = block_scope (pstate->expression_context_block); + int offset; + + if (scope[0] == '\0') + error (_("Couldn't find namespace scope for self::")); + + if (n_supers > 0) + { + int len; + std::vector offsets; + unsigned int current_len; + + current_len = cp_find_first_component (scope); + while (scope[current_len] != '\0') + { + offsets.push_back (current_len); + gdb_assert (scope[current_len] == ':'); + /* The "::". */ + current_len += 2; + current_len += cp_find_first_component (scope + + current_len); + } + + len = offsets.size (); + if (n_supers >= len) + error (_("Too many super:: uses from '%s'"), scope); + + offset = offsets[len - n_supers]; + } + else + offset = strlen (scope); + + return "::" + std::string (scope, offset) + "::" + ident; +} + +/* A helper to appropriately munge NAME and BLOCK depending on the + presence of a leading "::". */ + +static void +munge_name_and_block (const char **name, const struct block **block) +{ + /* If it is a global reference, skip the current block in favor of + the static block. */ + if (startswith (*name, "::")) + { + *name += 2; + *block = block_static_block (*block); + } +} + +/* Like lookup_symbol, but handles Rust namespace conventions, and + doesn't require field_of_this_result. */ + +struct block_symbol +rust_parser::lookup_symbol (const char *name, const struct block *block, + const domain_enum domain) +{ + struct block_symbol result; + + munge_name_and_block (&name, &block); + + result = ::lookup_symbol (name, block, domain, NULL); + if (result.symbol != NULL) + update_innermost_block (result); + return result; +} + +/* Look up a type, following Rust namespace conventions. */ + +struct type * +rust_parser::rust_lookup_type (const char *name) +{ + struct block_symbol result; + struct type *type; + + const struct block *block = pstate->expression_context_block; + munge_name_and_block (&name, &block); + + result = ::lookup_symbol (name, block, STRUCT_DOMAIN, NULL); + if (result.symbol != NULL) + { + update_innermost_block (result); + return SYMBOL_TYPE (result.symbol); + } + + type = lookup_typename (language (), name, NULL, 1); + if (type != NULL) + return type; + + /* Last chance, try a built-in type. */ + return language_lookup_primitive_type (language (), arch (), name); +} + +/* A helper that updates the innermost block as appropriate. */ + +void +rust_parser::update_innermost_block (struct block_symbol sym) +{ + if (symbol_read_needs_frame (sym.symbol)) + pstate->block_tracker->update (sym); +} + +/* Lex a hex number with at least MIN digits and at most MAX + digits. */ + +uint32_t +rust_parser::lex_hex (int min, int max) +{ + uint32_t result = 0; + int len = 0; + /* We only want to stop at MAX if we're lexing a byte escape. */ + int check_max = min == max; + + while ((check_max ? len <= max : 1) + && ((pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'f') + || (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'F') + || (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9'))) + { + result *= 16; + if (pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'f') + result = result + 10 + pstate->lexptr[0] - 'a'; + else if (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'F') + result = result + 10 + pstate->lexptr[0] - 'A'; + else + result = result + pstate->lexptr[0] - '0'; + ++pstate->lexptr; + ++len; + } + + if (len < min) + error (_("Not enough hex digits seen")); + if (len > max) + { + gdb_assert (min != max); + error (_("Overlong hex escape")); + } + + return result; +} + +/* Lex an escape. IS_BYTE is true if we're lexing a byte escape; + otherwise we're lexing a character escape. */ + +uint32_t +rust_parser::lex_escape (int is_byte) +{ + uint32_t result; + + gdb_assert (pstate->lexptr[0] == '\\'); + ++pstate->lexptr; + switch (pstate->lexptr[0]) + { + case 'x': + ++pstate->lexptr; + result = lex_hex (2, 2); + break; + + case 'u': + if (is_byte) + error (_("Unicode escape in byte literal")); + ++pstate->lexptr; + if (pstate->lexptr[0] != '{') + error (_("Missing '{' in Unicode escape")); + ++pstate->lexptr; + result = lex_hex (1, 6); + /* Could do range checks here. */ + if (pstate->lexptr[0] != '}') + error (_("Missing '}' in Unicode escape")); + ++pstate->lexptr; + break; + + case 'n': + result = '\n'; + ++pstate->lexptr; + break; + case 'r': + result = '\r'; + ++pstate->lexptr; + break; + case 't': + result = '\t'; + ++pstate->lexptr; + break; + case '\\': + result = '\\'; + ++pstate->lexptr; + break; + case '0': + result = '\0'; + ++pstate->lexptr; + break; + case '\'': + result = '\''; + ++pstate->lexptr; + break; + case '"': + result = '"'; + ++pstate->lexptr; + break; + + default: + error (_("Invalid escape \\%c in literal"), pstate->lexptr[0]); + } + + return result; +} + +/* Lex a character constant. */ + +int +rust_parser::lex_character () +{ + int is_byte = 0; + uint32_t value; + + if (pstate->lexptr[0] == 'b') + { + is_byte = 1; + ++pstate->lexptr; + } + gdb_assert (pstate->lexptr[0] == '\''); + ++pstate->lexptr; + /* This should handle UTF-8 here. */ + if (pstate->lexptr[0] == '\\') + value = lex_escape (is_byte); + else + { + value = pstate->lexptr[0] & 0xff; + ++pstate->lexptr; + } + + if (pstate->lexptr[0] != '\'') + error (_("Unterminated character literal")); + ++pstate->lexptr; + + current_int_val.val = value; + current_int_val.type = get_type (is_byte ? "u8" : "char"); + + return INTEGER; +} + +/* Return the offset of the double quote if STR looks like the start + of a raw string, or 0 if STR does not start a raw string. */ + +static int +starts_raw_string (const char *str) +{ + const char *save = str; + + if (str[0] != 'r') + return 0; + ++str; + while (str[0] == '#') + ++str; + if (str[0] == '"') + return str - save; + return 0; +} + +/* Return true if STR looks like the end of a raw string that had N + hashes at the start. */ + +static bool +ends_raw_string (const char *str, int n) +{ + int i; + + gdb_assert (str[0] == '"'); + for (i = 0; i < n; ++i) + if (str[i + 1] != '#') + return false; + return true; +} + +/* Lex a string constant. */ + +int +rust_parser::lex_string () +{ + int is_byte = pstate->lexptr[0] == 'b'; + int raw_length; + + if (is_byte) + ++pstate->lexptr; + raw_length = starts_raw_string (pstate->lexptr); + pstate->lexptr += raw_length; + gdb_assert (pstate->lexptr[0] == '"'); + ++pstate->lexptr; + + while (1) + { + uint32_t value; + + if (raw_length > 0) + { + if (pstate->lexptr[0] == '"' && ends_raw_string (pstate->lexptr, + raw_length - 1)) + { + /* Exit with lexptr pointing after the final "#". */ + pstate->lexptr += raw_length; + break; + } + else if (pstate->lexptr[0] == '\0') + error (_("Unexpected EOF in string")); + + value = pstate->lexptr[0] & 0xff; + if (is_byte && value > 127) + error (_("Non-ASCII value in raw byte string")); + obstack_1grow (&obstack, value); + + ++pstate->lexptr; + } + else if (pstate->lexptr[0] == '"') + { + /* Make sure to skip the quote. */ + ++pstate->lexptr; + break; + } + else if (pstate->lexptr[0] == '\\') + { + value = lex_escape (is_byte); + + if (is_byte) + obstack_1grow (&obstack, value); + else + convert_between_encodings ("UTF-32", "UTF-8", (gdb_byte *) &value, + sizeof (value), sizeof (value), + &obstack, translit_none); + } + else if (pstate->lexptr[0] == '\0') + error (_("Unexpected EOF in string")); + else + { + value = pstate->lexptr[0] & 0xff; + if (is_byte && value > 127) + error (_("Non-ASCII value in byte string")); + obstack_1grow (&obstack, value); + ++pstate->lexptr; + } + } + + current_string_val.length = obstack_object_size (&obstack); + current_string_val.ptr = (const char *) obstack_finish (&obstack); + return is_byte ? BYTESTRING : STRING; +} + +/* Return true if STRING starts with whitespace followed by a digit. */ + +static bool +space_then_number (const char *string) +{ + const char *p = string; + + while (p[0] == ' ' || p[0] == '\t') + ++p; + if (p == string) + return false; + + return *p >= '0' && *p <= '9'; +} + +/* Return true if C can start an identifier. */ + +static bool +rust_identifier_start_p (char c) +{ + return ((c >= 'a' && c <= 'z') + || (c >= 'A' && c <= 'Z') + || c == '_' + || c == '$'); +} + +/* Lex an identifier. */ + +int +rust_parser::lex_identifier () +{ + const char *start = pstate->lexptr; + unsigned int length; + const struct token_info *token; + int i; + int is_gdb_var = pstate->lexptr[0] == '$'; + + gdb_assert (rust_identifier_start_p (pstate->lexptr[0])); + + ++pstate->lexptr; + + /* For the time being this doesn't handle Unicode rules. Non-ASCII + identifiers are gated anyway. */ + while ((pstate->lexptr[0] >= 'a' && pstate->lexptr[0] <= 'z') + || (pstate->lexptr[0] >= 'A' && pstate->lexptr[0] <= 'Z') + || pstate->lexptr[0] == '_' + || (is_gdb_var && pstate->lexptr[0] == '$') + || (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9')) + ++pstate->lexptr; + + + length = pstate->lexptr - start; + token = NULL; + for (i = 0; i < ARRAY_SIZE (identifier_tokens); ++i) + { + if (length == strlen (identifier_tokens[i].name) + && strncmp (identifier_tokens[i].name, start, length) == 0) + { + token = &identifier_tokens[i]; + break; + } + } + + if (token != NULL) + { + if (token->value == 0) + { + /* Leave the terminating token alone. */ + pstate->lexptr = start; + return 0; + } + } + else if (token == NULL + && (strncmp (start, "thread", length) == 0 + || strncmp (start, "task", length) == 0) + && space_then_number (pstate->lexptr)) + { + /* "task" or "thread" followed by a number terminates the + parse, per gdb rules. */ + pstate->lexptr = start; + return 0; + } + + if (token == NULL || (pstate->parse_completion && pstate->lexptr[0] == '\0')) + { + current_string_val.length = length; + current_string_val.ptr = start; + } + + if (pstate->parse_completion && pstate->lexptr[0] == '\0') + { + /* Prevent rustyylex from returning two COMPLETE tokens. */ + pstate->prev_lexptr = pstate->lexptr; + return COMPLETE; + } + + if (token != NULL) + return token->value; + if (is_gdb_var) + return GDBVAR; + return IDENT; +} + +/* Lex an operator. */ + +int +rust_parser::lex_operator () +{ + const struct token_info *token = NULL; + int i; + + for (i = 0; i < ARRAY_SIZE (operator_tokens); ++i) + { + if (strncmp (operator_tokens[i].name, pstate->lexptr, + strlen (operator_tokens[i].name)) == 0) + { + pstate->lexptr += strlen (operator_tokens[i].name); + token = &operator_tokens[i]; + break; + } + } + + if (token != NULL) + { + current_opcode = token->opcode; + return token->value; + } + + return *pstate->lexptr++; +} + +/* Lex a number. */ + +int +rust_parser::lex_number () +{ + regmatch_t subexps[NUM_SUBEXPRESSIONS]; + int match; + int is_integer = 0; + int could_be_decimal = 1; + int implicit_i32 = 0; + const char *type_name = NULL; + struct type *type; + int end_index; + int type_index = -1; + int i; + + match = regexec (&number_regex, pstate->lexptr, ARRAY_SIZE (subexps), + subexps, 0); + /* Failure means the regexp is broken. */ + gdb_assert (match == 0); + + if (subexps[INT_TEXT].rm_so != -1) + { + /* Integer part matched. */ + is_integer = 1; + end_index = subexps[INT_TEXT].rm_eo; + if (subexps[INT_TYPE].rm_so == -1) + { + type_name = "i32"; + implicit_i32 = 1; + } + else + { + type_index = INT_TYPE; + could_be_decimal = 0; + } + } + else if (subexps[FLOAT_TYPE1].rm_so != -1) + { + /* Found floating point type suffix. */ + end_index = subexps[FLOAT_TYPE1].rm_so; + type_index = FLOAT_TYPE1; + } + else if (subexps[FLOAT_TYPE2].rm_so != -1) + { + /* Found floating point type suffix. */ + end_index = subexps[FLOAT_TYPE2].rm_so; + type_index = FLOAT_TYPE2; + } + else + { + /* Any other floating point match. */ + end_index = subexps[0].rm_eo; + type_name = "f64"; + } + + /* We need a special case if