]> git.ipfire.org Git - thirdparty/bash.git/blob - parse.y
Bash-4.1 distribution source
[thirdparty/bash.git] / parse.y
1 /* parse.y - Yacc grammar for bash. */
2
3 /* Copyright (C) 1989-2009 Free Software Foundation, Inc.
4
5 This file is part of GNU Bash, the Bourne Again SHell.
6
7 Bash is free software: you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation, either version 3 of the License, or
10 (at your option) any later version.
11
12 Bash is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
16
17 You should have received a copy of the GNU General Public License
18 along with Bash. If not, see <http://www.gnu.org/licenses/>.
19 */
20
21 %{
22 #include "config.h"
23
24 #include "bashtypes.h"
25 #include "bashansi.h"
26
27 #include "filecntl.h"
28
29 #if defined (HAVE_UNISTD_H)
30 # include <unistd.h>
31 #endif
32
33 #if defined (HAVE_LOCALE_H)
34 # include <locale.h>
35 #endif
36
37 #include <stdio.h>
38 #include "chartypes.h"
39 #include <signal.h>
40
41 #include "memalloc.h"
42
43 #include "bashintl.h"
44
45 #define NEED_STRFTIME_DECL /* used in externs.h */
46
47 #include "shell.h"
48 #include "trap.h"
49 #include "flags.h"
50 #include "parser.h"
51 #include "mailcheck.h"
52 #include "test.h"
53 #include "builtins.h"
54 #include "builtins/common.h"
55 #include "builtins/builtext.h"
56
57 #include "shmbutil.h"
58
59 #if defined (READLINE)
60 # include "bashline.h"
61 # include <readline/readline.h>
62 #endif /* READLINE */
63
64 #if defined (HISTORY)
65 # include "bashhist.h"
66 # include <readline/history.h>
67 #endif /* HISTORY */
68
69 #if defined (JOB_CONTROL)
70 # include "jobs.h"
71 #endif /* JOB_CONTROL */
72
73 #if defined (ALIAS)
74 # include "alias.h"
75 #else
76 typedef void *alias_t;
77 #endif /* ALIAS */
78
79 #if defined (PROMPT_STRING_DECODE)
80 # ifndef _MINIX
81 # include <sys/param.h>
82 # endif
83 # include <time.h>
84 # if defined (TM_IN_SYS_TIME)
85 # include <sys/types.h>
86 # include <sys/time.h>
87 # endif /* TM_IN_SYS_TIME */
88 # include "maxpath.h"
89 #endif /* PROMPT_STRING_DECODE */
90
91 #define RE_READ_TOKEN -99
92 #define NO_EXPANSION -100
93
94 #ifdef DEBUG
95 # define YYDEBUG 1
96 #else
97 # define YYDEBUG 0
98 #endif
99
100 #if defined (HANDLE_MULTIBYTE)
101 # define last_shell_getc_is_singlebyte \
102 ((shell_input_line_index > 1) \
103 ? shell_input_line_property[shell_input_line_index - 1] \
104 : 1)
105 # define MBTEST(x) ((x) && last_shell_getc_is_singlebyte)
106 #else
107 # define last_shell_getc_is_singlebyte 1
108 # define MBTEST(x) ((x))
109 #endif
110
111 #if defined (EXTENDED_GLOB)
112 extern int extended_glob;
113 #endif
114
115 extern int eof_encountered;
116 extern int no_line_editing, running_under_emacs;
117 extern int current_command_number;
118 extern int sourcelevel, parse_and_execute_level;
119 extern int posixly_correct;
120 extern int last_command_exit_value;
121 extern pid_t last_command_subst_pid;
122 extern char *shell_name, *current_host_name;
123 extern char *dist_version;
124 extern int patch_level;
125 extern int dump_translatable_strings, dump_po_strings;
126 extern sh_builtin_func_t *last_shell_builtin, *this_shell_builtin;
127 #if defined (BUFFERED_INPUT)
128 extern int bash_input_fd_changed;
129 #endif
130
131 extern int errno;
132 /* **************************************************************** */
133 /* */
134 /* "Forward" declarations */
135 /* */
136 /* **************************************************************** */
137
138 #ifdef DEBUG
139 static void debug_parser __P((int));
140 #endif
141
142 static int yy_getc __P((void));
143 static int yy_ungetc __P((int));
144
145 #if defined (READLINE)
146 static int yy_readline_get __P((void));
147 static int yy_readline_unget __P((int));
148 #endif
149
150 static int yy_string_get __P((void));
151 static int yy_string_unget __P((int));
152 static void rewind_input_string __P((void));
153 static int yy_stream_get __P((void));
154 static int yy_stream_unget __P((int));
155
156 static int shell_getc __P((int));
157 static void shell_ungetc __P((int));
158 static void discard_until __P((int));
159
160 #if defined (ALIAS) || defined (DPAREN_ARITHMETIC)
161 static void push_string __P((char *, int, alias_t *));
162 static void pop_string __P((void));
163 static void free_string_list __P((void));
164 #endif
165
166 static char *read_a_line __P((int));
167
168 static int reserved_word_acceptable __P((int));
169 static int yylex __P((void));
170 static int alias_expand_token __P((char *));
171 static int time_command_acceptable __P((void));
172 static int special_case_tokens __P((char *));
173 static int read_token __P((int));
174 static char *parse_matched_pair __P((int, int, int, int *, int));
175 static char *parse_comsub __P((int, int, int, int *, int));
176 #if defined (ARRAY_VARS)
177 static char *parse_compound_assignment __P((int *));
178 #endif
179 #if defined (DPAREN_ARITHMETIC) || defined (ARITH_FOR_COMMAND)
180 static int parse_dparen __P((int));
181 static int parse_arith_cmd __P((char **, int));
182 #endif
183 #if defined (COND_COMMAND)
184 static void cond_error __P((void));
185 static COND_COM *cond_expr __P((void));
186 static COND_COM *cond_or __P((void));
187 static COND_COM *cond_and __P((void));
188 static COND_COM *cond_term __P((void));
189 static int cond_skip_newlines __P((void));
190 static COMMAND *parse_cond_command __P((void));
191 #endif
192 #if defined (ARRAY_VARS)
193 static int token_is_assignment __P((char *, int));
194 static int token_is_ident __P((char *, int));
195 #endif
196 static int read_token_word __P((int));
197 static void discard_parser_constructs __P((int));
198
199 static char *error_token_from_token __P((int));
200 static char *error_token_from_text __P((void));
201 static void print_offending_line __P((void));
202 static void report_syntax_error __P((char *));
203
204 static void handle_eof_input_unit __P((void));
205 static void prompt_again __P((void));
206 #if 0
207 static void reset_readline_prompt __P((void));
208 #endif
209 static void print_prompt __P((void));
210
211 #if defined (HANDLE_MULTIBYTE)
212 static void set_line_mbstate __P((void));
213 static char *shell_input_line_property = NULL;
214 #else
215 # define set_line_mbstate()
216 #endif
217
218 extern int yyerror __P((const char *));
219
220 #ifdef DEBUG
221 extern int yydebug;
222 #endif
223
224 /* Default prompt strings */
225 char *primary_prompt = PPROMPT;
226 char *secondary_prompt = SPROMPT;
227
228 /* PROMPT_STRING_POINTER points to one of these, never to an actual string. */
229 char *ps1_prompt, *ps2_prompt;
230
231 /* Handle on the current prompt string. Indirectly points through
232 ps1_ or ps2_prompt. */
233 char **prompt_string_pointer = (char **)NULL;
234 char *current_prompt_string;
235
236 /* Non-zero means we expand aliases in commands. */
237 int expand_aliases = 0;
238
239 /* If non-zero, the decoded prompt string undergoes parameter and
240 variable substitution, command substitution, arithmetic substitution,
241 string expansion, process substitution, and quote removal in
242 decode_prompt_string. */
243 int promptvars = 1;
244
245 /* If non-zero, $'...' and $"..." are expanded when they appear within
246 a ${...} expansion, even when the expansion appears within double
247 quotes. */
248 int extended_quote = 1;
249
250 /* The number of lines read from input while creating the current command. */
251 int current_command_line_count;
252
253 /* The token that currently denotes the end of parse. */
254 int shell_eof_token;
255
256 /* The token currently being read. */
257 int current_token;
258
259 /* The current parser state. */
260 int parser_state;
261
262 /* Variables to manage the task of reading here documents, because we need to
263 defer the reading until after a complete command has been collected. */
264 static REDIRECT *redir_stack[10];
265 int need_here_doc;
266
267 /* Where shell input comes from. History expansion is performed on each
268 line when the shell is interactive. */
269 static char *shell_input_line = (char *)NULL;
270 static int shell_input_line_index;
271 static int shell_input_line_size; /* Amount allocated for shell_input_line. */
272 static int shell_input_line_len; /* strlen (shell_input_line) */
273
274 /* Either zero or EOF. */
275 static int shell_input_line_terminator;
276
277 /* The line number in a script on which a function definition starts. */
278 static int function_dstart;
279
280 /* The line number in a script on which a function body starts. */
281 static int function_bstart;
282
283 /* The line number in a script at which an arithmetic for command starts. */
284 static int arith_for_lineno;
285
286 /* The decoded prompt string. Used if READLINE is not defined or if
287 editing is turned off. Analogous to current_readline_prompt. */
288 static char *current_decoded_prompt;
289
290 /* The last read token, or NULL. read_token () uses this for context
291 checking. */
292 static int last_read_token;
293
294 /* The token read prior to last_read_token. */
295 static int token_before_that;
296
297 /* The token read prior to token_before_that. */
298 static int two_tokens_ago;
299
300 static int global_extglob;
301
302 /* The line number in a script where the word in a `case WORD', `select WORD'
303 or `for WORD' begins. This is a nested command maximum, since the array
304 index is decremented after a case, select, or for command is parsed. */
305 #define MAX_CASE_NEST 128
306 static int word_lineno[MAX_CASE_NEST];
307 static int word_top = -1;
308
309 /* If non-zero, it is the token that we want read_token to return
310 regardless of what text is (or isn't) present to be read. This
311 is reset by read_token. If token_to_read == WORD or
312 ASSIGNMENT_WORD, yylval.word should be set to word_desc_to_read. */
313 static int token_to_read;
314 static WORD_DESC *word_desc_to_read;
315
316 static REDIRECTEE source;
317 static REDIRECTEE redir;
318 %}
319
320 %union {
321 WORD_DESC *word; /* the word that we read. */
322 int number; /* the number that we read. */
323 WORD_LIST *word_list;
324 COMMAND *command;
325 REDIRECT *redirect;
326 ELEMENT element;
327 PATTERN_LIST *pattern;
328 }
329
330 /* Reserved words. Members of the first group are only recognized
331 in the case that they are preceded by a list_terminator. Members
332 of the second group are for [[...]] commands. Members of the
333 third group are recognized only under special circumstances. */
334 %token IF THEN ELSE ELIF FI CASE ESAC FOR SELECT WHILE UNTIL DO DONE FUNCTION COPROC
335 %token COND_START COND_END COND_ERROR
336 %token IN BANG TIME TIMEOPT
337
338 /* More general tokens. yylex () knows how to make these. */
339 %token <word> WORD ASSIGNMENT_WORD REDIR_WORD
340 %token <number> NUMBER
341 %token <word_list> ARITH_CMD ARITH_FOR_EXPRS
342 %token <command> COND_CMD
343 %token AND_AND OR_OR GREATER_GREATER LESS_LESS LESS_AND LESS_LESS_LESS
344 %token GREATER_AND SEMI_SEMI SEMI_AND SEMI_SEMI_AND
345 %token LESS_LESS_MINUS AND_GREATER AND_GREATER_GREATER LESS_GREATER
346 %token GREATER_BAR BAR_AND
347
348 /* The types that the various syntactical units return. */
349
350 %type <command> inputunit command pipeline pipeline_command
351 %type <command> list list0 list1 compound_list simple_list simple_list1
352 %type <command> simple_command shell_command
353 %type <command> for_command select_command case_command group_command
354 %type <command> arith_command
355 %type <command> cond_command
356 %type <command> arith_for_command
357 %type <command> coproc
358 %type <command> function_def function_body if_command elif_clause subshell
359 %type <redirect> redirection redirection_list
360 %type <element> simple_command_element
361 %type <word_list> word_list pattern
362 %type <pattern> pattern_list case_clause_sequence case_clause
363 %type <number> timespec
364 %type <number> list_terminator
365
366 %start inputunit
367
368 %left '&' ';' '\n' yacc_EOF
369 %left AND_AND OR_OR
370 %right '|' BAR_AND
371 %%
372
373 inputunit: simple_list simple_list_terminator
374 {
375 /* Case of regular command. Discard the error
376 safety net,and return the command just parsed. */
377 global_command = $1;
378 eof_encountered = 0;
379 /* discard_parser_constructs (0); */
380 if (parser_state & PST_CMDSUBST)
381 parser_state |= PST_EOFTOKEN;
382 YYACCEPT;
383 }
384 | '\n'
385 {
386 /* Case of regular command, but not a very
387 interesting one. Return a NULL command. */
388 global_command = (COMMAND *)NULL;
389 if (parser_state & PST_CMDSUBST)
390 parser_state |= PST_EOFTOKEN;
391 YYACCEPT;
392 }
393 | error '\n'
394 {
395 /* Error during parsing. Return NULL command. */
396 global_command = (COMMAND *)NULL;
397 eof_encountered = 0;
398 /* discard_parser_constructs (1); */
399 if (interactive && parse_and_execute_level == 0)
400 {
401 YYACCEPT;
402 }
403 else
404 {
405 YYABORT;
406 }
407 }
408 | yacc_EOF
409 {
410 /* Case of EOF seen by itself. Do ignoreeof or
411 not. */
412 global_command = (COMMAND *)NULL;
413 handle_eof_input_unit ();
414 YYACCEPT;
415 }
416 ;
417
418 word_list: WORD
419 { $$ = make_word_list ($1, (WORD_LIST *)NULL); }
420 | word_list WORD
421 { $$ = make_word_list ($2, $1); }
422 ;
423
424 redirection: '>' WORD
425 {
426 source.dest = 1;
427 redir.filename = $2;
428 $$ = make_redirection (source, r_output_direction, redir, 0);
429 }
430 | '<' WORD
431 {
432 source.dest = 0;
433 redir.filename = $2;
434 $$ = make_redirection (source, r_input_direction, redir, 0);
435 }
436 | NUMBER '>' WORD
437 {
438 source.dest = $1;
439 redir.filename = $3;
440 $$ = make_redirection (source, r_output_direction, redir, 0);
441 }
442 | NUMBER '<' WORD
443 {
444 source.dest = $1;
445 redir.filename = $3;
446 $$ = make_redirection (source, r_input_direction, redir, 0);
447 }
448 | REDIR_WORD '>' WORD
449 {
450 source.filename = $1;
451 redir.filename = $3;
452 $$ = make_redirection (source, r_output_direction, redir, REDIR_VARASSIGN);
453 }
454 | REDIR_WORD '<' WORD
455 {
456 source.filename = $1;
457 redir.filename = $3;
458 $$ = make_redirection (source, r_input_direction, redir, REDIR_VARASSIGN);
459 }
460 | GREATER_GREATER WORD
461 {
462 source.dest = 1;
463 redir.filename = $2;
464 $$ = make_redirection (source, r_appending_to, redir, 0);
465 }
466 | NUMBER GREATER_GREATER WORD
467 {
468 source.dest = $1;
469 redir.filename = $3;
470 $$ = make_redirection (source, r_appending_to, redir, 0);
471 }
472 | REDIR_WORD GREATER_GREATER WORD
473 {
474 source.filename = $1;
475 redir.filename = $3;
476 $$ = make_redirection (source, r_appending_to, redir, REDIR_VARASSIGN);
477 }
478 | GREATER_BAR WORD
479 {
480 source.dest = 1;
481 redir.filename = $2;
482 $$ = make_redirection (source, r_output_force, redir, 0);
483 }
484 | NUMBER GREATER_BAR WORD
485 {
486 source.dest = $1;
487 redir.filename = $3;
488 $$ = make_redirection (source, r_output_force, redir, 0);
489 }
490 | REDIR_WORD GREATER_BAR WORD
491 {
492 source.filename = $1;
493 redir.filename = $3;
494 $$ = make_redirection (source, r_output_force, redir, REDIR_VARASSIGN);
495 }
496 | LESS_GREATER WORD
497 {
498 source.dest = 0;
499 redir.filename = $2;
500 $$ = make_redirection (source, r_input_output, redir, 0);
501 }
502 | NUMBER LESS_GREATER WORD
503 {
504 source.dest = $1;
505 redir.filename = $3;
506 $$ = make_redirection (source, r_input_output, redir, 0);
507 }
508 | REDIR_WORD LESS_GREATER WORD
509 {
510 source.filename = $1;
511 redir.filename = $3;
512 $$ = make_redirection (source, r_input_output, redir, REDIR_VARASSIGN);
513 }
514 | LESS_LESS WORD
515 {
516 source.dest = 0;
517 redir.filename = $2;
518 $$ = make_redirection (source, r_reading_until, redir, 0);
519 redir_stack[need_here_doc++] = $$;
520 }
521 | NUMBER LESS_LESS WORD
522 {
523 source.dest = $1;
524 redir.filename = $3;
525 $$ = make_redirection (source, r_reading_until, redir, 0);
526 redir_stack[need_here_doc++] = $$;
527 }
528 | REDIR_WORD LESS_LESS WORD
529 {
530 source.filename = $1;
531 redir.filename = $3;
532 $$ = make_redirection (source, r_reading_until, redir, REDIR_VARASSIGN);
533 redir_stack[need_here_doc++] = $$;
534 }
535 | LESS_LESS_MINUS WORD
536 {
537 source.dest = 0;
538 redir.filename = $2;
539 $$ = make_redirection (source, r_deblank_reading_until, redir, 0);
540 redir_stack[need_here_doc++] = $$;
541 }
542 | NUMBER LESS_LESS_MINUS WORD
543 {
544 source.dest = $1;
545 redir.filename = $3;
546 $$ = make_redirection (source, r_deblank_reading_until, redir, 0);
547 redir_stack[need_here_doc++] = $$;
548 }
549 | REDIR_WORD LESS_LESS_MINUS WORD
550 {
551 source.filename = $1;
552 redir.filename = $3;
553 $$ = make_redirection (source, r_deblank_reading_until, redir, REDIR_VARASSIGN);
554 redir_stack[need_here_doc++] = $$;
555 }
556 | LESS_LESS_LESS WORD
557 {
558 source.dest = 0;
559 redir.filename = $2;
560 $$ = make_redirection (source, r_reading_string, redir, 0);
561 }
562 | NUMBER LESS_LESS_LESS WORD
563 {
564 source.dest = $1;
565 redir.filename = $3;
566 $$ = make_redirection (source, r_reading_string, redir, 0);
567 }
568 | REDIR_WORD LESS_LESS_LESS WORD
569 {
570 source.filename = $1;
571 redir.filename = $3;
572 $$ = make_redirection (source, r_reading_string, redir, REDIR_VARASSIGN);
573 }
574 | LESS_AND NUMBER
575 {
576 source.dest = 0;
577 redir.dest = $2;
578 $$ = make_redirection (source, r_duplicating_input, redir, 0);
579 }
580 | NUMBER LESS_AND NUMBER
581 {
582 source.dest = $1;
583 redir.dest = $3;
584 $$ = make_redirection (source, r_duplicating_input, redir, 0);
585 }
586 | REDIR_WORD LESS_AND NUMBER
587 {
588 source.filename = $1;
589 redir.dest = $3;
590 $$ = make_redirection (source, r_duplicating_input, redir, REDIR_VARASSIGN);
591 }
592 | GREATER_AND NUMBER
593 {
594 source.dest = 1;
595 redir.dest = $2;
596 $$ = make_redirection (source, r_duplicating_output, redir, 0);
597 }
598 | NUMBER GREATER_AND NUMBER
599 {
600 source.dest = $1;
601 redir.dest = $3;
602 $$ = make_redirection (source, r_duplicating_output, redir, 0);
603 }
604 | REDIR_WORD GREATER_AND NUMBER
605 {
606 source.filename = $1;
607 redir.dest = $3;
608 $$ = make_redirection (source, r_duplicating_output, redir, REDIR_VARASSIGN);
609 }
610 | LESS_AND WORD
611 {
612 source.dest = 0;
613 redir.filename = $2;
614 $$ = make_redirection (source, r_duplicating_input_word, redir, 0);
615 }
616 | NUMBER LESS_AND WORD
617 {
618 source.dest = $1;
619 redir.filename = $3;
620 $$ = make_redirection (source, r_duplicating_input_word, redir, 0);
621 }
622 | REDIR_WORD LESS_AND WORD
623 {
624 source.filename = $1;
625 redir.filename = $3;
626 $$ = make_redirection (source, r_duplicating_input_word, redir, REDIR_VARASSIGN);
627 }
628 | GREATER_AND WORD
629 {
630 source.dest = 1;
631 redir.filename = $2;
632 $$ = make_redirection (source, r_duplicating_output_word, redir, 0);
633 }
634 | NUMBER GREATER_AND WORD
635 {
636 source.dest = $1;
637 redir.filename = $3;
638 $$ = make_redirection (source, r_duplicating_output_word, redir, 0);
639 }
640 | REDIR_WORD GREATER_AND WORD
641 {
642 source.filename = $1;
643 redir.filename = $3;
644 $$ = make_redirection (source, r_duplicating_output_word, redir, REDIR_VARASSIGN);
645 }
646 | GREATER_AND '-'
647 {
648 source.dest = 1;
649 redir.dest = 0;
650 $$ = make_redirection (source, r_close_this, redir, 0);
651 }
652 | NUMBER GREATER_AND '-'
653 {
654 source.dest = $1;
655 redir.dest = 0;
656 $$ = make_redirection (source, r_close_this, redir, 0);
657 }
658 | REDIR_WORD GREATER_AND '-'
659 {
660 source.filename = $1;
661 redir.dest = 0;
662 $$ = make_redirection (source, r_close_this, redir, REDIR_VARASSIGN);
663 }
664 | LESS_AND '-'
665 {
666 source.dest = 0;
667 redir.dest = 0;
668 $$ = make_redirection (source, r_close_this, redir, 0);
669 }
670 | NUMBER LESS_AND '-'
671 {
672 source.dest = $1;
673 redir.dest = 0;
674 $$ = make_redirection (source, r_close_this, redir, 0);
675 }
676 | REDIR_WORD LESS_AND '-'
677 {
678 source.filename = $1;
679 redir.dest = 0;
680 $$ = make_redirection (source, r_close_this, redir, REDIR_VARASSIGN);
681 }
682 | AND_GREATER WORD
683 {
684 source.dest = 1;
685 redir.filename = $2;
686 $$ = make_redirection (source, r_err_and_out, redir, 0);
687 }
688 | AND_GREATER_GREATER WORD
689 {
690 source.dest = 1;
691 redir.filename = $2;
692 $$ = make_redirection (source, r_append_err_and_out, redir, 0);
693 }
694 ;
695
696 simple_command_element: WORD
697 { $$.word = $1; $$.redirect = 0; }
698 | ASSIGNMENT_WORD
699 { $$.word = $1; $$.redirect = 0; }
700 | redirection
701 { $$.redirect = $1; $$.word = 0; }
702 ;
703
704 redirection_list: redirection
705 {
706 $$ = $1;
707 }
708 | redirection_list redirection
709 {
710 register REDIRECT *t;
711
712 for (t = $1; t->next; t = t->next)
713 ;
714 t->next = $2;
715 $$ = $1;
716 }
717 ;
718
719 simple_command: simple_command_element
720 { $$ = make_simple_command ($1, (COMMAND *)NULL); }
721 | simple_command simple_command_element
722 { $$ = make_simple_command ($2, $1); }
723 ;
724
725 command: simple_command
726 { $$ = clean_simple_command ($1); }
727 | shell_command
728 { $$ = $1; }
729 | shell_command redirection_list
730 {
731 COMMAND *tc;
732
733 tc = $1;
734 if (tc->redirects)
735 {
736 register REDIRECT *t;
737 for (t = tc->redirects; t->next; t = t->next)
738 ;
739 t->next = $2;
740 }
741 else
742 tc->redirects = $2;
743 $$ = $1;
744 }
745 | function_def
746 { $$ = $1; }
747 | coproc
748 { $$ = $1; }
749 ;
750
751 shell_command: for_command
752 { $$ = $1; }
753 | case_command
754 { $$ = $1; }
755 | WHILE compound_list DO compound_list DONE
756 { $$ = make_while_command ($2, $4); }
757 | UNTIL compound_list DO compound_list DONE
758 { $$ = make_until_command ($2, $4); }
759 | select_command
760 { $$ = $1; }
761 | if_command
762 { $$ = $1; }
763 | subshell
764 { $$ = $1; }
765 | group_command
766 { $$ = $1; }
767 | arith_command
768 { $$ = $1; }
769 | cond_command
770 { $$ = $1; }
771 | arith_for_command
772 { $$ = $1; }
773 ;
774
775 for_command: FOR WORD newline_list DO compound_list DONE
776 {
777 $$ = make_for_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $5, word_lineno[word_top]);
778 if (word_top > 0) word_top--;
779 }
780 | FOR WORD newline_list '{' compound_list '}'
781 {
782 $$ = make_for_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $5, word_lineno[word_top]);
783 if (word_top > 0) word_top--;
784 }
785 | FOR WORD ';' newline_list DO compound_list DONE
786 {
787 $$ = make_for_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $6, word_lineno[word_top]);
788 if (word_top > 0) word_top--;
789 }
790 | FOR WORD ';' newline_list '{' compound_list '}'
791 {
792 $$ = make_for_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $6, word_lineno[word_top]);
793 if (word_top > 0) word_top--;
794 }
795 | FOR WORD newline_list IN word_list list_terminator newline_list DO compound_list DONE
796 {
797 $$ = make_for_command ($2, REVERSE_LIST ($5, WORD_LIST *), $9, word_lineno[word_top]);
798 if (word_top > 0) word_top--;
799 }
800 | FOR WORD newline_list IN word_list list_terminator newline_list '{' compound_list '}'
801 {
802 $$ = make_for_command ($2, REVERSE_LIST ($5, WORD_LIST *), $9, word_lineno[word_top]);
803 if (word_top > 0) word_top--;
804 }
805 | FOR WORD newline_list IN list_terminator newline_list DO compound_list DONE
806 {
807 $$ = make_for_command ($2, (WORD_LIST *)NULL, $8, word_lineno[word_top]);
808 if (word_top > 0) word_top--;
809 }
810 | FOR WORD newline_list IN list_terminator newline_list '{' compound_list '}'
811 {
812 $$ = make_for_command ($2, (WORD_LIST *)NULL, $8, word_lineno[word_top]);
813 if (word_top > 0) word_top--;
814 }
815 ;
816
817 arith_for_command: FOR ARITH_FOR_EXPRS list_terminator newline_list DO compound_list DONE
818 {
819 $$ = make_arith_for_command ($2, $6, arith_for_lineno);
820 if (word_top > 0) word_top--;
821 }
822 | FOR ARITH_FOR_EXPRS list_terminator newline_list '{' compound_list '}'
823 {
824 $$ = make_arith_for_command ($2, $6, arith_for_lineno);
825 if (word_top > 0) word_top--;
826 }
827 | FOR ARITH_FOR_EXPRS DO compound_list DONE
828 {
829 $$ = make_arith_for_command ($2, $4, arith_for_lineno);
830 if (word_top > 0) word_top--;
831 }
832 | FOR ARITH_FOR_EXPRS '{' compound_list '}'
833 {
834 $$ = make_arith_for_command ($2, $4, arith_for_lineno);
835 if (word_top > 0) word_top--;
836 }
837 ;
838
839 select_command: SELECT WORD newline_list DO list DONE
840 {
841 $$ = make_select_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $5, word_lineno[word_top]);
842 if (word_top > 0) word_top--;
843 }
844 | SELECT WORD newline_list '{' list '}'
845 {
846 $$ = make_select_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $5, word_lineno[word_top]);
847 if (word_top > 0) word_top--;
848 }
849 | SELECT WORD ';' newline_list DO list DONE
850 {
851 $$ = make_select_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $6, word_lineno[word_top]);
852 if (word_top > 0) word_top--;
853 }
854 | SELECT WORD ';' newline_list '{' list '}'
855 {
856 $$ = make_select_command ($2, add_string_to_list ("\"$@\"", (WORD_LIST *)NULL), $6, word_lineno[word_top]);
857 if (word_top > 0) word_top--;
858 }
859 | SELECT WORD newline_list IN word_list list_terminator newline_list DO list DONE
860 {
861 $$ = make_select_command ($2, REVERSE_LIST ($5, WORD_LIST *), $9, word_lineno[word_top]);
862 if (word_top > 0) word_top--;
863 }
864 | SELECT WORD newline_list IN word_list list_terminator newline_list '{' list '}'
865 {
866 $$ = make_select_command ($2, REVERSE_LIST ($5, WORD_LIST *), $9, word_lineno[word_top]);
867 if (word_top > 0) word_top--;
868 }
869 ;
870
871 case_command: CASE WORD newline_list IN newline_list ESAC
872 {
873 $$ = make_case_command ($2, (PATTERN_LIST *)NULL, word_lineno[word_top]);
874 if (word_top > 0) word_top--;
875 }
876 | CASE WORD newline_list IN case_clause_sequence newline_list ESAC
877 {
878 $$ = make_case_command ($2, $5, word_lineno[word_top]);
879 if (word_top > 0) word_top--;
880 }
881 | CASE WORD newline_list IN case_clause ESAC
882 {
883 $$ = make_case_command ($2, $5, word_lineno[word_top]);
884 if (word_top > 0) word_top--;
885 }
886 ;
887
888 function_def: WORD '(' ')' newline_list function_body
889 { $$ = make_function_def ($1, $5, function_dstart, function_bstart); }
890
891 | FUNCTION WORD '(' ')' newline_list function_body
892 { $$ = make_function_def ($2, $6, function_dstart, function_bstart); }
893
894 | FUNCTION WORD newline_list function_body
895 { $$ = make_function_def ($2, $4, function_dstart, function_bstart); }
896 ;
897
898 function_body: shell_command
899 { $$ = $1; }
900 | shell_command redirection_list
901 {
902 COMMAND *tc;
903
904 tc = $1;
905 /* According to Posix.2 3.9.5, redirections
906 specified after the body of a function should
907 be attached to the function and performed when
908 the function is executed, not as part of the
909 function definition command. */
910 /* XXX - I don't think it matters, but we might
911 want to change this in the future to avoid
912 problems differentiating between a function
913 definition with a redirection and a function
914 definition containing a single command with a
915 redirection. The two are semantically equivalent,
916 though -- the only difference is in how the
917 command printing code displays the redirections. */
918 if (tc->redirects)
919 {
920 register REDIRECT *t;
921 for (t = tc->redirects; t->next; t = t->next)
922 ;
923 t->next = $2;
924 }
925 else
926 tc->redirects = $2;
927 $$ = $1;
928 }
929 ;
930
931 subshell: '(' compound_list ')'
932 {
933 $$ = make_subshell_command ($2);
934 $$->flags |= CMD_WANT_SUBSHELL;
935 }
936 ;
937
938 coproc: COPROC shell_command
939 {
940 $$ = make_coproc_command ("COPROC", $2);
941 $$->flags |= CMD_WANT_SUBSHELL|CMD_COPROC_SUBSHELL;
942 }
943 | COPROC shell_command redirection_list
944 {
945 COMMAND *tc;
946
947 tc = $2;
948 if (tc->redirects)
949 {
950 register REDIRECT *t;
951 for (t = tc->redirects; t->next; t = t->next)
952 ;
953 t->next = $3;
954 }
955 else
956 tc->redirects = $3;
957 $$ = make_coproc_command ("COPROC", $2);
958 $$->flags |= CMD_WANT_SUBSHELL|CMD_COPROC_SUBSHELL;
959 }
960 | COPROC WORD shell_command
961 {
962 $$ = make_coproc_command ($2->word, $3);
963 $$->flags |= CMD_WANT_SUBSHELL|CMD_COPROC_SUBSHELL;
964 }
965 | COPROC WORD shell_command redirection_list
966 {
967 COMMAND *tc;
968
969 tc = $3;
970 if (tc->redirects)
971 {
972 register REDIRECT *t;
973 for (t = tc->redirects; t->next; t = t->next)
974 ;
975 t->next = $4;
976 }
977 else
978 tc->redirects = $4;
979 $$ = make_coproc_command ($2->word, $3);
980 $$->flags |= CMD_WANT_SUBSHELL|CMD_COPROC_SUBSHELL;
981 }
982 | COPROC simple_command
983 {
984 $$ = make_coproc_command ("COPROC", clean_simple_command ($2));
985 $$->flags |= CMD_WANT_SUBSHELL|CMD_COPROC_SUBSHELL;
986 }
987 ;
988
989 if_command: IF compound_list THEN compound_list FI
990 { $$ = make_if_command ($2, $4, (COMMAND *)NULL); }
991 | IF compound_list THEN compound_list ELSE compound_list FI
992 { $$ = make_if_command ($2, $4, $6); }
993 | IF compound_list THEN compound_list elif_clause FI
994 { $$ = make_if_command ($2, $4, $5); }
995 ;
996
997
998 group_command: '{' compound_list '}'
999 { $$ = make_group_command ($2); }
1000 ;
1001
1002 arith_command: ARITH_CMD
1003 { $$ = make_arith_command ($1); }
1004 ;
1005
1006 cond_command: COND_START COND_CMD COND_END
1007 { $$ = $2; }
1008 ;
1009
1010 elif_clause: ELIF compound_list THEN compound_list
1011 { $$ = make_if_command ($2, $4, (COMMAND *)NULL); }
1012 | ELIF compound_list THEN compound_list ELSE compound_list
1013 { $$ = make_if_command ($2, $4, $6); }
1014 | ELIF compound_list THEN compound_list elif_clause
1015 { $$ = make_if_command ($2, $4, $5); }
1016 ;
1017
1018 case_clause: pattern_list
1019 | case_clause_sequence pattern_list
1020 { $2->next = $1; $$ = $2; }
1021 ;
1022
1023 pattern_list: newline_list pattern ')' compound_list
1024 { $$ = make_pattern_list ($2, $4); }
1025 | newline_list pattern ')' newline_list
1026 { $$ = make_pattern_list ($2, (COMMAND *)NULL); }
1027 | newline_list '(' pattern ')' compound_list
1028 { $$ = make_pattern_list ($3, $5); }
1029 | newline_list '(' pattern ')' newline_list
1030 { $$ = make_pattern_list ($3, (COMMAND *)NULL); }
1031 ;
1032
1033 case_clause_sequence: pattern_list SEMI_SEMI
1034 { $$ = $1; }
1035 | case_clause_sequence pattern_list SEMI_SEMI
1036 { $2->next = $1; $$ = $2; }
1037 | pattern_list SEMI_AND
1038 { $1->flags |= CASEPAT_FALLTHROUGH; $$ = $1; }
1039 | case_clause_sequence pattern_list SEMI_AND
1040 { $2->flags |= CASEPAT_FALLTHROUGH; $2->next = $1; $$ = $2; }
1041 | pattern_list SEMI_SEMI_AND
1042 { $1->flags |= CASEPAT_TESTNEXT; $$ = $1; }
1043 | case_clause_sequence pattern_list SEMI_SEMI_AND
1044 { $2->flags |= CASEPAT_TESTNEXT; $2->next = $1; $$ = $2; }
1045 ;
1046
1047 pattern: WORD
1048 { $$ = make_word_list ($1, (WORD_LIST *)NULL); }
1049 | pattern '|' WORD
1050 { $$ = make_word_list ($3, $1); }
1051 ;
1052
1053 /* A list allows leading or trailing newlines and
1054 newlines as operators (equivalent to semicolons).
1055 It must end with a newline or semicolon.
1056 Lists are used within commands such as if, for, while. */
1057
1058 list: newline_list list0
1059 {
1060 $$ = $2;
1061 if (need_here_doc)
1062 gather_here_documents ();
1063 }
1064 ;
1065
1066 compound_list: list
1067 | newline_list list1
1068 {
1069 $$ = $2;
1070 }
1071 ;
1072
1073 list0: list1 '\n' newline_list
1074 | list1 '&' newline_list
1075 {
1076 if ($1->type == cm_connection)
1077 $$ = connect_async_list ($1, (COMMAND *)NULL, '&');
1078 else
1079 $$ = command_connect ($1, (COMMAND *)NULL, '&');
1080 }
1081 | list1 ';' newline_list
1082
1083 ;
1084
1085 list1: list1 AND_AND newline_list list1
1086 { $$ = command_connect ($1, $4, AND_AND); }
1087 | list1 OR_OR newline_list list1
1088 { $$ = command_connect ($1, $4, OR_OR); }
1089 | list1 '&' newline_list list1
1090 {
1091 if ($1->type == cm_connection)
1092 $$ = connect_async_list ($1, $4, '&');
1093 else
1094 $$ = command_connect ($1, $4, '&');
1095 }
1096 | list1 ';' newline_list list1
1097 { $$ = command_connect ($1, $4, ';'); }
1098 | list1 '\n' newline_list list1
1099 { $$ = command_connect ($1, $4, ';'); }
1100 | pipeline_command
1101 { $$ = $1; }
1102 ;
1103
1104 simple_list_terminator: '\n'
1105 | yacc_EOF
1106 ;
1107
1108 list_terminator:'\n'
1109 { $$ = '\n'; }
1110 | ';'
1111 { $$ = ';'; }
1112 | yacc_EOF
1113 { $$ = yacc_EOF; }
1114 ;
1115
1116 newline_list:
1117 | newline_list '\n'
1118 ;
1119
1120 /* A simple_list is a list that contains no significant newlines
1121 and no leading or trailing newlines. Newlines are allowed
1122 only following operators, where they are not significant.