the final character is ".". In this + case we might need to parse an integer. For example, "23.f()" is + a request for a trait method call, not a syntax error involving + the floating point number "23.". */ + gdb_assert (subexps[0].rm_eo > 0); + if (pstate->lexptr[subexps[0].rm_eo - 1] == '.') + { + const char *next = skip_spaces (&pstate->lexptr[subexps[0].rm_eo]); + + if (rust_identifier_start_p (*next) || *next == '.') + { + --subexps[0].rm_eo; + is_integer = 1; + end_index = subexps[0].rm_eo; + type_name = "i32"; + could_be_decimal = 1; + implicit_i32 = 1; + } + } + + /* Compute the type name if we haven't already. */ + std::string type_name_holder; + if (type_name == NULL) + { + gdb_assert (type_index != -1); + type_name_holder = std::string ((pstate->lexptr + + subexps[type_index].rm_so), + (subexps[type_index].rm_eo + - subexps[type_index].rm_so)); + type_name = type_name_holder.c_str (); + } + + /* Look up the type. */ + type = get_type (type_name); + + /* Copy the text of the number and remove the "_"s. */ + std::string number; + for (i = 0; i < end_index && pstate->lexptr[i]; ++i) + { + if (pstate->lexptr[i] == '_') + could_be_decimal = 0; + else + number.push_back (pstate->lexptr[i]); + } + + /* Advance past the match. */ + pstate->lexptr += subexps[0].rm_eo; + + /* Parse the number. */ + if (is_integer) + { + uint64_t value; + int radix = 10; + int offset = 0; + + if (number[0] == '0') + { + if (number[1] == 'x') + radix = 16; + else if (number[1] == 'o') + radix = 8; + else if (number[1] == 'b') + radix = 2; + if (radix != 10) + { + offset = 2; + could_be_decimal = 0; + } + } + + value = strtoulst (number.c_str () + offset, NULL, radix); + if (implicit_i32 && value >= ((uint64_t) 1) << 31) + type = get_type ("i64"); + + current_int_val.val = value; + current_int_val.type = type; + } + else + { + current_float_val.type = type; + bool parsed = parse_float (number.c_str (), number.length (), + current_float_val.type, + current_float_val.val.data ()); + gdb_assert (parsed); + } + + return is_integer ? (could_be_decimal ? DECIMAL_INTEGER : INTEGER) : FLOAT; +} + +/* The lexer. */ + +int +rust_parser::lex_one_token () +{ + /* Skip all leading whitespace. */ + while (pstate->lexptr[0] == ' ' + || pstate->lexptr[0] == '\t' + || pstate->lexptr[0] == '\r' + || pstate->lexptr[0] == '\n') + ++pstate->lexptr; + + /* If we hit EOF and we're completing, then return COMPLETE -- maybe + we're completing an empty string at the end of a field_expr. + But, we don't want to return two COMPLETE tokens in a row. */ + if (pstate->lexptr[0] == '\0' && pstate->lexptr == pstate->prev_lexptr) + return 0; + pstate->prev_lexptr = pstate->lexptr; + if (pstate->lexptr[0] == '\0') + { + if (pstate->parse_completion) + { + current_string_val.length =0; + current_string_val.ptr = ""; + return COMPLETE; + } + return 0; + } + + if (pstate->lexptr[0] >= '0' && pstate->lexptr[0] <= '9') + return lex_number (); + else if (pstate->lexptr[0] == 'b' && pstate->lexptr[1] == '\'') + return lex_character (); + else if (pstate->lexptr[0] == 'b' && pstate->lexptr[1] == '"') + return lex_string (); + else if (pstate->lexptr[0] == 'b' && starts_raw_string (pstate->lexptr + 1)) + return lex_string (); + else if (starts_raw_string (pstate->lexptr)) + return lex_string (); + else if (rust_identifier_start_p (pstate->lexptr[0])) + return lex_identifier (); + else if (pstate->lexptr[0] == '"') + return lex_string (); + else if (pstate->lexptr[0] == '\'') + return lex_character (); + else if (pstate->lexptr[0] == '}' || pstate->lexptr[0] == ']') + { + /* Falls through to lex_operator. */ + --paren_depth; + } + else if (pstate->lexptr[0] == '(' || pstate->lexptr[0] == '{') + { + /* Falls through to lex_operator. */ + ++paren_depth; + } + else if (pstate->lexptr[0] == ',' && pstate->comma_terminates + && paren_depth == 0) + return 0; + + return lex_operator (); +} + +/* Push back a single character to be re-lexed. */ + +void +rust_parser::push_back (char c) +{ + /* Can't be called before any lexing. */ + gdb_assert (pstate->prev_lexptr != NULL); + + --pstate->lexptr; + gdb_assert (*pstate->lexptr == c); +} + + + +/* Parse a tuple or paren expression. */ + +operation_up +rust_parser::parse_tuple () +{ + assume ('('); + + if (current_token == ')') + { + lex (); + struct type *unit = get_type ("()"); + return make_operation (unit, 0); + } + + operation_up expr = parse_expr (); + if (current_token == ')') + { + /* Parenthesized expression. */ + lex (); + return expr; + } + + std::vector ops; + ops.push_back (std::move (expr)); + while (current_token != ')') + { + if (current_token != ',') + error (_("',' or ')' expected")); + lex (); + + /* A trailing "," is ok. */ + if (current_token != ')') + ops.push_back (parse_expr ()); + } + + assume (')'); + + error (_("Tuple expressions not supported yet")); +} + +/* Parse an array expression. */ + +operation_up +rust_parser::parse_array () +{ + assume ('['); + + if (current_token == KW_MUT) + lex (); + + operation_up result; + operation_up expr = parse_expr (); + if (current_token == ';') + { + lex (); + operation_up rhs = parse_expr (); + result = make_operation (std::move (expr), + std::move (rhs)); + } + else if (current_token == ',') + { + std::vector ops; + ops.push_back (std::move (expr)); + while (current_token != ']') + { + if (current_token != ',') + error (_("',' or ']' expected")); + lex (); + ops.push_back (parse_expr ()); + } + ops.shrink_to_fit (); + int len = ops.size () - 1; + result = make_operation (0, len, std::move (ops)); + } + else if (current_token != ']') + error (_("',', ';', or ']' expected")); + + require (']'); + + return result; +} + +/* Turn a name into an operation. */ + +operation_up +rust_parser::name_to_operation (const std::string &name) +{ + struct block_symbol sym = lookup_symbol (name.c_str (), + pstate->expression_context_block, + VAR_DOMAIN); + if (sym.symbol != nullptr && SYMBOL_CLASS (sym.symbol) != LOC_TYPEDEF) + return make_operation (sym); + + struct type *type = nullptr; + + if (sym.symbol != nullptr) + { + gdb_assert (SYMBOL_CLASS (sym.symbol) == LOC_TYPEDEF); + type = SYMBOL_TYPE (sym.symbol); + } + if (type == nullptr) + type = rust_lookup_type (name.c_str ()); + if (type == nullptr) + error (_("No symbol '%s' in current context"), name.c_str ()); + + if (type->code () == TYPE_CODE_STRUCT && type->num_fields () == 0) + { + /* A unit-like struct. */ + operation_up result (new rust_aggregate_operation (type, {}, {})); + return result; + } + else + return make_operation (type); +} + +/* Parse a struct expression. */ + +operation_up +rust_parser::parse_struct_expr (struct type *type) +{ + assume ('{'); + + if (type->code () != TYPE_CODE_STRUCT + || rust_tuple_type_p (type) + || rust_tuple_struct_type_p (type)) + error (_("Struct expression applied to non-struct type")); + + std::vector> field_v; + while (current_token != '}' && current_token != DOTDOT) + { + if (current_token != IDENT) + error (_("'}', '..', or identifier expected")); + + std::string name = get_string (); + lex (); + + operation_up expr; + if (current_token == ',' || current_token == '}' + || current_token == DOTDOT) + expr = name_to_operation (name); + else + { + require (':'); + expr = parse_expr (); + } + field_v.emplace_back (std::move (name), std::move (expr)); + + /* A trailing "," is ok. */ + if (current_token == ',') + lex (); + } + + operation_up others; + if (current_token == DOTDOT) + { + lex (); + others = parse_expr (); + } + + require ('}'); + + return make_operation (type, + std::move (others), + std::move (field_v)); +} + +/* Used by the operator precedence parser. */ +struct rustop_item +{ + rustop_item (int token_, int precedence_, enum exp_opcode opcode_, + operation_up &&op_) + : token (token_), + precedence (precedence_), + opcode (opcode_), + op (std::move (op_)) + { + } + + /* The token value. */ + int token; + /* Precedence of this operator. */ + int precedence; + /* This is used only for assign-modify. */ + enum exp_opcode opcode; + /* The right hand side of this operation. */ + operation_up op; +}; + +/* An operator precedence parser for binary operations, including + "as". */ + +operation_up +rust_parser::parse_binop (bool required) +{ + /* All the binary operators. Each one is of the form + OPERATION(TOKEN, PRECEDENCE, TYPE) + TOKEN is the corresponding operator token. + PRECEDENCE is a value indicating relative precedence. + TYPE is the operation type corresponding to the operator. + Assignment operations are handled specially, not via this + table; they have precedence 0. */ +#define ALL_OPS \ + OPERATION ('*', 10, mul_operation) \ + OPERATION ('/', 10, div_operation) \ + OPERATION ('%', 10, rem_operation) \ + OPERATION ('@', 9, repeat_operation) \ + OPERATION ('+', 8, add_operation) \ + OPERATION ('-', 8, sub_operation) \ + OPERATION (LSH, 7, lsh_operation) \ + OPERATION (RSH, 7, rsh_operation) \ + OPERATION ('&', 6, bitwise_and_operation) \ + OPERATION ('^', 5, bitwise_xor_operation) \ + OPERATION ('|', 4, bitwise_ior_operation) \ + OPERATION (EQEQ, 3, equal_operation) \ + OPERATION (NOTEQ, 3, notequal_operation) \ + OPERATION ('<', 3, less_operation) \ + OPERATION (LTEQ, 3, leq_operation) \ + OPERATION ('>', 3, gtr_operation) \ + OPERATION (GTEQ, 3, geq_operation) \ + OPERATION (ANDAND, 2, logical_and_operation) \ + OPERATION (OROR, 1, logical_or_operation) + + operation_up start = parse_atom (required); + if (start == nullptr) + { + gdb_assert (!required); + return start; + } + + std::vector operator_stack; + operator_stack.emplace_back (0, -1, OP_NULL, std::move (start)); + + while (true) + { + int this_token = current_token; + enum exp_opcode compound_assign_op = OP_NULL; + int precedence = -2; + + switch (this_token) + { +#define OPERATION(TOKEN, PRECEDENCE, TYPE) \ + case TOKEN: \ + precedence = PRECEDENCE; \ + lex (); \ + break; + + ALL_OPS + +#undef OPERATION + + case COMPOUND_ASSIGN: + compound_assign_op = current_opcode; + /* FALLTHROUGH */ + case '=': + precedence = 0; + lex (); + break; + + /* "as" must be handled specially. */ + case KW_AS: + { + lex (); + rustop_item &lhs = operator_stack.back (); + struct type *type = parse_type (); + lhs.op = make_operation (std::move (lhs.op), + type); + } + /* Bypass the rest of the loop. */ + continue; + + default: + /* Arrange to pop the entire stack. */ + precedence = -2; + break; + } + + while (precedence < operator_stack.back ().precedence + && operator_stack.size () > 1) + { + rustop_item rhs = std::move (operator_stack.back ()); + operator_stack.pop_back (); + + rustop_item &lhs = operator_stack.back (); + + switch (rhs.token) + { +#define OPERATION(TOKEN, PRECEDENCE, TYPE) \ + case TOKEN: \ + lhs.op = make_operation (std::move (lhs.op), \ + std::move (rhs.op)); \ + break; + + ALL_OPS + +#undef OPERATION + + case '=': + case COMPOUND_ASSIGN: + { + if (rhs.token == '=') + lhs.op = (make_operation + (std::move (lhs.op), std::move (rhs.op))); + else + lhs.op = (make_operation + (rhs.opcode, std::move (lhs.op), + std::move (rhs.op))); + + struct type *unit_type = get_type ("()"); + + operation_up nil (new long_const_operation (unit_type, 0)); + lhs.op = (make_operation + (std::move (lhs.op), std::move (nil))); + } + break; + + default: + gdb_assert_not_reached ("bad binary operator"); + } + } + + if (precedence == -2) + break; + + operator_stack.emplace_back (this_token, precedence, compound_assign_op, + parse_atom (true)); + } + + gdb_assert (operator_stack.size () == 1); + return std::move (operator_stack[0].op); +#undef ALL_OPS +} + +/* Parse a range expression. */ + +operation_up +rust_parser::parse_range () +{ + enum range_flag kind = (RANGE_HIGH_BOUND_DEFAULT + | RANGE_LOW_BOUND_DEFAULT); + + operation_up lhs; + if (current_token != DOTDOT && current_token != DOTDOTEQ) + { + lhs = parse_binop (true); + kind &= ~RANGE_LOW_BOUND_DEFAULT; + } + + if (current_token == DOTDOT) + kind |= RANGE_HIGH_BOUND_EXCLUSIVE; + else if (current_token != DOTDOTEQ) + return lhs; + lex (); + + /* A "..