1123
1124 This is what an inputunit consists of. */
1125
1126 simple_list: simple_list1
1127 {
1128 $$ = $1;
1129 if (need_here_doc)
1130 gather_here_documents ();
1131 if ((parser_state & PST_CMDSUBST) && current_token == shell_eof_token)
1132 {
1133 global_command = $1;
1134 eof_encountered = 0;
1135 rewind_input_string ();
1136 YYACCEPT;
1137 }
1138 }
1139 | simple_list1 '&'
1140 {
1141 if ($1->type == cm_connection)
1142 $$ = connect_async_list ($1, (COMMAND *)NULL, '&');
1143 else
1144 $$ = command_connect ($1, (COMMAND *)NULL, '&');
1145 if (need_here_doc)
1146 gather_here_documents ();
1147 if ((parser_state & PST_CMDSUBST) && current_token == shell_eof_token)
1148 {
1149 global_command = $1;
1150 eof_encountered = 0;
1151 rewind_input_string ();
1152 YYACCEPT;
1153 }
1154 }
1155 | simple_list1 ';'
1156 {
1157 $$ = $1;
1158 if (need_here_doc)
1159 gather_here_documents ();
1160 if ((parser_state & PST_CMDSUBST) && current_token == shell_eof_token)
1161 {
1162 global_command = $1;
1163 eof_encountered = 0;
1164 rewind_input_string ();
1165 YYACCEPT;
1166 }
1167 }
1168 ;
1169
1170 simple_list1: simple_list1 AND_AND newline_list simple_list1
1171 { $$ = command_connect ($1, $4, AND_AND); }
1172 | simple_list1 OR_OR newline_list simple_list1
1173 { $$ = command_connect ($1, $4, OR_OR); }
1174 | simple_list1 '&' simple_list1
1175 {
1176 if ($1->type == cm_connection)
1177 $$ = connect_async_list ($1, $3, '&');
1178 else
1179 $$ = command_connect ($1, $3, '&');
1180 }
1181 | simple_list1 ';' simple_list1
1182 { $$ = command_connect ($1, $3, ';'); }
1183
1184 | pipeline_command
1185 { $$ = $1; }
1186 ;
1187
1188 pipeline_command: pipeline
1189 { $$ = $1; }
1190 | BANG pipeline
1191 {
1192 if ($2)
1193 $2->flags |= CMD_INVERT_RETURN;
1194 $$ = $2;
1195 }
1196 | timespec pipeline
1197 {
1198 if ($2)
1199 $2->flags |= $1;
1200 $$ = $2;
1201 }
1202 | timespec BANG pipeline
1203 {
1204 if ($3)
1205 $3->flags |= $1|CMD_INVERT_RETURN;
1206 $$ = $3;
1207 }
1208 | BANG timespec pipeline
1209 {
1210 if ($3)
1211 $3->flags |= $2|CMD_INVERT_RETURN;
1212 $$ = $3;
1213 }
1214 | timespec list_terminator
1215 {
1216 ELEMENT x;
1217
1218 /* Boy, this is unclean. `time' by itself can
1219 time a null command. We cheat and push a
1220 newline back if the list_terminator was a newline
1221 to avoid the double-newline problem (one to
1222 terminate this, one to terminate the command) */
1223 x.word = 0;
1224 x.redirect = 0;
1225 $$ = make_simple_command (x, (COMMAND *)NULL);
1226 $$->flags |= $1;
1227 /* XXX - let's cheat and push a newline back */
1228 if ($2 == '\n')
1229 token_to_read = '\n';
1230 }
1231
1232 ;
1233
1234 pipeline: pipeline '|' newline_list pipeline
1235 { $$ = command_connect ($1, $4, '|'); }
1236 | pipeline BAR_AND newline_list pipeline
1237 {
1238 /* Make cmd1 |& cmd2 equivalent to cmd1 2>&1 | cmd2 */
1239 COMMAND *tc;
1240 REDIRECTEE rd, sd;
1241 REDIRECT *r;
1242
1243 tc = $1->type == cm_simple ? (COMMAND *)$1->value.Simple : $1;
1244 sd.dest = 2;
1245 rd.dest = 1;
1246 r = make_redirection (sd, r_duplicating_output, rd, 0);
1247 if (tc->redirects)
1248 {
1249 register REDIRECT *t;
1250 for (t = tc->redirects; t->next; t = t->next)
1251 ;
1252 t->next = r;
1253 }
1254 else
1255 tc->redirects = r;
1256
1257 $$ = command_connect ($1, $4, '|');
1258 }
1259 | command
1260 { $$ = $1; }
1261 ;
1262
1263 timespec: TIME
1264 { $$ = CMD_TIME_PIPELINE; }
1265 | TIME TIMEOPT
1266 { $$ = CMD_TIME_PIPELINE|CMD_TIME_POSIX; }
1267 ;
1268 %%
1269
1270 /* Initial size to allocate for tokens, and the
1271 amount to grow them by. */
1272 #define TOKEN_DEFAULT_INITIAL_SIZE 496
1273 #define TOKEN_DEFAULT_GROW_SIZE 512
1274
1275 /* Should we call prompt_again? */
1276 #define SHOULD_PROMPT() \
1277 (interactive && (bash_input.type == st_stdin || bash_input.type == st_stream))
1278
1279 #if defined (ALIAS)
1280 # define expanding_alias() (pushed_string_list && pushed_string_list->expander)
1281 #else
1282 # define expanding_alias() 0
1283 #endif
1284
1285 /* Global var is non-zero when end of file has been reached. */
1286 int EOF_Reached = 0;
1287
1288 #ifdef DEBUG
1289 static void
1290 debug_parser (i)
1291 int i;
1292 {
1293 #if YYDEBUG != 0
1294 yydebug = i;
1295 #endif
1296 }
1297 #endif
1298
1299 /* yy_getc () returns the next available character from input or EOF.
1300 yy_ungetc (c) makes `c' the next character to read.
1301 init_yy_io (get, unget, type, location) makes the function GET the
1302 installed function for getting the next character, makes UNGET the
1303 installed function for un-getting a character, sets the type of stream
1304 (either string or file) from TYPE, and makes LOCATION point to where
1305 the input is coming from. */
1306
1307 /* Unconditionally returns end-of-file. */
1308 int
1309 return_EOF ()
1310 {
1311 return (EOF);
1312 }
1313
1314 /* Variable containing the current get and unget functions.
1315 See ./input.h for a clearer description. */
1316 BASH_INPUT bash_input;
1317
1318 /* Set all of the fields in BASH_INPUT to NULL. Free bash_input.name if it
1319 is non-null, avoiding a memory leak. */
1320 void
1321 initialize_bash_input ()
1322 {
1323 bash_input.type = st_none;
1324 FREE (bash_input.name);
1325 bash_input.name = (char *)NULL;
1326 bash_input.location.file = (FILE *)NULL;
1327 bash_input.location.string = (char *)NULL;
1328 bash_input.getter = (sh_cget_func_t *)NULL;
1329 bash_input.ungetter = (sh_cunget_func_t *)NULL;
1330 }
1331
1332 /* Set the contents of the current bash input stream from
1333 GET, UNGET, TYPE, NAME, and LOCATION. */
1334 void
1335 init_yy_io (get, unget, type, name, location)
1336 sh_cget_func_t *get;
1337 sh_cunget_func_t *unget;
1338 enum stream_type type;
1339 const char *name;
1340 INPUT_STREAM location;
1341 {
1342 bash_input.type = type;
1343 FREE (bash_input.name);
1344 bash_input.name = name ? savestring (name) : (char *)NULL;
1345
1346 /* XXX */
1347 #if defined (CRAY)
1348 memcpy((char *)&bash_input.location.string, (char *)&location.string, sizeof(location));
1349 #else
1350 bash_input.location = location;
1351 #endif
1352 bash_input.getter = get;
1353 bash_input.ungetter = unget;
1354 }
1355
1356 char *
1357 yy_input_name ()
1358 {
1359 return (bash_input.name ? bash_input.name : "stdin");
1360 }
1361
1362 /* Call this to get the next character of input. */
1363 static int
1364 yy_getc ()
1365 {
1366 return (*(bash_input.getter)) ();
1367 }
1368
1369 /* Call this to unget C. That is, to make C the next character
1370 to be read. */
1371 static int
1372 yy_ungetc (c)
1373 int c;
1374 {
1375 return (*(bash_input.ungetter)) (c);
1376 }
1377
1378 #if defined (BUFFERED_INPUT)
1379 #ifdef INCLUDE_UNUSED
1380 int
1381 input_file_descriptor ()
1382 {
1383 switch (bash_input.type)
1384 {
1385 case st_stream:
1386 return (fileno (bash_input.location.file));
1387 case st_bstream:
1388 return (bash_input.location.buffered_fd);
1389 case st_stdin:
1390 default:
1391 return (fileno (stdin));
1392 }
1393 }
1394 #endif
1395 #endif /* BUFFERED_INPUT */
1396
1397 /* **************************************************************** */
1398 /* */
1399 /* Let input be read from readline (). */
1400 /* */
1401 /* **************************************************************** */
1402
1403 #if defined (READLINE)
1404 char *current_readline_prompt = (char *)NULL;
1405 char *current_readline_line = (char *)NULL;
1406 int current_readline_line_index = 0;
1407
1408 static int
1409 yy_readline_get ()
1410 {
1411 SigHandler *old_sigint;
1412 int line_len;
1413 unsigned char c;
1414
1415 if (!current_readline_line)
1416 {
1417 if (!bash_readline_initialized)
1418 initialize_readline ();
1419
1420 #if defined (JOB_CONTROL)
1421 if (job_control)
1422 give_terminal_to (shell_pgrp, 0);
1423 #endif /* JOB_CONTROL */
1424
1425 old_sigint = (SigHandler *)NULL;
1426 if (signal_is_ignored (SIGINT) == 0)
1427 {
1428 old_sigint = (SigHandler *)set_signal_handler (SIGINT, sigint_sighandler);
1429 interrupt_immediately++;
1430 }
1431 terminate_immediately = 1;
1432
1433 current_readline_line = readline (current_readline_prompt ?
1434 current_readline_prompt : "");
1435
1436 terminate_immediately = 0;
1437 if (signal_is_ignored (SIGINT) == 0 && old_sigint)
1438 {
1439 interrupt_immediately--;
1440 set_signal_handler (SIGINT, old_sigint);
1441 }
1442
1443 #if 0
1444 /* Reset the prompt to the decoded value of prompt_string_pointer. */
1445 reset_readline_prompt ();
1446 #endif
1447
1448 if (current_readline_line == 0)
1449 return (EOF);
1450
1451 current_readline_line_index = 0;
1452 line_len = strlen (current_readline_line);
1453
1454 current_readline_line = (char *)xrealloc (current_readline_line, 2 + line_len);
1455 current_readline_line[line_len++] = '\n';
1456 current_readline_line[line_len] = '\0';
1457 }
1458
1459 if (current_readline_line[current_readline_line_index] == 0)
1460 {
1461 free (current_readline_line);
1462 current_readline_line = (char *)NULL;
1463 return (yy_readline_get ());
1464 }
1465 else
1466 {
1467 c = current_readline_line[current_readline_line_index++];
1468 return (c);
1469 }
1470 }
1471
1472 static int
1473 yy_readline_unget (c)
1474 int c;
1475 {
1476 if (current_readline_line_index && current_readline_line)
1477 current_readline_line[--current_readline_line_index] = c;
1478 return (c);
1479 }
1480
1481 void
1482 with_input_from_stdin ()
1483 {
1484 INPUT_STREAM location;
1485
1486 if (bash_input.type != st_stdin && stream_on_stack (st_stdin) == 0)
1487 {
1488 location.string = current_readline_line;
1489 init_yy_io (yy_readline_get, yy_readline_unget,
1490 st_stdin, "readline stdin", location);
1491 }
1492 }
1493
1494 #else /* !READLINE */
1495
1496 void
1497 with_input_from_stdin ()
1498 {
1499 with_input_from_stream (stdin, "stdin");
1500 }
1501 #endif /* !READLINE */
1502
1503 /* **************************************************************** */
1504 /* */
1505 /* Let input come from STRING. STRING is zero terminated. */
1506 /* */
1507 /* **************************************************************** */
1508
1509 static int
1510 yy_string_get ()
1511 {
1512 register char *string;
1513 register unsigned char c;
1514
1515 string = bash_input.location.string;
1516
1517 /* If the string doesn't exist, or is empty, EOF found. */
1518 if (string && *string)
1519 {
1520 c = *string++;
1521 bash_input.location.string = string;
1522 return (c);
1523 }
1524 else
1525 return (EOF);
1526 }
1527
1528 static int
1529 yy_string_unget (c)
1530 int c;
1531 {
1532 *(--bash_input.location.string) = c;
1533 return (c);
1534 }
1535
1536 void
1537 with_input_from_string (string, name)
1538 char *string;
1539 const char *name;
1540 {
1541 INPUT_STREAM location;
1542
1543 location.string = string;
1544 init_yy_io (yy_string_get, yy_string_unget, st_string, name, location);
1545 }
1546
1547 /* Count the number of characters we've consumed from bash_input.location.string
1548 and read into shell_input_line, but have not returned from shell_getc.
1549 That is the true input location. Rewind bash_input.location.string by
1550 that number of characters, so it points to the last character actually
1551 consumed by the parser. */
1552 static void
1553 rewind_input_string ()
1554 {
1555 int xchars;
1556
1557 /* number of unconsumed characters in the input -- XXX need to take newlines
1558 into account, e.g., $(...\n) */
1559 xchars = shell_input_line_len - shell_input_line_index;
1560 if (bash_input.location.string[-1] == '\n')
1561 xchars++;
1562
1563 /* XXX - how to reflect bash_input.location.string back to string passed to
1564 parse_and_execute or xparse_dolparen? xparse_dolparen needs to know how
1565 far into the string we parsed. parse_and_execute knows where bash_input.
1566 location.string is, and how far from orig_string that is -- that's the
1567 number of characters the command consumed. */
1568
1569 /* bash_input.location.string - xchars should be where we parsed to */
1570 /* need to do more validation on xchars value for sanity -- test cases. */
1571 bash_input.location.string -= xchars;
1572 }
1573
1574 /* **************************************************************** */
1575 /* */
1576 /* Let input come from STREAM. */
1577 /* */
1578 /* **************************************************************** */
1579
1580 /* These two functions used to test the value of the HAVE_RESTARTABLE_SYSCALLS
1581 define, and just use getc/ungetc if it was defined, but since bash
1582 installs its signal handlers without the SA_RESTART flag, some signals
1583 (like SIGCHLD, SIGWINCH, etc.) received during a read(2) will not cause
1584 the read to be restarted. We need to restart it ourselves. */
1585
1586 static int
1587 yy_stream_get ()
1588 {
1589 int result;
1590
1591 result = EOF;
1592 if (bash_input.location.file)
1593 {
1594 if (interactive)
1595 {
1596 interrupt_immediately++;
1597 terminate_immediately++;
1598 }
1599 result = getc_with_restart (bash_input.location.file);
1600 if (interactive)
1601 {
1602 interrupt_immediately--;
1603 terminate_immediately--;
1604 }
1605 }
1606 return (result);
1607 }
1608
1609 static int
1610 yy_stream_unget (c)
1611 int c;
1612 {
1613 return (ungetc_with_restart (c, bash_input.location.file));
1614 }
1615
1616 void
1617 with_input_from_stream (stream, name)
1618 FILE *stream;
1619 const char *name;
1620 {
1621 INPUT_STREAM location;
1622
1623 location.file = stream;
1624 init_yy_io (yy_stream_get, yy_stream_unget, st_stream, name, location);
1625 }
1626
1627 typedef struct stream_saver {
1628 struct stream_saver *next;
1629 BASH_INPUT bash_input;
1630 int line;
1631 #if defined (BUFFERED_INPUT)
1632 BUFFERED_STREAM *bstream;
1633 #endif /* BUFFERED_INPUT */
1634 } STREAM_SAVER;
1635
1636 /* The globally known line number. */
1637 int line_number = 0;
1638
1639 #if defined (COND_COMMAND)
1640 static int cond_lineno;
1641 static int cond_token;
1642 #endif
1643
1644 STREAM_SAVER *stream_list = (STREAM_SAVER *)NULL;
1645
1646 void
1647 push_stream (reset_lineno)
1648 int reset_lineno;
1649 {
1650 STREAM_SAVER *saver = (STREAM_SAVER *)xmalloc (sizeof (STREAM_SAVER));
1651
1652 xbcopy ((char *)&bash_input, (char *)&(saver->bash_input), sizeof (BASH_INPUT));
1653
1654 #if defined (BUFFERED_INPUT)
1655 saver->bstream = (BUFFERED_STREAM *)NULL;
1656 /* If we have a buffered stream, clear out buffers[fd]. */
1657 if (bash_input.type == st_bstream && bash_input.location.buffered_fd >= 0)
1658 saver->bstream = set_buffered_stream (bash_input.location.buffered_fd,
1659 (BUFFERED_STREAM *)NULL);
1660 #endif /* BUFFERED_INPUT */
1661
1662 saver->line = line_number;
1663 bash_input.name = (char *)NULL;
1664 saver->next = stream_list;
1665 stream_list = saver;
1666 EOF_Reached = 0;
1667 if (reset_lineno)
1668 line_number = 0;
1669 }
1670
1671 void
1672 pop_stream ()
1673 {
1674 if (!stream_list)
1675 EOF_Reached = 1;
1676 else
1677 {
1678 STREAM_SAVER *saver = stream_list;
1679
1680 EOF_Reached = 0;
1681 stream_list = stream_list->next;
1682
1683 init_yy_io (saver->bash_input.getter,
1684 saver->bash_input.ungetter,
1685 saver->bash_input.type,
1686 saver->bash_input.name,
1687 saver->bash_input.location);
1688
1689 #if defined (BUFFERED_INPUT)
1690 /* If we have a buffered stream, restore buffers[fd]. */
1691 /* If the input file descriptor was changed while this was on the
1692 save stack, update the buffered fd to the new file descriptor and
1693 re-establish the buffer <-> bash_input fd correspondence. */
1694 if (bash_input.type == st_bstream && bash_input.location.buffered_fd >= 0)
1695 {
1696 if (bash_input_fd_changed)
1697 {
1698 bash_input_fd_changed = 0;
1699 if (default_buffered_input >= 0)
1700 {
1701 bash_input.location.buffered_fd = default_buffered_input;
1702 saver->bstream->b_fd = default_buffered_input;
1703 SET_CLOSE_ON_EXEC (default_buffered_input);
1704 }
1705 }
1706 /* XXX could free buffered stream returned as result here. */
1707 set_buffered_stream (bash_input.location.buffered_fd, saver->bstream);
1708 }
1709 #endif /* BUFFERED_INPUT */
1710
1711 line_number = saver->line;
1712
1713 FREE (saver->bash_input.name);
1714 free (saver);
1715 }
1716 }
1717
1718 /* Return 1 if a stream of type TYPE is saved on the stack. */
1719 int
1720 stream_on_stack (type)
1721 enum stream_type type;
1722 {
1723 register STREAM_SAVER *s;
1724
1725 for (s = stream_list; s; s = s->next)
1726 if (s->bash_input.type == type)
1727 return 1;
1728 return 0;
1729 }
1730
1731 /* Save the current token state and return it in a malloced array. */
1732 int *
1733 save_token_state ()
1734 {
1735 int *ret;
1736
1737 ret = (int *)xmalloc (4 * sizeof (int));
1738 ret[0] = last_read_token;
1739 ret[1] = token_before_that;
1740 ret[2] = two_tokens_ago;
1741 ret[3] = current_token;
1742 return ret;
1743 }
1744
1745 void
1746 restore_token_state (ts)
1747 int *ts;
1748 {
1749 if (ts == 0)
1750 return;
1751 last_read_token = ts[0];
1752 token_before_that = ts[1];
1753 two_tokens_ago = ts[2];
1754 current_token = ts[3];
1755 }
1756
1757 /*
1758 * This is used to inhibit alias expansion and reserved word recognition
1759 * inside case statement pattern lists. A `case statement pattern list' is:
1760 *
1761 * everything between the `in' in a `case word in' and the next ')'
1762 * or `esac'
1763 * everything between a `;;' and the next `)' or `esac'
1764 */
1765
1766 #if defined (ALIAS) || defined (DPAREN_ARITHMETIC)
1767
1768 #define END_OF_ALIAS 0
1769
1770 /*
1771 * Pseudo-global variables used in implementing token-wise alias expansion.
1772 */
1773
1774 /*
1775 * Pushing and popping strings. This works together with shell_getc to
1776 * implement alias expansion on a per-token basis.
1777 */
1778
1779 typedef struct string_saver {
1780 struct string_saver *next;
1781 int expand_alias; /* Value to set expand_alias to when string is popped. */
1782 char *saved_line;
1783 #if defined (ALIAS)
1784 alias_t *expander; /* alias that caused this line to be pushed. */
1785 #endif
1786 int saved_line_size, saved_line_index, saved_line_terminator;
1787 } STRING_SAVER;
1788
1789 STRING_SAVER *pushed_string_list = (STRING_SAVER *)NULL;
1790
1791 /*
1792 * Push the current shell_input_line onto a stack of such lines and make S
1793 * the current input. Used when expanding aliases. EXPAND is used to set
1794 * the value of expand_next_token when the string is popped, so that the
1795 * word after the alias in the original line is handled correctly when the
1796 * alias expands to multiple words. TOKEN is the token that was expanded
1797 * into S; it is saved and used to prevent infinite recursive expansion.
1798 */
1799 static void
1800 push_string (s, expand, ap)
1801 char *s;
1802 int expand;
1803 alias_t *ap;
1804 {
1805 STRING_SAVER *temp = (STRING_SAVER *)xmalloc (sizeof (STRING_SAVER));
1806
1807 temp->expand_alias = expand;
1808 temp->saved_line = shell_input_line;
1809 temp->saved_line_size = shell_input_line_size;
1810 temp->saved_line_index = shell_input_line_index;
1811 temp->saved_line_terminator = shell_input_line_terminator;
1812 #if defined (ALIAS)
1813 temp->expander = ap;
1814 #endif
1815 temp->next = pushed_string_list;
1816 pushed_string_list = temp;
1817
1818 #if defined (ALIAS)
1819 if (ap)
1820 ap->flags |= AL_BEINGEXPANDED;
1821 #endif
1822
1823 shell_input_line = s;
1824 shell_input_line_size = strlen (s);
1825 shell_input_line_index = 0;
1826 shell_input_line_terminator = '\0';
1827 #if 0
1828 parser_state &= ~PST_ALEXPNEXT; /* XXX */
1829 #endif
1830
1831 set_line_mbstate ();
1832 }
1833
1834 /*
1835 * Make the top of the pushed_string stack be the current shell input.
1836 * Only called when there is something on the stack. Called from shell_getc
1837 * when it thinks it has consumed the string generated by an alias expansion
1838 * and needs to return to the original input line.
1839 */
1840 static void
1841 pop_string ()
1842 {
1843 STRING_SAVER *t;
1844
1845 FREE (shell_input_line);
1846 shell_input_line = pushed_string_list->saved_line;
1847 shell_input_line_index = pushed_string_list->saved_line_index;
1848 shell_input_line_size = pushed_string_list->saved_line_size;
1849 shell_input_line_terminator = pushed_string_list->saved_line_terminator;
1850
1851 if (pushed_string_list->expand_alias)
1852 parser_state |= PST_ALEXPNEXT;
1853 else
1854 parser_state &= ~PST_ALEXPNEXT;
1855
1856 t = pushed_string_list;
1857 pushed_string_list = pushed_string_list->next;
1858
1859 #if defined (ALIAS)
1860 if (t->expander)
1861 t->expander->flags &= ~AL_BEINGEXPANDED;
1862 #endif
1863
1864 free ((char *)t);
1865
1866 set_line_mbstate ();
1867 }
1868
1869 static void
1870 free_string_list ()
1871 {
1872 register STRING_SAVER *t, *t1;
1873
1874 for (t = pushed_string_list; t; )
1875 {
1876 t1 = t->next;
1877 FREE (t->saved_line);
1878 #if defined (ALIAS)
1879 if (t->expander)
1880 t->expander->flags &= ~AL_BEINGEXPANDED;
1881 #endif
1882 free ((char *)t);
1883 t = t1;
1884 }
1885 pushed_string_list = (STRING_SAVER *)NULL;
1886 }
1887
1888 #endif /* ALIAS || DPAREN_ARITHMETIC */
1889
1890 void
1891 free_pushed_string_input ()
1892 {
1893 #if defined (ALIAS) || defined (DPAREN_ARITHMETIC)
1894 free_string_list ();
1895 #endif
1896 }
1897
1898 /* Return a line of text, taken from wherever yylex () reads input.
1899 If there is no more input, then we return NULL. If REMOVE_QUOTED_NEWLINE
1900 is non-zero, we remove unquoted \<newline> pairs. This is used by
1901 read_secondary_line to read here documents. */
1902 static char *
1903 read_a_line (remove_quoted_newline)
1904 int remove_quoted_newline;
1905 {
1906 static char *line_buffer = (char *)NULL;
1907 static int buffer_size = 0;
1908 int indx, c, peekc, pass_next;
1909
1910 #if defined (READLINE)
1911 if (no_line_editing && SHOULD_PROMPT ())
1912 #else
1913 if (SHOULD_PROMPT ())
1914 #endif
1915 print_prompt ();
1916
1917 pass_next = indx = 0;
1918 while (1)
1919 {
1920 /* Allow immediate exit if interrupted during input. */
1921 QUIT;
1922
1923 c = yy_getc ();
1924
1925 /* Ignore null bytes in input. */
1926 if (c == 0)
1927 {
1928 #if 0
1929 internal_warning ("read_a_line: ignored null byte in input");
1930 #endif
1931 continue;
1932 }
1933
1934 /* If there is no more input, then we return NULL. */
1935 if (c == EOF)
1936 {
1937 if (interactive && bash_input.type == st_stream)
1938 clearerr (stdin);
1939 if (indx == 0)
1940 return ((char *)NULL);
1941 c = '\n';
1942 }
1943
1944 /* `+2' in case the final character in the buffer is a newline. */
1945 RESIZE_MALLOCED_BUFFER (line_buffer, indx, 2, buffer_size, 128);
1946
1947 /* IF REMOVE_QUOTED_NEWLINES is non-zero, we are reading a
1948 here document with an unquoted delimiter. In this case,
1949 the line will be expanded as if it were in double quotes.
1950 We allow a backslash to escape the next character, but we
1951 need to treat the backslash specially only if a backslash
1952 quoting a backslash-newline pair appears in the line. */
1953 if (pass_next)
1954 {
1955 line_buffer[indx++] = c;
1956 pass_next = 0;
1957 }
1958 else if (c == '\\' && remove_quoted_newline)
1959 {
1960 peekc = yy_getc ();
1961 if (peekc == '\n')
1962 {
1963 line_number++;
1964 continue; /* Make the unquoted \<newline> pair disappear. */
1965 }
1966 else
1967 {
1968 yy_ungetc (peekc);
1969 pass_next = 1;
1970 line_buffer[indx++] = c; /* Preserve the backslash. */
1971 }
1972 }
1973 else
1974 line_buffer[indx++] = c;
1975
1976 if (c == '\n')
1977 {
1978 line_buffer[indx] = '\0';
1979 return (line_buffer);
1980 }
1981 }
1982 }
1983
1984 /* Return a line as in read_a_line (), but insure that the prompt is
1985 the secondary prompt. This is used to read the lines of a here
1986 document. REMOVE_QUOTED_NEWLINE is non-zero if we should remove
1987 newlines quoted with backslashes while reading the line. It is
1988 non-zero unless the delimiter of the here document was quoted. */
1989 char *
1990 read_secondary_line (remove_quoted_newline)
1991 int remove_quoted_newline;
1992 {
1993 char *ret;
1994 int n, c;
1995
1996 prompt_string_pointer = &ps2_prompt;
1997 if (SHOULD_PROMPT())
1998 prompt_again ();
1999 ret = read_a_line (remove_quoted_newline);
2000 #if defined (HISTORY)
2001 if (ret && remember_on_history && (parser_state & PST_HEREDOC))
2002 {
2003 /* To make adding the the here-document body right, we need to rely
2004 on history_delimiting_chars() returning \n for the first line of
2005 the here-document body and the null string for the second and
2006 subsequent lines, so we avoid double newlines.
2007 current_command_line_count == 2 for the first line of the body. */
2008
2009 current_command_line_count++;
2010 maybe_add_history (ret);
2011 }
2012 #endif /* HISTORY */
2013 return ret;
2014 }
2015
2016 /* **************************************************************** */
2017 /* */
2018 /* YYLEX () */
2019 /* */
2020 /* **************************************************************** */
2021
2022 /* Reserved words. These are only recognized as the first word of a
2023 command. */
2024 STRING_INT_ALIST word_token_alist[] = {
2025 { "if", IF },
2026 { "then", THEN },
2027 { "else", ELSE },
2028 { "elif", ELIF },
2029 { "fi", FI },
2030 { "case", CASE },
2031 { "esac", ESAC },
2032 { "for", FOR },
2033 #if defined (SELECT_COMMAND)
2034 { "select", SELECT },
2035 #endif
2036 { "while", WHILE },
2037 { "until", UNTIL },
2038 { "do", DO },
2039 { "done", DONE },
2040 { "in", IN },
2041 { "function", FUNCTION },
2042 #if defined (COMMAND_TIMING)
2043 { "time", TIME },
2044 #endif
2045 { "{", '{' },
2046 { "}", '}' },
2047 { "!", BANG },
2048 #if defined (COND_COMMAND)
2049 { "[[", COND_START },
2050 { "]]", COND_END },
2051 #endif
2052 #if defined (COPROCESS_SUPPORT)
2053 { "coproc", COPROC },
2054 #endif
2055 { (char *)NULL, 0}
2056 };
2057
2058 /* other tokens that can be returned by read_token() */
2059 STRING_INT_ALIST other_token_alist[] = {
2060 /* Multiple-character tokens with special values */
2061 { "-p", TIMEOPT },
2062 { "&&", AND_AND },
2063 { "||", OR_OR },
2064 { ">>", GREATER_GREATER },
2065 { "<<", LESS_LESS },
2066 { "<&", LESS_AND },
2067 { ">&", GREATER_AND },
2068 { ";;", SEMI_SEMI },
2069 { ";&", SEMI_AND },
2070 { ";;&", SEMI_SEMI_AND },
2071 { "<<-", LESS_LESS_MINUS },
2072 { "<<<", LESS_LESS_LESS },
2073 { "&>", AND_GREATER },
2074 { "&>>", AND_GREATER_GREATER },
2075 { "<>", LESS_GREATER },
2076 { ">|", GREATER_BAR },
2077 { "|&", BAR_AND },
2078 { "EOF", yacc_EOF },
2079 /* Tokens whose value is the character itself */
2080 { ">", '>' },
2081 { "<", '<' },
2082 { "-", '-' },
2083 { "{", '{' },
2084 { "}", '}' },
2085 { ";", ';' },
2086 { "(", '(' },
2087 { ")", ')' },
2088 { "|", '|' },
2089 { "&", '&' },
2090 { "newline", '\n' },
2091 { (char *)NULL, 0}
2092 };
2093
2094 /* others not listed here:
2095 WORD look at yylval.word
2096 ASSIGNMENT_WORD look at yylval.word
2097 NUMBER look at yylval.number
2098 ARITH_CMD look at yylval.word_list
2099 ARITH_FOR_EXPRS look at yylval.word_list
2100 COND_CMD look at yylval.command
2101 */
2102
2103 /* These are used by read_token_word, but appear up here so that shell_getc
2104 can use them to decide when to add otherwise blank lines to the history. */
2105
2106 /* The primary delimiter stack. */
2107 struct dstack dstack = { (char *)NULL, 0, 0 };
2108
2109 /* A temporary delimiter stack to be used when decoding prompt strings.
2110 This is needed because command substitutions in prompt strings (e.g., PS2)
2111 can screw up the parser's quoting state. */
2112 static struct dstack temp_dstack = { (char *)NULL, 0, 0 };
2113
2114 /* Macro for accessing the top delimiter on the stack. Returns the
2115 delimiter or zero if none. */
2116 #define current_delimiter(ds) \
2117 (ds.delimiter_depth ? ds.delimiters[ds.delimiter_depth - 1] : 0)
2118
2119 #define push_delimiter(ds, character) \
2120 do \
2121 { \
2122 if (ds.delimiter_depth + 2 > ds.delimiter_space) \
2123 ds.delimiters = (char *)xrealloc \
2124 (ds.delimiters, (ds.delimiter_space += 10) * sizeof (char)); \
2125 ds.delimiters[ds.delimiter_depth] = character; \
2126 ds.delimiter_depth++; \
2127 } \
2128 while (0)
2129
2130 #define pop_delimiter(ds) ds.delimiter_depth--
2131
2132 /* Return the next shell input character. This always reads characters
2133 from shell_input_line; when that line is exhausted, it is time to
2134 read the next line. This is called by read_token when the shell is
2135 processing normal command input. */
2136
2137 /* This implements one-character lookahead/lookbehind across physical input
2138 lines, to avoid something being lost because it's pushed back with
2139 shell_ungetc when we're at the start of a line. */
2140 static int eol_ungetc_lookahead = 0;
2141
2142 static int
2143 shell_getc (remove_quoted_newline)
2144 int remove_quoted_newline;
2145 {
2146 register int i;
2147 int c;
2148 unsigned char uc;
2149
2150 QUIT;
2151
2152 if (sigwinch_received)
2153 {
2154 sigwinch_received = 0;
2155 get_new_window_size (0, (int *)0, (int *)0);
2156 }
2157
2158 if (eol_ungetc_lookahead)
2159 {
2160 c = eol_ungetc_lookahead;
2161 eol_ungetc_lookahead = 0;
2162 return (c);
2163 }
2164
2165 #if defined (ALIAS) || defined (DPAREN_ARITHMETIC)
2166 /* If shell_input_line[shell_input_line_index] == 0, but there is
2167 something on the pushed list of strings, then we don't want to go
2168 off and get another line. We let the code down below handle it. */
2169
2170 if (!shell_input_line || ((!shell_input_line[shell_input_line_index]) &&
2171 (pushed_string_list == (STRING_SAVER *)NULL)))
2172 #else /* !ALIAS && !DPAREN_ARITHMETIC */
2173 if (!shell_input_line || !shell_input_line[shell_input_line_index])
2174 #endif /* !ALIAS && !DPAREN_ARITHMETIC */
2175 {
2176 line_number++;
2177
2178 restart_read:
2179
2180 /* Allow immediate exit if interrupted during input. */
2181 QUIT;
2182
2183 i = 0;
2184 shell_input_line_terminator = 0;
2185
2186 /* If the shell is interatctive, but not currently printing a prompt
2187 (interactive_shell && interactive == 0), we don't want to print
2188 notifies or cleanup the jobs -- we want to defer it until we do
2189 print the next prompt. */
2190 if (interactive_shell == 0 || SHOULD_PROMPT())
2191 {
2192 #if defined (JOB_CONTROL)
2193 /* This can cause a problem when reading a command as the result
2194 of a trap, when the trap is called from flush_child. This call
2195 had better not cause jobs to disappear from the job table in
2196 that case, or we will have big trouble. */
2197 notify_and_cleanup ();
2198 #else /* !JOB_CONTROL */
2199 cleanup_dead_jobs ();
2200 #endif /* !JOB_CONTROL */
2201 }
2202
2203 #if defined (READLINE)
2204 if (no_line_editing && SHOULD_PROMPT())
2205 #else
2206 if (SHOULD_PROMPT())
2207 #endif
2208 print_prompt ();
2209
2210 if (bash_input.type == st_stream)
2211 clearerr (stdin);
2212
2213 while (1)
2214 {
2215 c = yy_getc ();
2216
2217 /* Allow immediate exit if interrupted during input. */
2218 QUIT;
2219
2220 if (c == '\0')
2221 {
2222 #if 0
2223 internal_warning ("shell_getc: ignored null byte in input");
2224 #endif
2225 continue;
2226 }
2227
2228 RESIZE_MALLOCED_BUFFER (shell_input_line, i, 2, shell_input_line_size, 256);
2229
2230 if (c == EOF)
2231 {
2232 if (bash_input.type == st_stream)
2233 clearerr (stdin);
2234
2235 if (i == 0)
2236 shell_input_line_terminator = EOF;
2237
2238 shell_input_line[i] = '\0';
2239 break;
2240 }
2241
2242 shell_input_line[i++] = c;
2243
2244 if (c == '\n')
2245 {
2246 shell_input_line[--i] = '\0';
2247 current_command_line_count++;
2248 break;
2249 }
2250 }
2251
2252 shell_input_line_index = 0;
2253 shell_input_line_len = i; /* == strlen (shell_input_line) */
2254
2255 set_line_mbstate ();
2256
2257 #if defined (HISTORY)
2258 if (remember_on_history && shell_input_line && shell_input_line[0])
2259 {
2260 char *expansions;
2261 # if defined (BANG_HISTORY)
2262 int old_hist;
2263
2264 /* If the current delimiter is a single quote, we should not be
2265 performing history expansion, even if we're on a different
2266 line from the original single quote. */
2267 old_hist = history_expansion_inhibited;
2268 if (current_delimiter (dstack) == '\'')
2269 history_expansion_inhibited = 1;
2270 # endif
2271 expansions = pre_process_line (shell_input_line, 1, 1);
2272 # if defined (BANG_HISTORY)
2273 history_expansion_inhibited = old_hist;
2274 # endif
2275 if (expansions != shell_input_line)
2276 {
2277 free (shell_input_line);
2278 shell_input_line = expansions;
2279 shell_input_line_len = shell_input_line ?
2280 strlen (shell_input_line) : 0;
2281 if (!shell_input_line_len)
2282 current_command_line_count--;
2283
2284 /* We have to force the xrealloc below because we don't know
2285 the true allocated size of shell_input_line anymore. */
2286 shell_input_line_size = shell_input_line_len;
2287
2288 set_line_mbstate ();
2289 }
2290 }
2291 /* Try to do something intelligent with blank lines encountered while
2292 entering multi-line commands. XXX - this is grotesque */
2293 else if (remember_on_history && shell_input_line &&
2294 shell_input_line[0] == '\0' &&
2295 current_command_line_count > 1)
2296 {
2297 if (current_delimiter (dstack))
2298 /* We know shell_input_line[0] == 0 and we're reading some sort of
2299 quoted string. This means we've got a line consisting of only
2300 a newline in a quoted string. We want to make sure this line
2301 gets added to the history. */
2302 maybe_add_history (shell_input_line);
2303 else
2304 {
2305 char *hdcs;
2306 hdcs = history_delimiting_chars ();
2307 if (hdcs && hdcs[0] == ';')
2308 maybe_add_history (shell_input_line);
2309 }
2310 }
2311
2312 #endif /* HISTORY */
2313
2314 if (shell_input_line)
2315 {
2316 /* Lines that signify the end of the shell's input should not be
2317 echoed. */
2318 if (echo_input_at_read && (shell_input_line[0] ||
2319 shell_input_line_terminator != EOF))
2320 fprintf (stderr, "%s\n", shell_input_line);
2321 }
2322 else
2323 {
2324 shell_input_line_size = 0;
2325 prompt_string_pointer = &current_prompt_string;
2326 if (SHOULD_PROMPT ())
2327 prompt_again ();
2328 goto restart_read;
2329 }
2330
2331 /* Add the newline to the end of this string, iff the string does
2332 not already end in an EOF character. */
2333 if (shell_input_line_terminator != EOF)
2334 {
2335 if (shell_input_line_len + 3 > shell_input_line_size)
2336 shell_input_line = (char *)xrealloc (shell_input_line,
2337 1 + (shell_input_line_size += 2));
2338
2339 shell_input_line[shell_input_line_len] = '\n';
2340 shell_input_line[shell_input_line_len + 1] = '\0';
2341
2342 set_line_mbstate ();
2343 }
2344 }
2345
2346 uc = shell_input_line[shell_input_line_index];
2347
2348 if (uc)
2349 shell_input_line_index++;
2350
2351 #if defined (ALIAS) || defined (DPAREN_ARITHMETIC)
2352 /* If UC is NULL, we have reached the end of the current input string. If
2353 pushed_string_list is non-empty, it's time to pop to the previous string
2354 because we have fully consumed the result of the last alias expansion.