=" range requires a high bound, but otherwise it is + optional. */ + operation_up rhs = parse_binop ((kind & RANGE_HIGH_BOUND_EXCLUSIVE) == 0); + if (rhs != nullptr) + kind &= ~RANGE_HIGH_BOUND_DEFAULT; + + return make_operation (kind, + std::move (lhs), + std::move (rhs)); +} + +/* Parse an expression. */ + +operation_up +rust_parser::parse_expr () +{ + return parse_range (); +} + +/* Parse a sizeof expression. */ + +operation_up +rust_parser::parse_sizeof () +{ + assume (KW_SIZEOF); + + if (current_token == KW_MUT) + lex (); + + require ('('); + operation_up result = make_operation (parse_expr ()); + require (')'); + return result; +} + +/* Parse an address-of operation. */ + +operation_up +rust_parser::parse_addr () +{ + assume ('&'); + + if (current_token == KW_MUT) + lex (); + + return make_operation (parse_atom (true)); +} + +/* Parse a field expression. */ + +operation_up +rust_parser::parse_field (operation_up &&lhs) +{ + assume ('.'); + + operation_up result; + switch (current_token) + { + case IDENT: + case COMPLETE: + { + bool is_complete = current_token == COMPLETE; + auto struct_op = new rust_structop (std::move (lhs), get_string ()); + lex (); + if (is_complete) + { + completion_op.reset (struct_op); + pstate->mark_struct_expression (struct_op); + /* Throw to the outermost level of the parser. */ + error (_("not really an error")); + } + result.reset (struct_op); + } + break; + + case DECIMAL_INTEGER: + result = make_operation (current_int_val.val, + std::move (lhs)); + lex (); + break; + + case INTEGER: + error (_("'_' not allowed in integers in anonymous field references")); + + default: + error (_("field name expected")); + } + + return result; +} + +/* Parse an index expression. */ + +operation_up +rust_parser::parse_index (operation_up &&lhs) +{ + assume ('['); + operation_up rhs = parse_expr (); + require (']'); + + return make_operation (std::move (lhs), + std::move (rhs)); +} + +/* Parse a sequence of comma-separated expressions in parens. */ + +std::vector +rust_parser::parse_paren_args () +{ + assume ('('); + + std::vector args; + while (current_token != ')') + { + if (!args.empty ()) + { + if (current_token != ',') + error (_("',' or ')' expected")); + lex (); + } + + args.push_back (parse_expr ()); + } + + assume (')'); + + return args; +} + +/* Parse the parenthesized part of a function call. */ + +operation_up +rust_parser::parse_call (operation_up &&lhs) +{ + std::vector args = parse_paren_args (); + + return make_operation (std::move (lhs), + std::move (args)); +} + +/* Parse a list of types. */ + +std::vector +rust_parser::parse_type_list () +{ + std::vector result; + result.push_back (parse_type ()); + while (current_token == ',') + { + lex (); + result.push_back (parse_type ()); + } + return result; +} + +/* Parse a possibly-empty list of types, surrounded in parens. */ + +std::vector +rust_parser::parse_maybe_type_list () +{ + assume ('('); + std::vector types; + if (current_token != ')') + types = parse_type_list (); + require (')'); + return types; +} + +/* Parse an array type. */ + +struct type * +rust_parser::parse_array_type () +{ + assume ('['); + struct type *elt_type = parse_type (); + require (';'); + + if (current_token != INTEGER && current_token != DECIMAL_INTEGER) + error (_("integer expected")); + LONGEST val = current_int_val.val; + if (val < 0) + error (_("Negative array length")); + lex (); + require (']'); + + return lookup_array_range_type (elt_type, 0, val - 1); +} + +/* Parse a slice type. */ + +struct type * +rust_parser::parse_slice_type () +{ + assume ('&'); + + bool is_slice = current_token == '['; + if (is_slice) + lex (); + + struct type *target = parse_type (); + + if (is_slice) + { + require (']'); + return rust_slice_type ("&[*gdb*]", target, get_type ("usize")); + } + + /* For now we treat &x and *x identically. */ + return lookup_pointer_type (target); +} + +/* Parse a pointer type. */ + +struct type * +rust_parser::parse_pointer_type () +{ + assume ('*'); + + if (current_token == KW_MUT || current_token == KW_CONST) + lex (); + + struct type *target = parse_type (); + /* For the time being we ignore mut/const. */ + return lookup_pointer_type (target); +} + +/* Parse a function type. */ + +struct type * +rust_parser::parse_function_type () +{ + assume (KW_FN); + + if (current_token != '(') + error (_("'(' expected")); + + std::vector types = parse_maybe_type_list (); + + if (current_token != ARROW) + error (_("'->' expected")); + lex (); + + struct type *result_type = parse_type (); + + struct type **argtypes = nullptr; + if (!types.empty ()) + argtypes = types.data (); + + result_type = lookup_function_type_with_arguments (result_type, + types.size (), + argtypes); + return lookup_pointer_type (result_type); +} + +/* Parse a tuple type. */ + +struct type * +rust_parser::parse_tuple_type () +{ + std::vector types = parse_maybe_type_list (); + + auto_obstack obstack; + obstack_1grow (&obstack, '('); + for (int i = 0; i < types.size (); ++i) + { + std::string type_name = type_to_string (types[i]); + + if (i > 0) + obstack_1grow (&obstack, ','); + obstack_grow_str (&obstack, type_name.c_str ()); + } + + obstack_grow_str0 (&obstack, ")"); + const char *name = (const char *) obstack_finish (&obstack); + + /* We don't allow creating new tuple types (yet), but we do allow + looking up existing tuple types. */ + struct type *result = rust_lookup_type (name); + if (result == nullptr) + error (_("could not find tuple type '%s'"), name); + + return result; +} + +/* Parse a type. */ + +struct