2355 Do it transparently; just return the next character of the string popped
2356 to. */
2357 pop_alias:
2358 if (!uc && (pushed_string_list != (STRING_SAVER *)NULL))
2359 {
2360 pop_string ();
2361 uc = shell_input_line[shell_input_line_index];
2362 if (uc)
2363 shell_input_line_index++;
2364 }
2365 #endif /* ALIAS || DPAREN_ARITHMETIC */
2366
2367 if MBTEST(uc == '\\' && remove_quoted_newline && shell_input_line[shell_input_line_index] == '\n')
2368 {
2369 if (SHOULD_PROMPT ())
2370 prompt_again ();
2371 line_number++;
2372 /* XXX - what do we do here if we're expanding an alias whose definition
2373 ends with a newline? Recall that we inhibit the appending of a
2374 space in mk_alexpansion() if newline is the last character. */
2375 #if 0 /* XXX - bash-4.2 (jonathan@claggett.org) */
2376 if (expanding_alias () && shell_input_line[shell_input_line_index+1] == '\0')
2377 {
2378 uc = 0;
2379 goto pop_alias;
2380 }
2381 #endif
2382
2383 goto restart_read;
2384 }
2385
2386 if (!uc && shell_input_line_terminator == EOF)
2387 return ((shell_input_line_index != 0) ? '\n' : EOF);
2388
2389 return (uc);
2390 }
2391
2392 /* Put C back into the input for the shell. This might need changes for
2393 HANDLE_MULTIBYTE around EOLs. Since we (currently) never push back a
2394 character different than we read, shell_input_line_property doesn't need
2395 to change when manipulating shell_input_line. The define for
2396 last_shell_getc_is_singlebyte should take care of it, though. */
2397 static void
2398 shell_ungetc (c)
2399 int c;
2400 {
2401 if (shell_input_line && shell_input_line_index)
2402 shell_input_line[--shell_input_line_index] = c;
2403 else
2404 eol_ungetc_lookahead = c;
2405 }
2406
2407 #ifdef INCLUDE_UNUSED
2408 /* Back the input pointer up by one, effectively `ungetting' a character. */
2409 static void
2410 shell_ungetchar ()
2411 {
2412 if (shell_input_line && shell_input_line_index)
2413 shell_input_line_index--;
2414 }
2415 #endif
2416
2417 /* Discard input until CHARACTER is seen, then push that character back
2418 onto the input stream. */
2419 static void
2420 discard_until (character)
2421 int character;
2422 {
2423 int c;
2424
2425 while ((c = shell_getc (0)) != EOF && c != character)
2426 ;
2427
2428 if (c != EOF)
2429 shell_ungetc (c);
2430 }
2431
2432 void
2433 execute_variable_command (command, vname)
2434 char *command, *vname;
2435 {
2436 char *last_lastarg;
2437 sh_parser_state_t ps;
2438
2439 save_parser_state (&ps);
2440 last_lastarg = get_string_value ("_");
2441 if (last_lastarg)
2442 last_lastarg = savestring (last_lastarg);
2443
2444 parse_and_execute (savestring (command), vname, SEVAL_NONINT|SEVAL_NOHIST);
2445
2446 restore_parser_state (&ps);
2447 bind_variable ("_", last_lastarg, 0);
2448 FREE (last_lastarg);
2449
2450 if (token_to_read == '\n') /* reset_parser was called */
2451 token_to_read = 0;
2452 }
2453
2454 /* Place to remember the token. We try to keep the buffer
2455 at a reasonable size, but it can grow. */
2456 static char *token = (char *)NULL;
2457
2458 /* Current size of the token buffer. */
2459 static int token_buffer_size;
2460
2461 /* Command to read_token () explaining what we want it to do. */
2462 #define READ 0
2463 #define RESET 1
2464 #define prompt_is_ps1 \
2465 (!prompt_string_pointer || prompt_string_pointer == &ps1_prompt)
2466
2467 /* Function for yyparse to call. yylex keeps track of
2468 the last two tokens read, and calls read_token. */
2469 static int
2470 yylex ()
2471 {
2472 if (interactive && (current_token == 0 || current_token == '\n'))
2473 {
2474 /* Before we print a prompt, we might have to check mailboxes.
2475 We do this only if it is time to do so. Notice that only here
2476 is the mail alarm reset; nothing takes place in check_mail ()
2477 except the checking of mail. Please don't change this. */
2478 if (prompt_is_ps1 && time_to_check_mail ())
2479 {
2480 check_mail ();
2481 reset_mail_timer ();
2482 }
2483
2484 /* Avoid printing a prompt if we're not going to read anything, e.g.
2485 after resetting the parser with read_token (RESET). */
2486 if (token_to_read == 0 && SHOULD_PROMPT ())
2487 prompt_again ();
2488 }
2489
2490 two_tokens_ago = token_before_that;
2491 token_before_that = last_read_token;
2492 last_read_token = current_token;
2493 current_token = read_token (READ);
2494
2495 if ((parser_state & PST_EOFTOKEN) && current_token == shell_eof_token)
2496 {
2497 current_token = yacc_EOF;
2498 if (bash_input.type == st_string)
2499 rewind_input_string ();
2500 }
2501 parser_state &= ~PST_EOFTOKEN;
2502
2503 return (current_token);
2504 }
2505
2506 /* When non-zero, we have read the required tokens
2507 which allow ESAC to be the next one read. */
2508 static int esacs_needed_count;
2509
2510 void
2511 gather_here_documents ()
2512 {
2513 int r;
2514
2515 r = 0;
2516 while (need_here_doc)
2517 {
2518 parser_state |= PST_HEREDOC;
2519 make_here_document (redir_stack[r++], line_number);
2520 parser_state &= ~PST_HEREDOC;
2521 need_here_doc--;
2522 }
2523 }
2524
2525 /* When non-zero, an open-brace used to create a group is awaiting a close
2526 brace partner. */
2527 static int open_brace_count;
2528
2529 #define command_token_position(token) \
2530 (((token) == ASSIGNMENT_WORD) || (parser_state&PST_REDIRLIST) || \
2531 ((token) != SEMI_SEMI && (token) != SEMI_AND && (token) != SEMI_SEMI_AND && reserved_word_acceptable(token)))
2532
2533 #define assignment_acceptable(token) \
2534 (command_token_position(token) && ((parser_state & PST_CASEPAT) == 0))
2535
2536 /* Check to see if TOKEN is a reserved word and return the token
2537 value if it is. */
2538 #define CHECK_FOR_RESERVED_WORD(tok) \
2539 do { \
2540 if (!dollar_present && !quoted && \
2541 reserved_word_acceptable (last_read_token)) \
2542 { \
2543 int i; \
2544 for (i = 0; word_token_alist[i].word != (char *)NULL; i++) \
2545 if (STREQ (tok, word_token_alist[i].word)) \
2546 { \
2547 if ((parser_state & PST_CASEPAT) && (word_token_alist[i].token != ESAC)) \
2548 break; \
2549 if (word_token_alist[i].token == TIME && time_command_acceptable () == 0) \
2550 break; \
2551 if (word_token_alist[i].token == ESAC) \
2552 parser_state &= ~(PST_CASEPAT|PST_CASESTMT); \
2553 else if (word_token_alist[i].token == CASE) \
2554 parser_state |= PST_CASESTMT; \
2555 else if (word_token_alist[i].token == COND_END) \
2556 parser_state &= ~(PST_CONDCMD|PST_CONDEXPR); \
2557 else if (word_token_alist[i].token == COND_START) \
2558 parser_state |= PST_CONDCMD; \
2559 else if (word_token_alist[i].token == '{') \
2560 open_brace_count++; \
2561 else if (word_token_alist[i].token == '}' && open_brace_count) \
2562 open_brace_count--; \
2563 return (word_token_alist[i].token); \
2564 } \
2565 } \
2566 } while (0)
2567
2568 #if defined (ALIAS)
2569
2570 /* OK, we have a token. Let's try to alias expand it, if (and only if)
2571 it's eligible.
2572
2573 It is eligible for expansion if EXPAND_ALIASES is set, and
2574 the token is unquoted and the last token read was a command
2575 separator (or expand_next_token is set), and we are currently
2576 processing an alias (pushed_string_list is non-empty) and this
2577 token is not the same as the current or any previously
2578 processed alias.
2579
2580 Special cases that disqualify:
2581 In a pattern list in a case statement (parser_state & PST_CASEPAT). */
2582
2583 static char *
2584 mk_alexpansion (s)
2585 char *s;
2586 {
2587 int l;
2588 char *r;
2589
2590 l = strlen (s);
2591 r = xmalloc (l + 2);
2592 strcpy (r, s);
2593 #if 0 /* XXX - bash-4.2 */
2594 if (r[l -1] != ' ' && r[l -1] != '\n')
2595 #else
2596 if (r[l -1] != ' ')
2597 #endif
2598 r[l++] = ' ';
2599 r[l] = '\0';
2600 return r;
2601 }
2602
2603 static int
2604 alias_expand_token (tokstr)
2605 char *tokstr;
2606 {
2607 char *expanded;
2608 alias_t *ap;
2609
2610 if (((parser_state & PST_ALEXPNEXT) || command_token_position (last_read_token)) &&
2611 (parser_state & PST_CASEPAT) == 0)
2612 {
2613 ap = find_alias (tokstr);
2614
2615 /* Currently expanding this token. */
2616 if (ap && (ap->flags & AL_BEINGEXPANDED))
2617 return (NO_EXPANSION);
2618
2619 /* mk_alexpansion puts an extra space on the end of the alias expansion,
2620 so the lookahead by the parser works right. If this gets changed,
2621 make sure the code in shell_getc that deals with reaching the end of
2622 an expanded alias is changed with it. */
2623 expanded = ap ? mk_alexpansion (ap->value) : (char *)NULL;
2624
2625 if (expanded)
2626 {
2627 push_string (expanded, ap->flags & AL_EXPANDNEXT, ap);
2628 return (RE_READ_TOKEN);
2629 }
2630 else
2631 /* This is an eligible token that does not have an expansion. */
2632 return (NO_EXPANSION);
2633 }
2634 return (NO_EXPANSION);
2635 }
2636 #endif /* ALIAS */
2637
2638 static int
2639 time_command_acceptable ()
2640 {
2641 #if defined (COMMAND_TIMING)
2642 switch (last_read_token)
2643 {
2644 case 0:
2645 case ';':
2646 case '\n':
2647 case AND_AND:
2648 case OR_OR:
2649 case '&':
2650 case DO:
2651 case THEN:
2652 case ELSE:
2653 case '{': /* } */
2654 case '(': /* ) */
2655 return 1;
2656 default:
2657 return 0;
2658 }
2659 #else
2660 return 0;
2661 #endif /* COMMAND_TIMING */
2662 }
2663
2664 /* Handle special cases of token recognition:
2665 IN is recognized if the last token was WORD and the token
2666 before that was FOR or CASE or SELECT.
2667
2668 DO is recognized if the last token was WORD and the token
2669 before that was FOR or SELECT.
2670
2671 ESAC is recognized if the last token caused `esacs_needed_count'
2672 to be set
2673
2674 `{' is recognized if the last token as WORD and the token
2675 before that was FUNCTION, or if we just parsed an arithmetic
2676 `for' command.
2677
2678 `}' is recognized if there is an unclosed `{' present.
2679
2680 `-p' is returned as TIMEOPT if the last read token was TIME.
2681
2682 ']]' is returned as COND_END if the parser is currently parsing
2683 a conditional expression ((parser_state & PST_CONDEXPR) != 0)
2684
2685 `time' is returned as TIME if and only if it is immediately
2686 preceded by one of `;', `\n', `||', `&&', or `&'.
2687 */
2688
2689 static int
2690 special_case_tokens (tokstr)
2691 char *tokstr;
2692 {
2693 if ((last_read_token == WORD) &&
2694 #if defined (SELECT_COMMAND)
2695 ((token_before_that == FOR) || (token_before_that == CASE) || (token_before_that == SELECT)) &&
2696 #else
2697 ((token_before_that == FOR) || (token_before_that == CASE)) &&
2698 #endif
2699 (tokstr[0] == 'i' && tokstr[1] == 'n' && tokstr[2] == 0))
2700 {
2701 if (token_before_that == CASE)
2702 {
2703 parser_state |= PST_CASEPAT;
2704 esacs_needed_count++;
2705 }
2706 return (IN);
2707 }
2708
2709 if (last_read_token == WORD &&
2710 #if defined (SELECT_COMMAND)
2711 (token_before_that == FOR || token_before_that == SELECT) &&
2712 #else
2713 (token_before_that == FOR) &&
2714 #endif
2715 (tokstr[0] == 'd' && tokstr[1] == 'o' && tokstr[2] == '\0'))
2716 return (DO);
2717
2718 /* Ditto for ESAC in the CASE case.
2719 Specifically, this handles "case word in esac", which is a legal
2720 construct, certainly because someone will pass an empty arg to the
2721 case construct, and we don't want it to barf. Of course, we should
2722 insist that the case construct has at least one pattern in it, but
2723 the designers disagree. */
2724 if (esacs_needed_count)
2725 {
2726 esacs_needed_count--;
2727 if (STREQ (tokstr, "esac"))
2728 {
2729 parser_state &= ~PST_CASEPAT;
2730 return (ESAC);
2731 }
2732 }
2733
2734 /* The start of a shell function definition. */
2735 if (parser_state & PST_ALLOWOPNBRC)
2736 {
2737 parser_state &= ~PST_ALLOWOPNBRC;
2738 if (tokstr[0] == '{' && tokstr[1] == '\0') /* } */
2739 {
2740 open_brace_count++;
2741 function_bstart = line_number;
2742 return ('{'); /* } */
2743 }
2744 }
2745
2746 /* We allow a `do' after a for ((...)) without an intervening
2747 list_terminator */
2748 if (last_read_token == ARITH_FOR_EXPRS && tokstr[0] == 'd' && tokstr[1] == 'o' && !tokstr[2])
2749 return (DO);
2750 if (last_read_token == ARITH_FOR_EXPRS && tokstr[0] == '{' && tokstr[1] == '\0') /* } */
2751 {
2752 open_brace_count++;
2753 return ('{'); /* } */
2754 }
2755
2756 if (open_brace_count && reserved_word_acceptable (last_read_token) && tokstr[0] == '}' && !tokstr[1])
2757 {
2758 open_brace_count--; /* { */
2759 return ('}');
2760 }
2761
2762 #if defined (COMMAND_TIMING)
2763 /* Handle -p after `time'. */
2764 if (last_read_token == TIME && tokstr[0] == '-' && tokstr[1] == 'p' && !tokstr[2])
2765 return (TIMEOPT);
2766 #endif
2767
2768 #if 0
2769 #if defined (COMMAND_TIMING)
2770 if (STREQ (token, "time") && ((parser_state & PST_CASEPAT) == 0) && time_command_acceptable ())
2771 return (TIME);
2772 #endif /* COMMAND_TIMING */
2773 #endif
2774
2775 #if defined (COND_COMMAND) /* [[ */
2776 if ((parser_state & PST_CONDEXPR) && tokstr[0] == ']' && tokstr[1] == ']' && tokstr[2] == '\0')
2777 return (COND_END);
2778 #endif
2779
2780 return (-1);
2781 }
2782
2783 /* Called from shell.c when Control-C is typed at top level. Or
2784 by the error rule at top level. */
2785 void
2786 reset_parser ()
2787 {
2788 dstack.delimiter_depth = 0; /* No delimiters found so far. */
2789 open_brace_count = 0;
2790
2791 #if defined (EXTENDED_GLOB)
2792 /* Reset to global value of extended glob */
2793 if (parser_state & PST_EXTPAT)
2794 extended_glob = global_extglob;
2795 #endif
2796
2797 parser_state = 0;
2798
2799 #if defined (ALIAS) || defined (DPAREN_ARITHMETIC)
2800 if (pushed_string_list)
2801 free_string_list ();
2802 #endif /* ALIAS || DPAREN_ARITHMETIC */
2803
2804 if (shell_input_line)
2805 {
2806 free (shell_input_line);
2807 shell_input_line = (char *)NULL;
2808 shell_input_line_size = shell_input_line_index = 0;
2809 }
2810
2811 FREE (word_desc_to_read);
2812 word_desc_to_read = (WORD_DESC *)NULL;
2813
2814 current_token = '\n'; /* XXX */
2815 last_read_token = '\n';
2816 token_to_read = '\n';
2817 }
2818
2819 /* Read the next token. Command can be READ (normal operation) or
2820 RESET (to normalize state). */
2821 static int
2822 read_token (command)
2823 int command;
2824 {
2825 int character; /* Current character. */
2826 int peek_char; /* Temporary look-ahead character. */
2827 int result; /* The thing to return. */
2828
2829 if (command == RESET)
2830 {
2831 reset_parser ();
2832 return ('\n');
2833 }
2834
2835 if (token_to_read)
2836 {
2837 result = token_to_read;
2838 if (token_to_read == WORD || token_to_read == ASSIGNMENT_WORD)
2839 {
2840 yylval.word = word_desc_to_read;
2841 word_desc_to_read = (WORD_DESC *)NULL;
2842 }
2843 token_to_read = 0;
2844 return (result);
2845 }
2846
2847 #if defined (COND_COMMAND)
2848 if ((parser_state & (PST_CONDCMD|PST_CONDEXPR)) == PST_CONDCMD)
2849 {
2850 cond_lineno = line_number;
2851 parser_state |= PST_CONDEXPR;
2852 yylval.command = parse_cond_command ();
2853 if (cond_token != COND_END)
2854 {
2855 cond_error ();
2856 return (-1);
2857 }
2858 token_to_read = COND_END;
2859 parser_state &= ~(PST_CONDEXPR|PST_CONDCMD);
2860 return (COND_CMD);
2861 }
2862 #endif
2863
2864 #if defined (ALIAS)
2865 /* This is a place to jump back to once we have successfully expanded a
2866 token with an alias and pushed the string with push_string () */
2867 re_read_token:
2868 #endif /* ALIAS */
2869
2870 /* Read a single word from input. Start by skipping blanks. */
2871 while ((character = shell_getc (1)) != EOF && shellblank (character))
2872 ;
2873
2874 if (character == EOF)
2875 {
2876 EOF_Reached = 1;
2877 return (yacc_EOF);
2878 }
2879
2880 if MBTEST(character == '#' && (!interactive || interactive_comments))
2881 {
2882 /* A comment. Discard until EOL or EOF, and then return a newline. */
2883 discard_until ('\n');
2884 shell_getc (0);
2885 character = '\n'; /* this will take the next if statement and return. */
2886 }
2887
2888 if (character == '\n')
2889 {
2890 /* If we're about to return an unquoted newline, we can go and collect
2891 the text of any pending here document. */
2892 if (need_here_doc)
2893 gather_here_documents ();
2894
2895 #if defined (ALIAS)
2896 parser_state &= ~PST_ALEXPNEXT;
2897 #endif /* ALIAS */
2898
2899 parser_state &= ~PST_ASSIGNOK;
2900
2901 return (character);
2902 }
2903
2904 if (parser_state & PST_REGEXP)
2905 goto tokword;
2906
2907 /* Shell meta-characters. */
2908 if MBTEST(shellmeta (character) && ((parser_state & PST_DBLPAREN) == 0))
2909 {
2910 #if defined (ALIAS)
2911 /* Turn off alias tokenization iff this character sequence would
2912 not leave us ready to read a command. */
2913 if (character == '<' || character == '>')
2914 parser_state &= ~PST_ALEXPNEXT;
2915 #endif /* ALIAS */
2916
2917 parser_state &= ~PST_ASSIGNOK;
2918
2919 peek_char = shell_getc (1);
2920 if (character == peek_char)
2921 {
2922 switch (character)
2923 {
2924 case '<':
2925 /* If '<' then we could be at "<<" or at "<<-". We have to
2926 look ahead one more character. */
2927 peek_char = shell_getc (1);
2928 if MBTEST(peek_char == '-')
2929 return (LESS_LESS_MINUS);
2930 else if MBTEST(peek_char == '<')
2931 return (LESS_LESS_LESS);
2932 else
2933 {
2934 shell_ungetc (peek_char);
2935 return (LESS_LESS);
2936 }
2937
2938 case '>':
2939 return (GREATER_GREATER);
2940
2941 case ';':
2942 parser_state |= PST_CASEPAT;
2943 #if defined (ALIAS)
2944 parser_state &= ~PST_ALEXPNEXT;
2945 #endif /* ALIAS */
2946
2947 peek_char = shell_getc (1);
2948 if MBTEST(peek_char == '&')
2949 return (SEMI_SEMI_AND);
2950 else
2951 {
2952 shell_ungetc (peek_char);
2953 return (SEMI_SEMI);
2954 }
2955
2956 case '&':
2957 return (AND_AND);
2958
2959 case '|':
2960 return (OR_OR);
2961
2962 #if defined (DPAREN_ARITHMETIC) || defined (ARITH_FOR_COMMAND)
2963 case '(': /* ) */
2964 result = parse_dparen (character);
2965 if (result == -2)
2966 break;
2967 else
2968 return result;
2969 #endif
2970 }
2971 }
2972 else if MBTEST(character == '<' && peek_char == '&')
2973 return (LESS_AND);
2974 else if MBTEST(character == '>' && peek_char == '&')
2975 return (GREATER_AND);
2976 else if MBTEST(character == '<' && peek_char == '>')
2977 return (LESS_GREATER);
2978 else if MBTEST(character == '>' && peek_char == '|')
2979 return (GREATER_BAR);
2980 else if MBTEST(character == '&' && peek_char == '>')
2981 {
2982 peek_char = shell_getc (1);
2983 if MBTEST(peek_char == '>')
2984 return (AND_GREATER_GREATER);
2985 else
2986 {
2987 shell_ungetc (peek_char);
2988 return (AND_GREATER);
2989 }
2990 }
2991 else if MBTEST(character == '|' && peek_char == '&')
2992 return (BAR_AND);
2993 else if MBTEST(character == ';' && peek_char == '&')
2994 {
2995 parser_state |= PST_CASEPAT;
2996 #if defined (ALIAS)
2997 parser_state &= ~PST_ALEXPNEXT;
2998 #endif /* ALIAS */
2999 return (SEMI_AND);
3000 }
3001
3002 shell_ungetc (peek_char);
3003
3004 /* If we look like we are reading the start of a function
3005 definition, then let the reader know about it so that
3006 we will do the right thing with `{'. */
3007 if MBTEST(character == ')' && last_read_token == '(' && token_before_that == WORD)
3008 {
3009 parser_state |= PST_ALLOWOPNBRC;
3010 #if defined (ALIAS)
3011 parser_state &= ~PST_ALEXPNEXT;
3012 #endif /* ALIAS */
3013 function_dstart = line_number;
3014 }
3015
3016 /* case pattern lists may be preceded by an optional left paren. If
3017 we're not trying to parse a case pattern list, the left paren
3018 indicates a subshell. */
3019 if MBTEST(character == '(' && (parser_state & PST_CASEPAT) == 0) /* ) */
3020 parser_state |= PST_SUBSHELL;
3021 /*(*/
3022 else if MBTEST((parser_state & PST_CASEPAT) && character == ')')
3023 parser_state &= ~PST_CASEPAT;
3024 /*(*/
3025 else if MBTEST((parser_state & PST_SUBSHELL) && character == ')')
3026 parser_state &= ~PST_SUBSHELL;
3027
3028 #if defined (PROCESS_SUBSTITUTION)
3029 /* Check for the constructs which introduce process substitution.
3030 Shells running in `posix mode' don't do process substitution. */
3031 if MBTEST(posixly_correct || ((character != '>' && character != '<') || peek_char != '(')) /*)*/
3032 #endif /* PROCESS_SUBSTITUTION */
3033 return (character);
3034 }
3035
3036 /* Hack <&- (close stdin) case. Also <&N- (dup and close). */
3037 if MBTEST(character == '-' && (last_read_token == LESS_AND || last_read_token == GREATER_AND))
3038 return (character);
3039
3040 tokword:
3041 /* Okay, if we got this far, we have to read a word. Read one,
3042 and then check it against the known ones. */
3043 result = read_token_word (character);
3044 #if defined (ALIAS)
3045 if (result == RE_READ_TOKEN)
3046 goto re_read_token;
3047 #endif
3048 return result;
3049 }
3050
3051 /*
3052 * Match a $(...) or other grouping construct. This has to handle embedded
3053 * quoted strings ('', ``, "") and nested constructs. It also must handle
3054 * reprompting the user, if necessary, after reading a newline, and returning
3055 * correct error values if it reads EOF.
3056 */
3057 #define P_FIRSTCLOSE 0x01
3058 #define P_ALLOWESC 0x02
3059 #define P_DQUOTE 0x04
3060 #define P_COMMAND 0x08 /* parsing a command, so look for comments */
3061 #define P_BACKQUOTE 0x10 /* parsing a backquoted command substitution */
3062 #define P_ARRAYSUB 0x20 /* parsing a [...] array subscript for assignment */
3063
3064 /* Lexical state while parsing a grouping construct or $(...). */
3065 #define LEX_WASDOL 0x001
3066 #define LEX_CKCOMMENT 0x002
3067 #define LEX_INCOMMENT 0x004
3068 #define LEX_PASSNEXT 0x008
3069 #define LEX_RESWDOK 0x010
3070 #define LEX_CKCASE 0x020
3071 #define LEX_INCASE 0x040
3072 #define LEX_INHEREDOC 0x080
3073 #define LEX_HEREDELIM 0x100 /* reading here-doc delimiter */
3074 #define LEX_STRIPDOC 0x200 /* <<- strip tabs from here doc delim */
3075 #define LEX_INWORD 0x400
3076
3077 #define COMSUB_META(ch) ((ch) == ';' || (ch) == '&' || (ch) == '|')
3078
3079 #define CHECK_NESTRET_ERROR() \
3080 do { \
3081 if (nestret == &matched_pair_error) \
3082 { \
3083 free (ret); \
3084 return &matched_pair_error; \
3085 } \
3086 } while (0)
3087
3088 #define APPEND_NESTRET() \
3089 do { \
3090 if (nestlen) \
3091 { \
3092 RESIZE_MALLOCED_BUFFER (ret, retind, nestlen, retsize, 64); \
3093 strcpy (ret + retind, nestret); \
3094 retind += nestlen; \
3095 } \
3096 } while (0)
3097
3098 static char matched_pair_error;
3099
3100 static char *
3101 parse_matched_pair (qc, open, close, lenp, flags)
3102 int qc; /* `"' if this construct is within double quotes */
3103 int open, close;
3104 int *lenp, flags;
3105 {
3106 int count, ch, tflags;
3107 int nestlen, ttranslen, start_lineno;
3108 char *ret, *nestret, *ttrans;
3109 int retind, retsize, rflags;
3110
3111 /*itrace("parse_matched_pair[%d]: open = %c close = %c flags = %d", line_number, open, close, flags);*/
3112 count = 1;
3113 tflags = 0;
3114
3115 if ((flags & P_COMMAND) && qc != '`' && qc != '\'' && qc != '"' && (flags & P_DQUOTE) == 0)
3116 tflags |= LEX_CKCOMMENT;
3117
3118 /* RFLAGS is the set of flags we want to pass to recursive calls. */
3119 rflags = (qc == '"') ? P_DQUOTE : (flags & P_DQUOTE);
3120
3121 ret = (char *)xmalloc (retsize = 64);
3122 retind = 0;
3123
3124 start_lineno = line_number;
3125 while (count)
3126 {
3127 ch = shell_getc (qc != '\'' && (tflags & LEX_PASSNEXT) == 0);
3128
3129 if (ch == EOF)
3130 {
3131 free (ret);
3132 parser_error (start_lineno, _("unexpected EOF while looking for matching `%c'"), close);
3133 EOF_Reached = 1; /* XXX */
3134 return (&matched_pair_error);
3135 }
3136
3137 /* Possible reprompting. */
3138 if (ch == '\n' && SHOULD_PROMPT ())
3139 prompt_again ();
3140
3141 /* Don't bother counting parens or doing anything else if in a comment
3142 or part of a case statement */
3143 if (tflags & LEX_INCOMMENT)
3144 {
3145 /* Add this character. */
3146 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3147 ret[retind++] = ch;
3148
3149 if (ch == '\n')
3150 tflags &= ~LEX_INCOMMENT;
3151
3152 continue;
3153 }
3154
3155 /* Not exactly right yet, should handle shell metacharacters, too. If
3156 any changes are made to this test, make analogous changes to subst.c:
3157 extract_delimited_string(). */
3158 else if MBTEST((tflags & LEX_CKCOMMENT) && (tflags & LEX_INCOMMENT) == 0 && ch == '#' && (retind == 0 || ret[retind-1] == '\n' || shellblank (ret[retind - 1])))
3159 tflags |= LEX_INCOMMENT;
3160
3161 if (tflags & LEX_PASSNEXT) /* last char was backslash */
3162 {
3163 tflags &= ~LEX_PASSNEXT;
3164 if (qc != '\'' && ch == '\n') /* double-quoted \<newline> disappears. */
3165 {
3166 if (retind > 0)
3167 retind--; /* swallow previously-added backslash */
3168 continue;
3169 }
3170
3171 RESIZE_MALLOCED_BUFFER (ret, retind, 2, retsize, 64);
3172 if MBTEST(ch == CTLESC || ch == CTLNUL)
3173 ret[retind++] = CTLESC;
3174 ret[retind++] = ch;
3175 continue;
3176 }
3177 /* If we're reparsing the input (e.g., from parse_string_to_word_list),
3178 we've already prepended CTLESC to single-quoted results of $'...'.
3179 We may want to do this for other CTLESC-quoted characters in
3180 reparse, too. */
3181 else if MBTEST((parser_state & PST_REPARSE) && open == '\'' && (ch == CTLESC || ch == CTLNUL))
3182 {
3183 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3184 ret[retind++] = ch;
3185 continue;
3186 }
3187 else if MBTEST(ch == CTLESC || ch == CTLNUL) /* special shell escapes */
3188 {
3189 RESIZE_MALLOCED_BUFFER (ret, retind, 2, retsize, 64);
3190 ret[retind++] = CTLESC;
3191 ret[retind++] = ch;
3192 continue;
3193 }
3194 else if MBTEST(ch == close) /* ending delimiter */
3195 count--;
3196 /* handle nested ${...} specially. */
3197 else if MBTEST(open != close && (tflags & LEX_WASDOL) && open == '{' && ch == open) /* } */
3198 count++;
3199 else if MBTEST(((flags & P_FIRSTCLOSE) == 0) && ch == open) /* nested begin */
3200 count++;
3201
3202 /* Add this character. */
3203 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3204 ret[retind++] = ch;
3205
3206 /* If we just read the ending character, don't bother continuing. */
3207 if (count == 0)
3208 break;
3209
3210 if (open == '\'') /* '' inside grouping construct */
3211 {
3212 if MBTEST((flags & P_ALLOWESC) && ch == '\\')
3213 tflags |= LEX_PASSNEXT;
3214 continue;
3215 }
3216
3217 if MBTEST(ch == '\\') /* backslashes */
3218 tflags |= LEX_PASSNEXT;
3219
3220 #if 0
3221 /* The big hammer. Single quotes aren't special in double quotes. The
3222 problem is that Posix says the single quotes are semi-special:
3223 within a double-quoted ${...} construct "an even number of
3224 unescaped double-quotes or single-quotes, if any, shall occur." */
3225 if MBTEST(open == '{' && (flags & P_DQUOTE) && ch == '\'') /* } */
3226 continue;
3227 #endif
3228
3229 /* Could also check open == '`' if we want to parse grouping constructs
3230 inside old-style command substitution. */
3231 if (open != close) /* a grouping construct */
3232 {
3233 if MBTEST(shellquote (ch))
3234 {
3235 /* '', ``, or "" inside $(...) or other grouping construct. */
3236 push_delimiter (dstack, ch);
3237 if MBTEST((tflags & LEX_WASDOL) && ch == '\'') /* $'...' inside group */
3238 nestret = parse_matched_pair (ch, ch, ch, &nestlen, P_ALLOWESC|rflags);
3239 else
3240 nestret = parse_matched_pair (ch, ch, ch, &nestlen, rflags);
3241 pop_delimiter (dstack);
3242 CHECK_NESTRET_ERROR ();
3243
3244 if MBTEST((tflags & LEX_WASDOL) && ch == '\'' && (extended_quote || (rflags & P_DQUOTE) == 0))
3245 {
3246 /* Translate $'...' here. */
3247 ttrans = ansiexpand (nestret, 0, nestlen - 1, &ttranslen);
3248 xfree (nestret);
3249
3250 if ((rflags & P_DQUOTE) == 0)
3251 {
3252 nestret = sh_single_quote (ttrans);
3253 free (ttrans);
3254 nestlen = strlen (nestret);
3255 }
3256 else
3257 {
3258 nestret = ttrans;
3259 nestlen = ttranslen;
3260 }
3261 retind -= 2; /* back up before the $' */
3262 }
3263 else if MBTEST((tflags & LEX_WASDOL) && ch == '"' && (extended_quote || (rflags & P_DQUOTE) == 0))
3264 {
3265 /* Locale expand $"..." here. */
3266 ttrans = localeexpand (nestret, 0, nestlen - 1, start_lineno, &ttranslen);
3267 xfree (nestret);
3268
3269 nestret = sh_mkdoublequoted (ttrans, ttranslen, 0);
3270 free (ttrans);
3271 nestlen = ttranslen + 2;
3272 retind -= 2; /* back up before the $" */
3273 }
3274
3275 APPEND_NESTRET ();
3276 FREE (nestret);
3277 }
3278 else if ((flags & P_ARRAYSUB) && (tflags & LEX_WASDOL) && (ch == '(' || ch == '{' || ch == '[')) /* ) } ] */
3279 goto parse_dollar_word;
3280 }
3281 /* Parse an old-style command substitution within double quotes as a
3282 single word. */
3283 /* XXX - sh and ksh93 don't do this - XXX */
3284 else if MBTEST(open == '"' && ch == '`')
3285 {
3286 nestret = parse_matched_pair (0, '`', '`', &nestlen, rflags);
3287
3288 CHECK_NESTRET_ERROR ();
3289 APPEND_NESTRET ();
3290
3291 FREE (nestret);
3292 }
3293 else if MBTEST(open != '`' && (tflags & LEX_WASDOL) && (ch == '(' || ch == '{' || ch == '[')) /* ) } ] */
3294 /* check for $(), $[], or ${} inside quoted string. */
3295 {
3296 parse_dollar_word:
3297 if (open == ch) /* undo previous increment */
3298 count--;
3299 if (ch == '(') /* ) */
3300 nestret = parse_comsub (0, '(', ')', &nestlen, (rflags|P_COMMAND) & ~P_DQUOTE);
3301 else if (ch == '{') /* } */
3302 nestret = parse_matched_pair (0, '{', '}', &nestlen, P_FIRSTCLOSE|rflags);
3303 else if (ch == '[') /* ] */
3304 nestret = parse_matched_pair (0, '[', ']', &nestlen, rflags);
3305
3306 CHECK_NESTRET_ERROR ();
3307 APPEND_NESTRET ();
3308
3309 FREE (nestret);
3310 }
3311 if MBTEST(ch == '$')
3312 tflags |= LEX_WASDOL;
3313 else
3314 tflags &= ~LEX_WASDOL;
3315 }
3316
3317 ret[retind] = '\0';
3318 if (lenp)
3319 *lenp = retind;
3320 /*itrace("parse_matched_pair[%d]: returning %s", line_number, ret);*/
3321 return ret;
3322 }
3323
3324 /* Parse a $(...) command substitution. This is messier than I'd like, and
3325 reproduces a lot more of the token-reading code than I'd like. */
3326 static char *
3327 parse_comsub (qc, open, close, lenp, flags)
3328 int qc; /* `"' if this construct is within double quotes */
3329 int open, close;
3330 int *lenp, flags;
3331 {
3332 int count, ch, peekc, tflags, lex_rwlen, lex_wlen, lex_firstind;
3333 int nestlen, ttranslen, start_lineno;
3334 char *ret, *nestret, *ttrans, *heredelim;
3335 int retind, retsize, rflags, hdlen;
3336
3337 /*itrace("parse_comsub: qc = `%c' open = %c close = %c", qc, open, close);*/
3338 count = 1;
3339 tflags = LEX_RESWDOK;
3340
3341 if ((flags & P_COMMAND) && qc != '\'' && qc != '"' && (flags & P_DQUOTE) == 0)
3342 tflags |= LEX_CKCASE;
3343 if ((tflags & LEX_CKCASE) && (interactive == 0 || interactive_comments))
3344 tflags |= LEX_CKCOMMENT;
3345
3346 /* RFLAGS is the set of flags we want to pass to recursive calls. */
3347 rflags = (flags & P_DQUOTE);
3348
3349 ret = (char *)xmalloc (retsize = 64);
3350 retind = 0;
3351
3352 start_lineno = line_number;
3353 lex_rwlen = lex_wlen = 0;
3354
3355 heredelim = 0;
3356 lex_firstind = -1;
3357
3358 while (count)
3359 {
3360 comsub_readchar:
3361 ch = shell_getc (qc != '\'' && (tflags & LEX_PASSNEXT) == 0);
3362
3363 if (ch == EOF)
3364 {
3365 eof_error:
3366 free (ret);
3367 FREE (heredelim);
3368 parser_error (start_lineno, _("unexpected EOF while looking for matching `%c'"), close);
3369 EOF_Reached = 1; /* XXX */
3370 return (&matched_pair_error);
3371 }
3372
3373 /* If we hit the end of a line and are reading the contents of a here
3374 document, and it's not the same line that the document starts on,
3375 check for this line being the here doc delimiter. Otherwise, if
3376 we're in a here document, mark the next character as the beginning
3377 of a line. */
3378 if (ch == '\n')
3379 {
3380 if ((tflags & LEX_HEREDELIM) && heredelim)
3381 {
3382 tflags &= ~LEX_HEREDELIM;
3383 tflags |= LEX_INHEREDOC;
3384 lex_firstind = retind + 1;
3385 }
3386 else if (tflags & LEX_INHEREDOC)
3387 {
3388 int tind;
3389 tind = lex_firstind;
3390 while ((tflags & LEX_STRIPDOC) && ret[tind] == '\t')
3391 tind++;
3392 if (STREQN (ret + tind, heredelim, hdlen))
3393 {
3394 tflags &= ~(LEX_STRIPDOC|LEX_INHEREDOC);
3395 /*itrace("parse_comsub:%d: found here doc end `%s'", line_number, ret + tind);*/
3396 free (heredelim);
3397 heredelim = 0;
3398 lex_firstind = -1;
3399 }
3400 else
3401 lex_firstind = retind + 1;
3402 }
3403 }
3404
3405 /* Possible reprompting. */
3406 if (ch == '\n' && SHOULD_PROMPT ())
3407 prompt_again ();
3408
3409 /* XXX -- possibly allow here doc to be delimited by ending right
3410 paren. */
3411 if ((tflags & LEX_INHEREDOC) && ch == close && count == 1)
3412 {
3413 int tind;
3414 /*itrace("parse_comsub: in here doc, ch == close, retind - firstind = %d hdlen = %d retind = %d", retind-lex_firstind, hdlen, retind);*/
3415 tind = lex_firstind;
3416 while ((tflags & LEX_STRIPDOC) && ret[tind] == '\t')
3417 tind++;
3418 if (retind-tind == hdlen && STREQN (ret + tind, heredelim, hdlen))
3419 {
3420 tflags &= ~(LEX_STRIPDOC|LEX_INHEREDOC);
3421 /*itrace("parse_comsub:%d: found here doc end `%s'", line_number, ret + tind);*/
3422 free (heredelim);
3423 heredelim = 0;
3424 lex_firstind = -1;
3425 }
3426 }
3427
3428 /* Don't bother counting parens or doing anything else if in a comment */
3429 if (tflags & (LEX_INCOMMENT|LEX_INHEREDOC))
3430 {
3431 /* Add this character. */
3432 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3433 ret[retind++] = ch;
3434
3435 if ((tflags & LEX_INCOMMENT) && ch == '\n')
3436 {
3437 /*itrace("parse_comsub:%d: lex_incomment -> 0 ch = `%c'", line_number, ch);*/
3438 tflags &= ~LEX_INCOMMENT;
3439 }
3440
3441 continue;
3442 }
3443
3444 if (tflags & LEX_PASSNEXT) /* last char was backslash */
3445 {
3446 /*itrace("parse_comsub:%d: lex_passnext -> 0 ch = `%c' (%d)", line_number, ch, __LINE__);*/
3447 tflags &= ~LEX_PASSNEXT;
3448 if (qc != '\'' && ch == '\n') /* double-quoted \<newline> disappears. */
3449 {
3450 if (retind > 0)
3451 retind--; /* swallow previously-added backslash */
3452 continue;
3453 }
3454
3455 RESIZE_MALLOCED_BUFFER (ret, retind, 2, retsize, 64);
3456 if MBTEST(ch == CTLESC || ch == CTLNUL)
3457 ret[retind++] = CTLESC;
3458 ret[retind++] = ch;
3459 continue;
3460 }
3461
3462 /* If this is a shell break character, we are not in a word. If not,
3463 we either start or continue a word. */
3464 if MBTEST(shellbreak (ch))
3465 {
3466 tflags &= ~LEX_INWORD;
3467 /*itrace("parse_comsub:%d: lex_inword -> 0 ch = `%c' (%d)", line_number, ch, __LINE__);*/
3468 }
3469 else
3470 {
3471 if (tflags & LEX_INWORD)
3472 {
3473 lex_wlen++;
3474 /*itrace("parse_comsub:%d: lex_inword == 1 ch = `%c' lex_wlen = %d (%d)", line_number, ch, lex_wlen, __LINE__);*/
3475 }
3476 else
3477 {
3478 /*itrace("parse_comsub:%d: lex_inword -> 1 ch = `%c' (%d)", line_number, ch, __LINE__);*/
3479 tflags |= LEX_INWORD;
3480 lex_wlen = 0;
3481 }
3482 }
3483
3484 /* Skip whitespace */
3485 if MBTEST(shellblank (ch) && lex_rwlen == 0)
3486 {
3487 /* Add this character. */
3488 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3489 ret[retind++] = ch;
3490 continue;
3491 }
3492
3493 /* Either we are looking for the start of the here-doc delimiter
3494 (lex_firstind == -1) or we are reading one (lex_firstind >= 0).