type * +rust_parser::parse_type () +{ + switch (current_token) + { + case '[': + return parse_array_type (); + case '&': + return parse_slice_type (); + case '*': + return parse_pointer_type (); + case KW_FN: + return parse_function_type (); + case '(': + return parse_tuple_type (); + case KW_SELF: + case KW_SUPER: + case COLONCOLON: + case KW_EXTERN: + case IDENT: + { + std::string path = parse_path (false); + struct type *result = rust_lookup_type (path.c_str ()); + if (result == nullptr) + error (_("No type name '%s' in current context"), path.c_str ()); + return result; + } + default: + error (_("type expected")); + } +} + +/* Parse a path. */ + +std::string +rust_parser::parse_path (bool for_expr) +{ + unsigned n_supers = 0; + int first_token = current_token; + + switch (current_token) + { + case KW_SELF: + lex (); + if (current_token != COLONCOLON) + return "self"; + lex (); + /* FALLTHROUGH */ + case KW_SUPER: + while (current_token == KW_SUPER) + { + ++n_supers; + lex (); + if (current_token != COLONCOLON) + error (_("'::' expected")); + lex (); + } + break; + + case COLONCOLON: + lex (); + break; + + case KW_EXTERN: + /* This is a gdb extension to make it possible to refer to items + in other crates. It just bypasses adding the current crate + to the front of the name. */ + lex (); + break; + } + + if (current_token != IDENT) + error (_("identifier expected")); + std::string path = get_string (); + bool saw_ident = true; + lex (); + + /* The condition here lets us enter the loop even if we see + "ident<...>". */ + while (current_token == COLONCOLON || current_token == '<') + { + if (current_token == COLONCOLON) + { + lex (); + saw_ident = false; + + if (current_token == IDENT) + { + path = path + "::" + get_string (); + lex (); + saw_ident = true; + } + else if (current_token == COLONCOLON) + { + /* The code below won't detect this scenario. */ + error (_("unexpected '::'")); + } + } + + if (current_token != '<') + continue; + + /* Expression use name::<...>, whereas types use name<...>. */ + if (for_expr) + { + /* Expressions use "name::<...>", so if we saw an identifier + after the "::", we ignore the "<" here. */ + if (saw_ident) + break; + } + else + { + /* Types use "name<...>", so we need to have seen the + identifier. */ + if (!saw_ident) + break; + } + + lex (); + std::vector types = parse_type_list (); + if (current_token == '>') + lex (); + else if (current_token == RSH) + { + push_back ('>'); + lex (); + } + else + error (_("'>' expected")); + + path += "<"; + for (int i = 0; i < types.size (); ++i) + { + if (i > 0) + path += ","; + path += type_to_string (types[i]); + } + path += ">"; + break; + } + + switch (first_token) + { + case KW_SELF: + case KW_SUPER: + return super_name (path, n_supers); + + case COLONCOLON: + return crate_name (path); + + case KW_EXTERN: + return "::" + path; + + case IDENT: + return path; + + default: + gdb_assert_not_reached ("missing case in path parsing"); + } +} + +/* Handle the parsing for a string expression. */ + +operation_up +rust_parser::parse_string () +{ + gdb_assert (current_token == STRING); + + /* Wrap the raw string in the &str struct. */ + struct type *type = rust_lookup_type ("&str"); + if (type == nullptr) + error (_("Could not find type '&str'")); + + std::vector> field_v; + + size_t len = current_string_val.length; + operation_up str = make_operation (get_string ()); + operation_up addr + = make_operation (std::move (str)); + field_v.emplace_back ("data_ptr", std::move (addr)); + + struct type *valtype = get_type ("usize"); + operation_up lenop = make_operation (valtype, len); + field_v.emplace_back ("length", std::move (lenop)); + + return make_operation (type, + operation_up (), + std::move (field_v)); +} + +/* Parse a tuple struct expression. */ + +operation_up +rust_parser::parse_tuple_struct (struct type *type) +{ + std::vector args = parse_paren_args (); + + std::vector> field_v (args.size ()); + for (int i = 0; i < args.size (); ++i) + field_v[i] = { string_printf ("__%d", i), std::move (args[i]) }; + + return (make_operation + (type, operation_up (), std::move (field_v))); +} + +/* Parse a path expression. */ + +operation_up +rust_parser::parse_path_expr () +{ + std::string path = parse_path (true); + + if (current_token == '{') + { + struct type *type = rust_lookup_type (path.c_str ()); + if (type == nullptr) + error (_("Could not find type '%s'"), path.c_str ()); + + return parse_struct_expr (type); + } + else if (current_token == '(') + { + struct type *type = rust_lookup_type (path.c_str ()); + /* If this is actually a tuple struct expression, handle it + here. If it is a call, it will be handled elsewhere. */ + if (type != nullptr) + { + if (!rust_tuple_struct_type_p (type)) + error (_("Type %s is not a tuple struct"), path.c_str ()); + return parse_tuple_struct (type); + } + } + + return name_to_operation (path); +} + +/* Parse an atom. "Atom" isn't a Rust term, but this refers to a + single unitary item in the grammar; but here including some unary + prefix and postfix expressions. */ + +operation_up +rust_parser::parse_atom (bool required) +{ + operation_up result; + + switch (current_token) + { + case '(': + result = parse_tuple (); + break; + + case '[': + result = parse_array (); + break; + + case INTEGER: + case DECIMAL_INTEGER: + result = make_operation (current_int_val.type, + current_int_val.val); + lex (); + break; + + case FLOAT: + result = make_operation (current_float_val.type, + current_float_val.val); + lex (); + break; + + case STRING: + result = parse_string (); + break; + + case BYTESTRING: + result = make_operation (get_string ()); + lex (); + break; + + case KW_TRUE: + case KW_FALSE: + result = make_operation (current_token == KW_TRUE); + lex (); + break; + + case GDBVAR: + /* This is kind of a hacky approach. */ + { + pstate->push_dollar (current_string_val); + result = pstate->pop (); + lex (); + } + break; + + case KW_SELF: + case KW_SUPER: + case COLONCOLON: + case KW_EXTERN: + case IDENT: + result = parse_path_expr (); + break; + + case '*': + lex (); + result = make_operation (parse_atom (true)); + break; + case '+': + lex (); + result = make_operation (parse_atom (true)); + break; + case '-': + lex (); + result = make_operation (parse_atom (true)); + break; + case '!': + lex (); + result = make_operation (parse_atom (true)); + break; + case KW_SIZEOF: + result = parse_sizeof (); + break; + case '&': + result = parse_addr (); + break; + + default: + if (!required) + return {}; + error (_("unexpected token")); + } + + /* Now parse suffixes. */ + while (true) + { + switch (current_token) + { + case '.': + result = parse_field (std::move (result)); + break; + + case '[': + result = parse_index (std::move (result)); + break; + + case '(': + result = parse_call (std::move (result)); + break; + + default: + return result; + } + } +} + + + +/* The parser as exposed to gdb. */ + +int +rust_language::parser (struct parser_state *state) const +{ + rust_parser parser (state); + + operation_up result; + try + { + result = parser.parse_entry_point (); + } + catch (const gdb_exception &exc) + { + if (state->parse_completion) + { + result = std::move (parser.completion_op); + if (result == nullptr) + throw; + } + else + throw; + } + + state->set_operation (std::move (result)); + + return 0; +} + + + +#if GDB_SELF_TEST + +/* A test helper that lexes a string, expecting a single token. */ + +static void +rust_lex_test_one (rust_parser *parser, const char *input, int expected) +{ + int token; + + parser->reset (input); + + token = parser->lex_one_token (); + SELF_CHECK (token == expected); + + if (token) + { + token = parser->lex_one_token (); + SELF_CHECK (token == 0); + } +} + +/* Test that INPUT lexes as the integer VALUE. */ + +static void +rust_lex_int_test (rust_parser *parser, const char *input, + LONGEST value, int kind) +{ + rust_lex_test_one (parser, input, kind); + SELF_CHECK (parser->current_int_val.val == value); +} + +/* Test that INPUT throws an exception with text ERR. */ + +static void +rust_lex_exception_test (rust_parser *parser, const char *input, + const char *err) +{ + try + { + /* The "kind" doesn't matter. */ + rust_lex_test_one (parser, input, DECIMAL_INTEGER); + SELF_CHECK (0); + } + catch (const gdb_exception_error &except) + { + SELF_CHECK (strcmp (except.what (), err) == 0); + } +} + +/* Test that INPUT lexes as the identifier, string, or byte-string + VALUE. KIND holds the expected token kind. */ + +static void +rust_lex_stringish_test (rust_parser *parser, const char *input, + const char *value, int kind) +{ + rust_lex_test_one (parser, input, kind); + SELF_CHECK (parser->get_string () == value); +} + +/* Helper to test that a string parses as a given token sequence. */ + +static void +rust_lex_test_sequence (rust_parser *parser, const char *input, int len, + const int expected[]) +{ + int i; + + parser->reset (input); + + for (i = 0; i < len; ++i) + { + int token = parser->lex_one_token (); + SELF_CHECK (token == expected[i]); + } +} + +/* Tests for an integer-parsing corner case. */ + +static void +rust_lex_test_trailing_dot (rust_parser *parser) +{ + const int expected1[] = { DECIMAL_INTEGER, '.', IDENT, '(', ')', 0 }; + const int expected2[] = { INTEGER, '.', IDENT, '(', ')', 0 }; + const int expected3[] = { FLOAT, EQEQ, '(', ')', 0 }; + const int expected4[] = { DECIMAL_INTEGER, DOTDOT, DECIMAL_INTEGER, 0 }; + + rust_lex_test_sequence (parser, "23.g()", ARRAY_SIZE (expected1), expected1); + rust_lex_test_sequence (parser, "23_0.g()", ARRAY_SIZE (expected2), + expected2); + rust_lex_test_sequence (parser, "23.==()", ARRAY_SIZE (expected3), + expected3); + rust_lex_test_sequence (parser, "23..25", ARRAY_SIZE (expected4), expected4); +} + +/* Tests of completion. */ + +static void +rust_lex_test_completion (rust_parser *parser) +{ + const int expected[] = { IDENT, '.', COMPLETE, 0 }; + + parser->pstate->parse_completion = 1; + + rust_lex_test_sequence (parser, "something.wha", ARRAY_SIZE (expected), + expected); + rust_lex_test_sequence (parser, "something.", ARRAY_SIZE (expected), + expected); + + parser->pstate->parse_completion = 0; +} + +/* Test pushback. */ + +static void +rust_lex_test_push_back (rust_parser *parser) +{ + int token; + + parser->reset (">>="); + + token = parser->lex_one_token (); + SELF_CHECK (token == COMPOUND_ASSIGN); + SELF_CHECK (parser->current_opcode == BINOP_RSH); + + parser->push_back ('='); + + token = parser->lex_one_token (); + SELF_CHECK (token == '='); + + token = parser->lex_one_token (); + SELF_CHECK (token == 0); +} + +/* Unit test the lexer. */ + +static void +rust_lex_tests (void) +{ + int i; + + /* Set up dummy "parser", so that rust_type works. */ + struct parser_state ps (language_def (language_rust), target_gdbarch (), + nullptr, 0, 0, nullptr, 0, nullptr, false); + rust_parser parser (&ps); + + rust_lex_test_one (&parser, "", 0); + rust_lex_test_one (&parser, " \t \n \r ", 0); + rust_lex_test_one (&parser, "thread 23", 0); + rust_lex_test_one (&parser, "task 23", 0); + rust_lex_test_one (&parser, "th 104", 0); + rust_lex_test_one (&parser, "ta 97", 0); + + rust_lex_int_test (&parser, "'z'", 'z', INTEGER); + rust_lex_int_test (&parser, "'\\xff'", 0xff, INTEGER); + rust_lex_int_test (&parser, "'\\u{1016f}'", 0x1016f, INTEGER); + rust_lex_int_test (&parser, "b'z'", 'z', INTEGER); + rust_lex_int_test (&parser, "b'\\xfe'", 0xfe, INTEGER); + rust_lex_int_test (&parser, "b'\\xFE'", 0xfe, INTEGER); + rust_lex_int_test (&parser, "b'\\xfE'", 0xfe, INTEGER); + + /* Test all escapes in both modes. */ + rust_lex_int_test (&parser, "'\\n'", '\n', INTEGER); + rust_lex_int_test (&parser, "'\\r'", '\r', INTEGER); + rust_lex_int_test (&parser, "'\\t'", '\t', INTEGER); + rust_lex_int_test (&parser, "'\\\\'", '\\', INTEGER); + rust_lex_int_test (&parser, "'\\0'", '\0', INTEGER); + rust_lex_int_test (&parser, "'\\''", '\'', INTEGER); + rust_lex_int_test (&parser, "'\\\"'", '"', INTEGER); + + rust_lex_int_test (&parser, "b'\\n'", '\n', INTEGER); + rust_lex_int_test (&parser, "b'\\r'", '\r', INTEGER); + rust_lex_int_test (&parser, "b'\\t'", '\t', INTEGER); + rust_lex_int_test (&parser, "b'\\\\'", '\\', INTEGER); + rust_lex_int_test (&parser, "b'\\0'", '\0', INTEGER); + rust_lex_int_test (&parser, "b'\\''", '\'', INTEGER); + rust_lex_int_test (&parser, "b'\\\"'", '"', INTEGER); + + rust_lex_exception_test (&parser, "'z", "Unterminated character literal"); + rust_lex_exception_test (&parser, "b'\\x0'", "Not enough hex digits seen"); + rust_lex_exception_test (&parser, "b'\\u{0}'", + "Unicode escape in byte literal"); + rust_lex_exception_test (&parser, "'\\x0'", "Not enough hex digits seen"); + rust_lex_exception_test (&parser, "'\\u0'", "Missing '{' in Unicode escape"); + rust_lex_exception_test (&parser, "'\\u{0", "Missing '}' in Unicode escape"); + rust_lex_exception_test (&parser, "'\\u{0000007}", "Overlong hex escape"); + rust_lex_exception_test (&parser, "'\\u{}", "Not enough hex digits seen"); + rust_lex_exception_test (&parser, "'\\Q'", "Invalid escape \\Q in literal"); + rust_lex_exception_test (&parser, "b'\\Q'", "Invalid escape \\Q in literal"); + + rust_lex_int_test (&parser, "23", 23, DECIMAL_INTEGER); + rust_lex_int_test (&parser, "2_344__29", 234429, INTEGER); + rust_lex_int_test (&parser, "0x1f", 0x1f, INTEGER); + rust_lex_int_test (&parser, "23usize", 23, INTEGER); + rust_lex_int_test (&parser, "23i32", 23, INTEGER); + rust_lex_int_test (&parser, "0x1_f", 0x1f, INTEGER); + rust_lex_int_test (&parser, "0b1_101011__", 0x6b, INTEGER); + rust_lex_int_test (&parser, "0o001177i64", 639, INTEGER); + rust_lex_int_test (&parser, "0x123456789u64", 0x123456789ull, INTEGER); + + rust_lex_test_trailing_dot (&parser); + + rust_lex_test_one (&parser, "23.", FLOAT); + rust_lex_test_one (&parser, "23.99f32", FLOAT); + rust_lex_test_one (&parser, "23e7", FLOAT); + rust_lex_test_one (&parser, "23E-7", FLOAT); + rust_lex_test_one (&parser, "23e+7", FLOAT); + rust_lex_test_one (&parser, "23.99e+7f64", FLOAT); + rust_lex_test_one (&parser, "23.82f32", FLOAT); + + rust_lex_stringish_test (&parser, "hibob", "hibob", IDENT); + rust_lex_stringish_test (&parser, "hibob__93", "hibob__93", IDENT); + rust_lex_stringish_test (&parser, "thread", "thread", IDENT); + + rust_lex_stringish_test (&parser, "\"string\"", "string", STRING); + rust_lex_stringish_test (&parser, "\"str\\ting\"", "str\ting", STRING); + rust_lex_stringish_test (&parser, "\"str\\\"ing\"", "str\"ing", STRING); + rust_lex_stringish_test (&parser, "r\"str\\ing\"", "str\\ing", STRING); + rust_lex_stringish_test (&parser, "r#\"str\\ting\"#", "str\\ting", STRING); + rust_lex_stringish_test (&parser, "r###\"str\\\"ing\"###", "str\\\"ing", + STRING); + + rust_lex_stringish_test (&parser, "b\"string\"", "string", BYTESTRING); + rust_lex_stringish_test (&parser, "b\"\x73tring\"", "string", BYTESTRING); + rust_lex_stringish_test (&parser, "b\"str\\\"ing\"", "str\"ing", BYTESTRING); + rust_lex_stringish_test (&parser, "br####\"\\x73tring\"####", "\\x73tring", + BYTESTRING); + + for (i = 0; i < ARRAY_SIZE (identifier_tokens); ++i) + rust_lex_test_one (&parser, identifier_tokens[i].name, + identifier_tokens[i].value); + + for (i = 0; i < ARRAY_SIZE (operator_tokens); ++i) + rust_lex_test_one (&parser, operator_tokens[i].name, + operator_tokens[i].value); + + rust_lex_test_completion (&parser); + rust_lex_test_push_back (&parser); +} + +#endif /* GDB_SELF_TEST */ + + + +void _initialize_rust_exp (); +void +_initialize_rust_exp () +{ + int code = regcomp (&number_regex, number_regex_text, REG_EXTENDED); + /* If the regular expression was incorrect, it was a programming + error. */ + gdb_assert (code == 0); + +#if GDB_SELF_TEST + selftests::register_test ("rust-lex", rust_lex_tests); +#endif +} diff --git a/gdb/testsuite/ChangeLog b/gdb/testsuite/ChangeLog index d659ec4ed1a..995bb783bee 100644 --- a/gdb/testsuite/ChangeLog +++ b/gdb/testsuite/ChangeLog @@ -1,3 +1,8 @@ +2021-04-16 Tom Tromey + + * gdb.rust/simple.exp: Change error text. + * gdb.rust/expr.exp: Change error text. + 2021-04-16 Tom Tromey * gdb.rust/simple.exp: Add parens to 'as' test. diff --git a/gdb/testsuite/gdb.rust/expr.exp b/gdb/testsuite/gdb.rust/expr.exp index d77c780d9ce..d81b6fcbf57 100644 --- a/gdb/testsuite/gdb.rust/expr.exp +++ b/gdb/testsuite/gdb.rust/expr.exp @@ -136,6 +136,10 @@ gdb_test "print \[mut 23usize; 4\]" " = \\\[23, 23, 23, 23\\\]" # Test lexer corner cases. gdb_test "print 0x0 as *mut ()" " = \\\(\\*mut \\\(\\\)\\\) 0x0" gdb_test "print 0x0 as fn(i64) -> ()" " = \\\(\\*mut fn \\\(i64\\\) -> \\\(\\\)\\\) 0x0" -gdb_test "print r#" "syntax error in expression, near `#'\\." + +# The lexer doesn't treat this as a failure, but rather as two tokens, +# and we error out while trying to look up 'r'. This is fine, though +# -- what's important is that it isn't accepted. +gdb_test "print r#" "No symbol 'r' in current context" gdb_test "printf \"%d %d\\n\", 23+1, 23-1" "24 22" diff --git a/gdb/testsuite/gdb.rust/simple.exp b/gdb/testsuite/gdb.rust/simple.exp index 6d5cd319f41..0be0e94ac4c 100644 --- a/gdb/testsuite/gdb.rust/simple.exp +++ b/gdb/testsuite/gdb.rust/simple.exp @@ -165,7 +165,7 @@ gdb_test "print simple::HiBob(0xff, 5)" \ "Type simple::HiBob is not a tuple struct" gdb_test "print sizeof(simple::HiBob)" " = \[0-9\]+" gdb_test "print simple::HiBob + 5" \ - "Found type 'simple::HiBob', which can't be evaluated in this context" + "Attempt to use a type name as an expression" gdb_test "print nosuchsymbol" \ "No symbol 'nosuchsymbol' in current context"