3495 If this character is a shell break character and we are reading
3496 the delimiter, save it and note that we are now reading a here
3497 document. If we've found the start of the delimiter, note it by
3498 setting lex_firstind. Backslashes can quote shell metacharacters
3499 in here-doc delimiters. */
3500 if (tflags & LEX_HEREDELIM)
3501 {
3502 if (lex_firstind == -1 && shellbreak (ch) == 0)
3503 lex_firstind = retind;
3504 #if 0
3505 else if (heredelim && (tflags & LEX_PASSNEXT) == 0 && ch == '\n')
3506 {
3507 tflags |= LEX_INHEREDOC;
3508 tflags &= ~LEX_HEREDELIM;
3509 lex_firstind = retind + 1;
3510 }
3511 #endif
3512 else if (lex_firstind >= 0 && (tflags & LEX_PASSNEXT) == 0 && shellbreak (ch))
3513 {
3514 if (heredelim == 0)
3515 {
3516 nestret = substring (ret, lex_firstind, retind);
3517 heredelim = string_quote_removal (nestret, 0);
3518 free (nestret);
3519 hdlen = STRLEN(heredelim);
3520 /*itrace("parse_comsub:%d: found here doc delimiter `%s' (%d)", line_number, heredelim, hdlen);*/
3521 }
3522 if (ch == '\n')
3523 {
3524 tflags |= LEX_INHEREDOC;
3525 tflags &= ~LEX_HEREDELIM;
3526 lex_firstind = retind + 1;
3527 }
3528 else
3529 lex_firstind = -1;
3530 }
3531 }
3532
3533 /* Meta-characters that can introduce a reserved word. Not perfect yet. */
3534 if MBTEST((tflags & LEX_RESWDOK) == 0 && (tflags & LEX_CKCASE) && (tflags & LEX_INCOMMENT) == 0 && (shellmeta(ch) || ch == '\n'))
3535 {
3536 /* Add this character. */
3537 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3538 ret[retind++] = ch;
3539 peekc = shell_getc (1);
3540 if (ch == peekc && (ch == '&' || ch == '|' || ch == ';')) /* two-character tokens */
3541 {
3542 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3543 ret[retind++] = peekc;
3544 /*itrace("parse_comsub:%d: set lex_reswordok = 1, ch = `%c'", line_number, ch);*/
3545 tflags |= LEX_RESWDOK;
3546 lex_rwlen = 0;
3547 continue;
3548 }
3549 else if (ch == '\n' || COMSUB_META(ch))
3550 {
3551 shell_ungetc (peekc);
3552 /*itrace("parse_comsub:%d: set lex_reswordok = 1, ch = `%c'", line_number, ch);*/
3553 tflags |= LEX_RESWDOK;
3554 lex_rwlen = 0;
3555 continue;
3556 }
3557 else if (ch == EOF)
3558 goto eof_error;
3559 else
3560 {
3561 /* `unget' the character we just added and fall through */
3562 retind--;
3563 shell_ungetc (peekc);
3564 }
3565 }
3566
3567 /* If we can read a reserved word, try to read one. */
3568 if (tflags & LEX_RESWDOK)
3569 {
3570 if MBTEST(islower (ch))
3571 {
3572 /* Add this character. */
3573 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3574 ret[retind++] = ch;
3575 lex_rwlen++;
3576 continue;
3577 }
3578 else if MBTEST(lex_rwlen == 4 && shellbreak (ch))
3579 {
3580 if (STREQN (ret + retind - 4, "case", 4))
3581 {
3582 tflags |= LEX_INCASE;
3583 /*itrace("parse_comsub:%d: found `case', lex_incase -> 1 lex_reswdok -> 0", line_number);*/
3584 }
3585 else if (STREQN (ret + retind - 4, "esac", 4))
3586 {
3587 tflags &= ~LEX_INCASE;
3588 /*itrace("parse_comsub:%d: found `esac', lex_incase -> 0 lex_reswdok -> 0", line_number);*/
3589 }
3590 tflags &= ~LEX_RESWDOK;
3591 }
3592 else if MBTEST((tflags & LEX_CKCOMMENT) && ch == '#' && (lex_rwlen == 0 || ((tflags & LEX_INWORD) && lex_wlen == 0)))
3593 ; /* don't modify LEX_RESWDOK if we're starting a comment */
3594 else if MBTEST((tflags & LEX_INCASE) && ch != '\n')
3595 /* If we can read a reserved word and we're in case, we're at the
3596 point where we can read a new pattern list or an esac. We
3597 handle the esac case above. If we read a newline, we want to
3598 leave LEX_RESWDOK alone. If we read anything else, we want to
3599 turn off LEX_RESWDOK, since we're going to read a pattern list. */
3600 {
3601 tflags &= ~LEX_RESWDOK;
3602 /*itrace("parse_comsub:%d: lex_incase == 1 found `%c', lex_reswordok -> 0", line_number, ch);*/
3603 }
3604 else if MBTEST(shellbreak (ch) == 0)
3605 {
3606 tflags &= ~LEX_RESWDOK;
3607 /*itrace("parse_comsub:%d: found `%c', lex_reswordok -> 0", line_number, ch);*/
3608 }
3609 }
3610
3611 /* Might be the start of a here-doc delimiter */
3612 if MBTEST((tflags & LEX_INCOMMENT) == 0 && (tflags & LEX_CKCASE) && ch == '<')
3613 {
3614 /* Add this character. */
3615 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3616 ret[retind++] = ch;
3617 peekc = shell_getc (1);
3618 if (peekc == EOF)
3619 goto eof_error;
3620 if (peekc == ch)
3621 {
3622 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3623 ret[retind++] = peekc;
3624 peekc = shell_getc (1);
3625 if (peekc == EOF)
3626 goto eof_error;
3627 if (peekc == '-')
3628 {
3629 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3630 ret[retind++] = peekc;
3631 tflags |= LEX_STRIPDOC;
3632 }
3633 else
3634 shell_ungetc (peekc);
3635 if (peekc != '<')
3636 {
3637 tflags |= LEX_HEREDELIM;
3638 lex_firstind = -1;
3639 }
3640 continue;
3641 }
3642 else
3643 ch = peekc; /* fall through and continue XXX */
3644 }
3645 else if MBTEST((tflags & LEX_CKCOMMENT) && (tflags & LEX_INCOMMENT) == 0 && ch == '#' && (((tflags & LEX_RESWDOK) && lex_rwlen == 0) || ((tflags & LEX_INWORD) && lex_wlen == 0)))
3646 {
3647 /*itrace("parse_comsub:%d: lex_incomment -> 1 (%d)", line_number, __LINE__);*/
3648 tflags |= LEX_INCOMMENT;
3649 }
3650
3651 if MBTEST(ch == CTLESC || ch == CTLNUL) /* special shell escapes */
3652 {
3653 RESIZE_MALLOCED_BUFFER (ret, retind, 2, retsize, 64);
3654 ret[retind++] = CTLESC;
3655 ret[retind++] = ch;
3656 continue;
3657 }
3658 #if 0
3659 else if MBTEST((tflags & LEX_INCASE) && ch == close && close == ')')
3660 tflags &= ~LEX_INCASE; /* XXX */
3661 #endif
3662 else if MBTEST(ch == close && (tflags & LEX_INCASE) == 0) /* ending delimiter */
3663 {
3664 count--;
3665 /*itrace("parse_comsub:%d: found close: count = %d", line_number, count);*/
3666 }
3667 else if MBTEST(((flags & P_FIRSTCLOSE) == 0) && (tflags & LEX_INCASE) == 0 && ch == open) /* nested begin */
3668 {
3669 count++;
3670 /*itrace("parse_comsub:%d: found open: count = %d", line_number, count);*/
3671 }
3672
3673 /* Add this character. */
3674 RESIZE_MALLOCED_BUFFER (ret, retind, 1, retsize, 64);
3675 ret[retind++] = ch;
3676
3677 /* If we just read the ending character, don't bother continuing. */
3678 if (count == 0)
3679 break;
3680
3681 if MBTEST(ch == '\\') /* backslashes */
3682 tflags |= LEX_PASSNEXT;
3683
3684 if MBTEST(shellquote (ch))
3685 {
3686 /* '', ``, or "" inside $(...). */
3687 push_delimiter (dstack, ch);
3688 if MBTEST((tflags & LEX_WASDOL) && ch == '\'') /* $'...' inside group */
3689 nestret = parse_matched_pair (ch, ch, ch, &nestlen, P_ALLOWESC|rflags);
3690 else
3691 nestret = parse_matched_pair (ch, ch, ch, &nestlen, rflags);
3692 pop_delimiter (dstack);
3693 CHECK_NESTRET_ERROR ();
3694
3695 if MBTEST((tflags & LEX_WASDOL) && ch == '\'' && (extended_quote || (rflags & P_DQUOTE) == 0))
3696 {
3697 /* Translate $'...' here. */
3698 ttrans = ansiexpand (nestret, 0, nestlen - 1, &ttranslen);
3699 xfree (nestret);
3700
3701 if ((rflags & P_DQUOTE) == 0)
3702 {
3703 nestret = sh_single_quote (ttrans);
3704 free (ttrans);
3705 nestlen = strlen (nestret);
3706 }
3707 else
3708 {
3709 nestret = ttrans;
3710 nestlen = ttranslen;
3711 }
3712 retind -= 2; /* back up before the $' */
3713 }
3714 else if MBTEST((tflags & LEX_WASDOL) && ch == '"' && (extended_quote || (rflags & P_DQUOTE) == 0))
3715 {
3716 /* Locale expand $"..." here. */
3717 ttrans = localeexpand (nestret, 0, nestlen - 1, start_lineno, &ttranslen);
3718 xfree (nestret);
3719
3720 nestret = sh_mkdoublequoted (ttrans, ttranslen, 0);
3721 free (ttrans);
3722 nestlen = ttranslen + 2;
3723 retind -= 2; /* back up before the $" */
3724 }
3725
3726 APPEND_NESTRET ();
3727 FREE (nestret);
3728 }
3729 else if MBTEST((tflags & LEX_WASDOL) && (ch == '(' || ch == '{' || ch == '[')) /* ) } ] */
3730 /* check for $(), $[], or ${} inside command substitution. */
3731 {
3732 if ((tflags & LEX_INCASE) == 0 && open == ch) /* undo previous increment */
3733 count--;
3734 if (ch == '(') /* ) */
3735 nestret = parse_comsub (0, '(', ')', &nestlen, (rflags|P_COMMAND) & ~P_DQUOTE);
3736 else if (ch == '{') /* } */
3737 nestret = parse_matched_pair (0, '{', '}', &nestlen, P_FIRSTCLOSE|rflags);
3738 else if (ch == '[') /* ] */
3739 nestret = parse_matched_pair (0, '[', ']', &nestlen, rflags);
3740
3741 CHECK_NESTRET_ERROR ();
3742 APPEND_NESTRET ();
3743
3744 FREE (nestret);
3745 }
3746 if MBTEST(ch == '$')
3747 tflags |= LEX_WASDOL;
3748 else
3749 tflags &= ~LEX_WASDOL;
3750 }
3751
3752 FREE (heredelim);
3753 ret[retind] = '\0';
3754 if (lenp)
3755 *lenp = retind;
3756 /*itrace("parse_comsub:%d: returning `%s'", line_number, ret);*/
3757 return ret;
3758 }
3759
3760 /* XXX - this needs to handle functionality like subst.c:no_longjmp_on_fatal_error;
3761 maybe extract_command_subst should handle it. */
3762 char *
3763 xparse_dolparen (base, string, indp, flags)
3764 char *base;
3765 char *string;
3766 int *indp;
3767 int flags;
3768 {
3769 sh_parser_state_t ps;
3770 int orig_ind, nc, sflags;
3771 char *ret, *s, *ep, *ostring;
3772
3773 /*yydebug = 1;*/
3774 orig_ind = *indp;
3775 ostring = string;
3776
3777 sflags = SEVAL_NONINT|SEVAL_NOHIST|SEVAL_NOFREE;
3778 if (flags & SX_NOLONGJMP)
3779 sflags |= SEVAL_NOLONGJMP;
3780 save_parser_state (&ps);
3781
3782 /*(*/
3783 parser_state |= PST_CMDSUBST|PST_EOFTOKEN; /* allow instant ')' */ /*(*/
3784 shell_eof_token = ')';
3785 parse_string (string, "command substitution", sflags, &ep);
3786
3787 restore_parser_state (&ps);
3788 reset_parser ();
3789 if (interactive)
3790 token_to_read = 0;
3791
3792 /* Need to find how many characters parse_and_execute consumed, update
3793 *indp, if flags != 0, copy the portion of the string parsed into RET
3794 and return it. If flags & 1 (EX_NOALLOC) we can return NULL. */
3795
3796 /*(*/
3797 if (ep[-1] != ')')
3798 {
3799 #if DEBUG
3800 if (ep[-1] != '\n')
3801 itrace("xparse_dolparen:%d: ep[-1] != RPAREN (%d), ep = `%s'", line_number, ep[-1], ep);
3802 #endif
3803 while (ep > ostring && ep[-1] == '\n') ep--;
3804 }
3805
3806 nc = ep - ostring;
3807 *indp = ep - base - 1;
3808
3809 /*(*/
3810 #if DEBUG
3811 if (base[*indp] != ')')
3812 itrace("xparse_dolparen:%d: base[%d] != RPAREN (%d), base = `%s'", line_number, *indp, base[*indp], base);
3813 #endif
3814
3815 if (flags & SX_NOALLOC)
3816 return (char *)NULL;
3817
3818 if (nc == 0)
3819 {
3820 ret = xmalloc (1);
3821 ret[0] = '\0';
3822 }
3823 else
3824 ret = substring (ostring, 0, nc - 1);
3825
3826 return ret;
3827 }
3828
3829 #if defined (DPAREN_ARITHMETIC) || defined (ARITH_FOR_COMMAND)
3830 /* Parse a double-paren construct. It can be either an arithmetic
3831 command, an arithmetic `for' command, or a nested subshell. Returns
3832 the parsed token, -1 on error, or -2 if we didn't do anything and
3833 should just go on. */
3834 static int
3835 parse_dparen (c)
3836 int c;
3837 {
3838 int cmdtyp, sline;
3839 char *wval;
3840 WORD_DESC *wd;
3841
3842 #if defined (ARITH_FOR_COMMAND)
3843 if (last_read_token == FOR)
3844 {
3845 arith_for_lineno = line_number;
3846 cmdtyp = parse_arith_cmd (&wval, 0);
3847 if (cmdtyp == 1)
3848 {
3849 wd = alloc_word_desc ();
3850 wd->word = wval;
3851 yylval.word_list = make_word_list (wd, (WORD_LIST *)NULL);
3852 return (ARITH_FOR_EXPRS);
3853 }
3854 else
3855 return -1; /* ERROR */
3856 }
3857 #endif
3858
3859 #if defined (DPAREN_ARITHMETIC)
3860 if (reserved_word_acceptable (last_read_token))
3861 {
3862 sline = line_number;
3863
3864 cmdtyp = parse_arith_cmd (&wval, 0);
3865 if (cmdtyp == 1) /* arithmetic command */
3866 {
3867 wd = alloc_word_desc ();
3868 wd->word = wval;
3869 wd->flags = W_QUOTED|W_NOSPLIT|W_NOGLOB|W_DQUOTE;
3870 yylval.word_list = make_word_list (wd, (WORD_LIST *)NULL);
3871 return (ARITH_CMD);
3872 }
3873 else if (cmdtyp == 0) /* nested subshell */
3874 {
3875 push_string (wval, 0, (alias_t *)NULL);
3876 if ((parser_state & PST_CASEPAT) == 0)
3877 parser_state |= PST_SUBSHELL;
3878 return (c);
3879 }
3880 else /* ERROR */
3881 return -1;
3882 }
3883 #endif
3884
3885 return -2; /* XXX */
3886 }
3887
3888 /* We've seen a `(('. Look for the matching `))'. If we get it, return 1.
3889 If not, assume it's a nested subshell for backwards compatibility and
3890 return 0. In any case, put the characters we've consumed into a locally-
3891 allocated buffer and make *ep point to that buffer. Return -1 on an
3892 error, for example EOF. */
3893 static int
3894 parse_arith_cmd (ep, adddq)
3895 char **ep;
3896 int adddq;
3897 {
3898 int exp_lineno, rval, c;
3899 char *ttok, *tokstr;
3900 int ttoklen;
3901
3902 exp_lineno = line_number;
3903 ttok = parse_matched_pair (0, '(', ')', &ttoklen, 0);
3904 rval = 1;
3905 if (ttok == &matched_pair_error)
3906 return -1;
3907 /* Check that the next character is the closing right paren. If
3908 not, this is a syntax error. ( */
3909 c = shell_getc (0);
3910 if MBTEST(c != ')')
3911 rval = 0;
3912
3913 tokstr = (char *)xmalloc (ttoklen + 4);
3914
3915 /* if ADDDQ != 0 then (( ... )) -> "..." */
3916 if (rval == 1 && adddq) /* arith cmd, add double quotes */
3917 {
3918 tokstr[0] = '"';
3919 strncpy (tokstr + 1, ttok, ttoklen - 1);
3920 tokstr[ttoklen] = '"';
3921 tokstr[ttoklen+1] = '\0';
3922 }
3923 else if (rval == 1) /* arith cmd, don't add double quotes */
3924 {
3925 strncpy (tokstr, ttok, ttoklen - 1);
3926 tokstr[ttoklen-1] = '\0';
3927 }
3928 else /* nested subshell */
3929 {
3930 tokstr[0] = '(';
3931 strncpy (tokstr + 1, ttok, ttoklen - 1);
3932 tokstr[ttoklen] = ')';
3933 tokstr[ttoklen+1] = c;
3934 tokstr[ttoklen+2] = '\0';
3935 }
3936
3937 *ep = tokstr;
3938 FREE (ttok);
3939 return rval;
3940 }
3941 #endif /* DPAREN_ARITHMETIC || ARITH_FOR_COMMAND */
3942
3943 #if defined (COND_COMMAND)
3944 static void
3945 cond_error ()
3946 {
3947 char *etext;
3948
3949 if (EOF_Reached && cond_token != COND_ERROR) /* [[ */
3950 parser_error (cond_lineno, _("unexpected EOF while looking for `]]'"));
3951 else if (cond_token != COND_ERROR)
3952 {
3953 if (etext = error_token_from_token (cond_token))
3954 {
3955 parser_error (cond_lineno, _("syntax error in conditional expression: unexpected token `%s'"), etext);
3956 free (etext);
3957 }
3958 else
3959 parser_error (cond_lineno, _("syntax error in conditional expression"));
3960 }
3961 }
3962
3963 static COND_COM *
3964 cond_expr ()
3965 {
3966 return (cond_or ());
3967 }
3968
3969 static COND_COM *
3970 cond_or ()
3971 {
3972 COND_COM *l, *r;
3973
3974 l = cond_and ();
3975 if (cond_token == OR_OR)
3976 {
3977 r = cond_or ();
3978 l = make_cond_node (COND_OR, (WORD_DESC *)NULL, l, r);
3979 }
3980 return l;
3981 }
3982
3983 static COND_COM *
3984 cond_and ()
3985 {
3986 COND_COM *l, *r;
3987
3988 l = cond_term ();
3989 if (cond_token == AND_AND)
3990 {
3991 r = cond_and ();
3992 l = make_cond_node (COND_AND, (WORD_DESC *)NULL, l, r);
3993 }
3994 return l;
3995 }
3996
3997 static int
3998 cond_skip_newlines ()
3999 {
4000 while ((cond_token = read_token (READ)) == '\n')
4001 {
4002 if (SHOULD_PROMPT ())
4003 prompt_again ();
4004 }
4005 return (cond_token);
4006 }
4007
4008 #define COND_RETURN_ERROR() \
4009 do { cond_token = COND_ERROR; return ((COND_COM *)NULL); } while (0)
4010
4011 static COND_COM *
4012 cond_term ()
4013 {
4014 WORD_DESC *op;
4015 COND_COM *term, *tleft, *tright;
4016 int tok, lineno;
4017 char *etext;
4018
4019 /* Read a token. It can be a left paren, a `!', a unary operator, or a
4020 word that should be the first argument of a binary operator. Start by
4021 skipping newlines, since this is a compound command. */
4022 tok = cond_skip_newlines ();
4023 lineno = line_number;
4024 if (tok == COND_END)
4025 {
4026 COND_RETURN_ERROR ();
4027 }
4028 else if (tok == '(')
4029 {
4030 term = cond_expr ();
4031 if (cond_token != ')')
4032 {
4033 if (term)
4034 dispose_cond_node (term); /* ( */
4035 if (etext = error_token_from_token (cond_token))
4036 {
4037 parser_error (lineno, _("unexpected token `%s', expected `)'"), etext);
4038 free (etext);
4039 }
4040 else
4041 parser_error (lineno, _("expected `)'"));
4042 COND_RETURN_ERROR ();
4043 }
4044 term = make_cond_node (COND_EXPR, (WORD_DESC *)NULL, term, (COND_COM *)NULL);
4045 (void)cond_skip_newlines ();
4046 }
4047 else if (tok == BANG || (tok == WORD && (yylval.word->word[0] == '!' && yylval.word->word[1] == '\0')))
4048 {
4049 if (tok == WORD)
4050 dispose_word (yylval.word); /* not needed */
4051 term = cond_term ();
4052 if (term)
4053 term->flags |= CMD_INVERT_RETURN;
4054 }
4055 else if (tok == WORD && yylval.word->word[0] == '-' && yylval.word->word[2] == 0 && test_unop (yylval.word->word))
4056 {
4057 op = yylval.word;
4058 tok = read_token (READ);
4059 if (tok == WORD)
4060 {
4061 tleft = make_cond_node (COND_TERM, yylval.word, (COND_COM *)NULL, (COND_COM *)NULL);
4062 term = make_cond_node (COND_UNARY, op, tleft, (COND_COM *)NULL);
4063 }
4064 else
4065 {
4066 dispose_word (op);
4067 if (etext = error_token_from_token (tok))
4068 {
4069 parser_error (line_number, _("unexpected argument `%s' to conditional unary operator"), etext);
4070 free (etext);
4071 }
4072 else
4073 parser_error (line_number, _("unexpected argument to conditional unary operator"));
4074 COND_RETURN_ERROR ();
4075 }
4076
4077 (void)cond_skip_newlines ();
4078 }
4079 else if (tok == WORD) /* left argument to binary operator */
4080 {
4081 /* lhs */
4082 tleft = make_cond_node (COND_TERM, yylval.word, (COND_COM *)NULL, (COND_COM *)NULL);
4083
4084 /* binop */
4085 tok = read_token (READ);
4086 if (tok == WORD && test_binop (yylval.word->word))
4087 {
4088 op = yylval.word;
4089 if (op->word[0] == '=' && (op->word[1] == '\0' || (op->word[1] == '=' && op->word[2] == '\0')))
4090 parser_state |= PST_EXTPAT;
4091 else if (op->word[0] == '!' && op->word[1] == '=' && op->word[2] == '\0')
4092 parser_state |= PST_EXTPAT;
4093 }
4094 #if defined (COND_REGEXP)
4095 else if (tok == WORD && STREQ (yylval.word->word, "=~"))
4096 {
4097 op = yylval.word;
4098 parser_state |= PST_REGEXP;
4099 }
4100 #endif
4101 else if (tok == '<' || tok == '>')
4102 op = make_word_from_token (tok); /* ( */
4103 /* There should be a check before blindly accepting the `)' that we have
4104 seen the opening `('. */
4105 else if (tok == COND_END || tok == AND_AND || tok == OR_OR || tok == ')')
4106 {
4107 /* Special case. [[ x ]] is equivalent to [[ -n x ]], just like
4108 the test command. Similarly for [[ x && expr ]] or
4109 [[ x || expr ]] or [[ (x) ]]. */
4110 op = make_word ("-n");
4111 term = make_cond_node (COND_UNARY, op, tleft, (COND_COM *)NULL);
4112 cond_token = tok;
4113 return (term);
4114 }
4115 else
4116 {
4117 if (etext = error_token_from_token (tok))
4118 {
4119 parser_error (line_number, _("unexpected token `%s', conditional binary operator expected"), etext);
4120 free (etext);
4121 }
4122 else
4123 parser_error (line_number, _("conditional binary operator expected"));
4124 dispose_cond_node (tleft);
4125 COND_RETURN_ERROR ();
4126 }
4127
4128 /* rhs */
4129 if (parser_state & PST_EXTPAT)
4130 extended_glob = 1;
4131 tok = read_token (READ);
4132 if (parser_state & PST_EXTPAT)
4133 extended_glob = global_extglob;
4134 parser_state &= ~(PST_REGEXP|PST_EXTPAT);
4135
4136 if (tok == WORD)
4137 {
4138 tright = make_cond_node (COND_TERM, yylval.word, (COND_COM *)NULL, (COND_COM *)NULL);
4139 term = make_cond_node (COND_BINARY, op, tleft, tright);
4140 }
4141 else
4142 {
4143 if (etext = error_token_from_token (tok))
4144 {
4145 parser_error (line_number, _("unexpected argument `%s' to conditional binary operator"), etext);
4146 free (etext);
4147 }
4148 else
4149 parser_error (line_number, _("unexpected argument to conditional binary operator"));
4150 dispose_cond_node (tleft);
4151 dispose_word (op);
4152 COND_RETURN_ERROR ();
4153 }
4154
4155 (void)cond_skip_newlines ();
4156 }
4157 else
4158 {
4159 if (tok < 256)
4160 parser_error (line_number, _("unexpected token `%c' in conditional command"), tok);
4161 else if (etext = error_token_from_token (tok))
4162 {
4163 parser_error (line_number, _("unexpected token `%s' in conditional command"), etext);
4164 free (etext);
4165 }
4166 else
4167 parser_error (line_number, _("unexpected token %d in conditional command"), tok);
4168 COND_RETURN_ERROR ();
4169 }
4170 return (term);
4171 }
4172
4173 /* This is kind of bogus -- we slip a mini recursive-descent parser in
4174 here to handle the conditional statement syntax. */
4175 static COMMAND *
4176 parse_cond_command ()
4177 {
4178 COND_COM *cexp;
4179
4180 global_extglob = extended_glob;
4181 cexp = cond_expr ();
4182 return (make_cond_command (cexp));
4183 }
4184 #endif
4185
4186 #if defined (ARRAY_VARS)
4187 /* When this is called, it's guaranteed that we don't care about anything
4188 in t beyond i. We do save and restore the chars, though. */
4189 static int
4190 token_is_assignment (t, i)
4191 char *t;
4192 int i;
4193 {
4194 unsigned char c, c1;
4195 int r;
4196
4197 c = t[i]; c1 = t[i+1];
4198 t[i] = '='; t[i+1] = '\0';
4199 r = assignment (t, (parser_state & PST_COMPASSIGN) != 0);
4200 t[i] = c; t[i+1] = c1;
4201 return r;
4202 }
4203
4204 /* XXX - possible changes here for `+=' */
4205 static int
4206 token_is_ident (t, i)
4207 char *t;
4208 int i;
4209 {
4210 unsigned char c;
4211 int r;
4212
4213 c = t[i];
4214 t[i] = '\0';
4215 r = legal_identifier (t);
4216 t[i] = c;
4217 return r;
4218 }
4219 #endif
4220
4221 static int
4222 read_token_word (character)
4223 int character;
4224 {
4225 /* The value for YYLVAL when a WORD is read. */
4226 WORD_DESC *the_word;
4227
4228 /* Index into the token that we are building. */
4229 int token_index;
4230
4231 /* ALL_DIGITS becomes zero when we see a non-digit. */
4232 int all_digit_token;
4233
4234 /* DOLLAR_PRESENT becomes non-zero if we see a `$'. */
4235 int dollar_present;
4236
4237 /* COMPOUND_ASSIGNMENT becomes non-zero if we are parsing a compound
4238 assignment. */
4239 int compound_assignment;
4240
4241 /* QUOTED becomes non-zero if we see one of ("), ('), (`), or (\). */
4242 int quoted;
4243
4244 /* Non-zero means to ignore the value of the next character, and just
4245 to add it no matter what. */
4246 int pass_next_character;
4247
4248 /* The current delimiting character. */
4249 int cd;
4250 int result, peek_char;
4251 char *ttok, *ttrans;
4252 int ttoklen, ttranslen;
4253 intmax_t lvalue;
4254
4255 if (token_buffer_size < TOKEN_DEFAULT_INITIAL_SIZE)
4256 token = (char *)xrealloc (token, token_buffer_size = TOKEN_DEFAULT_INITIAL_SIZE);
4257
4258 token_index = 0;
4259 all_digit_token = DIGIT (character);
4260 dollar_present = quoted = pass_next_character = compound_assignment = 0;
4261
4262 for (;;)
4263 {
4264 if (character == EOF)
4265 goto got_token;
4266
4267 if (pass_next_character)
4268 {
4269 pass_next_character = 0;
4270 goto got_escaped_character;
4271 }
4272
4273 cd = current_delimiter (dstack);
4274
4275 /* Handle backslashes. Quote lots of things when not inside of
4276 double-quotes, quote some things inside of double-quotes. */
4277 if MBTEST(character == '\\')
4278 {
4279 peek_char = shell_getc (0);
4280
4281 /* Backslash-newline is ignored in all cases except
4282 when quoted with single quotes. */
4283 if (peek_char == '\n')
4284 {
4285 character = '\n';
4286 goto next_character;
4287 }
4288 else
4289 {
4290 shell_ungetc (peek_char);
4291
4292 /* If the next character is to be quoted, note it now. */
4293 if (cd == 0 || cd == '`' ||
4294 (cd == '"' && peek_char >= 0 && (sh_syntaxtab[peek_char] & CBSDQUOTE)))
4295 pass_next_character++;
4296
4297 quoted = 1;
4298 goto got_character;
4299 }
4300 }
4301
4302 /* Parse a matched pair of quote characters. */
4303 if MBTEST(shellquote (character))
4304 {
4305 push_delimiter (dstack, character);
4306 ttok = parse_matched_pair (character, character, character, &ttoklen, (character == '`') ? P_COMMAND : 0);
4307 pop_delimiter (dstack);
4308 if (ttok == &matched_pair_error)
4309 return -1; /* Bail immediately. */
4310 RESIZE_MALLOCED_BUFFER (token, token_index, ttoklen + 2,
4311 token_buffer_size, TOKEN_DEFAULT_GROW_SIZE);
4312 token[token_index++] = character;
4313 strcpy (token + token_index, ttok);
4314 token_index += ttoklen;
4315 all_digit_token = 0;
4316 quoted = 1;
4317 dollar_present |= (character == '"' && strchr (ttok, '$') != 0);
4318 FREE (ttok);
4319 goto next_character;
4320 }
4321
4322 #ifdef COND_REGEXP
4323 /* When parsing a regexp as a single word inside a conditional command,
4324 we need to special-case characters special to both the shell and
4325 regular expressions. Right now, that is only '(' and '|'. */ /*)*/
4326 if MBTEST((parser_state & PST_REGEXP) && (character == '(' || character == '|')) /*)*/
4327 {
4328 if (character == '|')
4329 goto got_character;
4330
4331 push_delimiter (dstack, character);
4332 ttok = parse_matched_pair (cd, '(', ')', &ttoklen, 0);
4333 pop_delimiter (dstack);
4334 if (ttok == &matched_pair_error)
4335 return -1; /* Bail immediately. */
4336 RESIZE_MALLOCED_BUFFER (token, token_index, ttoklen + 2,
4337 token_buffer_size, TOKEN_DEFAULT_GROW_SIZE);
4338 token[token_index++] = character;
4339 strcpy (token + token_index, ttok);
4340 token_index += ttoklen;
4341 FREE (ttok);
4342 dollar_present = all_digit_token = 0;
4343 goto next_character;
4344 }
4345 #endif /* COND_REGEXP */
4346
4347 #ifdef EXTENDED_GLOB
4348 /* Parse a ksh-style extended pattern matching specification. */
4349 if MBTEST(extended_glob && PATTERN_CHAR (character))
4350 {
4351 peek_char = shell_getc (1);
4352 if MBTEST(peek_char == '(') /* ) */
4353 {
4354 push_delimiter (dstack, peek_char);
4355 ttok = parse_matched_pair (cd, '(', ')', &ttoklen, 0);
4356 pop_delimiter (dstack);
4357 if (ttok == &matched_pair_error)
4358 return -1; /* Bail immediately. */
4359 RESIZE_MALLOCED_BUFFER (token, token_index, ttoklen + 2,
4360 token_buffer_size,
4361 TOKEN_DEFAULT_GROW_SIZE);
4362 token[token_index++] = character;
4363 token[token_index++] = peek_char;
4364 strcpy (token + token_index, ttok);
4365 token_index += ttoklen;
4366 FREE (ttok);
4367 dollar_present = all_digit_token = 0;
4368 goto next_character;
4369 }
4370 else
4371 shell_ungetc (peek_char);
4372 }
4373 #endif /* EXTENDED_GLOB */
4374
4375 /* If the delimiter character is not single quote, parse some of
4376 the shell expansions that must be read as a single word. */
4377 if (shellexp (character))
4378 {
4379 peek_char = shell_getc (1);
4380 /* $(...), <(...), >(...), $((...)), ${...}, and $[...] constructs */
4381 if MBTEST(peek_char == '(' || \
4382 ((peek_char == '{' || peek_char == '[') && character == '$')) /* ) ] } */
4383 {
4384 if (peek_char == '{') /* } */
4385 ttok = parse_matched_pair (cd, '{', '}', &ttoklen, P_FIRSTCLOSE);
4386 else if (peek_char == '(') /* ) */
4387 {
4388 /* XXX - push and pop the `(' as a delimiter for use by
4389 the command-oriented-history code. This way newlines
4390 appearing in the $(...) string get added to the
4391 history literally rather than causing a possibly-
4392 incorrect `;' to be added. ) */
4393 push_delimiter (dstack, peek_char);
4394 ttok = parse_comsub (cd, '(', ')', &ttoklen, P_COMMAND);
4395 pop_delimiter (dstack);
4396 }
4397 else
4398 ttok = parse_matched_pair (cd, '[', ']', &ttoklen, 0);
4399 if (ttok == &matched_pair_error)
4400 return -1; /* Bail immediately. */
4401 RESIZE_MALLOCED_BUFFER (token, token_index, ttoklen + 2,
4402 token_buffer_size,
4403 TOKEN_DEFAULT_GROW_SIZE);
4404 token[token_index++] = character;
4405 token[token_index++] = peek_char;
4406 strcpy (token + token_index, ttok);
4407 token_index += ttoklen;
4408 FREE (ttok);
4409 dollar_present = 1;
4410 all_digit_token = 0;
4411 goto next_character;
4412 }
4413 /* This handles $'...' and $"..." new-style quoted strings. */
4414 else if MBTEST(character == '$' && (peek_char == '\'' || peek_char == '"'))
4415 {
4416 int first_line;
4417
4418 first_line = line_number;
4419 push_delimiter (dstack, peek_char);
4420 ttok = parse_matched_pair (peek_char, peek_char, peek_char,
4421 &ttoklen,
4422 (peek_char == '\'') ? P_ALLOWESC : 0);
4423 pop_delimiter (dstack);
4424 if (ttok == &matched_pair_error)
4425 return -1;
4426 if (peek_char == '\'')
4427 {
4428 ttrans = ansiexpand (ttok, 0, ttoklen - 1, &ttranslen);
4429 free (ttok);
4430
4431 /* Insert the single quotes and correctly quote any
4432 embedded single quotes (allowed because P_ALLOWESC was
4433 passed to parse_matched_pair). */
4434 ttok = sh_single_quote (ttrans);
4435 free (ttrans);
4436 ttranslen = strlen (ttok);
4437 ttrans = ttok;
4438 }
4439 else
4440 {
4441 /* Try to locale-expand the converted string. */
4442 ttrans = localeexpand (ttok, 0, ttoklen - 1, first_line, &ttranslen);
4443 free (ttok);
4444
4445 /* Add the double quotes back */
4446 ttok = sh_mkdoublequoted (ttrans, ttranslen, 0);
4447 free (ttrans);
4448 ttranslen += 2;
4449 ttrans = ttok;
4450 }
4451
4452 RESIZE_MALLOCED_BUFFER (token, token_index, ttranslen + 2,
4453 token_buffer_size,
4454 TOKEN_DEFAULT_GROW_SIZE);
4455 strcpy (token + token_index, ttrans);
4456 token_index += ttranslen;
4457 FREE (ttrans);
4458 quoted = 1;
4459 all_digit_token = 0;
4460 goto next_character;
4461 }
4462 /* This could eventually be extended to recognize all of the
4463 shell's single-character parameter expansions, and set flags.*/
4464 else if MBTEST(character == '$' && peek_char == '$')
4465 {
4466 ttok = (char *)xmalloc (3);
4467 ttok[0] = ttok[1] = '$';
4468 ttok[2] = '\0';
4469 RESIZE_MALLOCED_BUFFER (token, token_index, 3,
4470 token_buffer_size,
4471 TOKEN_DEFAULT_GROW_SIZE);
4472 strcpy (token + token_index, ttok);
4473 token_index += 2;
4474 dollar_present = 1;
4475 all_digit_token = 0;
4476 FREE (ttok);
4477 goto next_character;
4478 }
4479 else
4480 shell_ungetc (peek_char);
4481 }
4482
4483 #if defined (ARRAY_VARS)
4484 /* Identify possible array subscript assignment; match [...]. If
4485 parser_state&PST_COMPASSIGN, we need to parse [sub]=words treating
4486 `sub' as if it were enclosed in double quotes. */
4487 else if MBTEST(character == '[' && /* ] */
4488 ((token_index > 0 && assignment_acceptable (last_read_token) && token_is_ident (token, token_index)) ||
4489 (token_index == 0 && (parser_state&PST_COMPASSIGN))))
4490 {
4491 ttok = parse_matched_pair (cd, '[', ']', &ttoklen, P_ARRAYSUB);
4492 if (ttok == &matched_pair_error)
4493 return -1; /* Bail immediately. */
4494 RESIZE_MALLOCED_BUFFER (token, token_index, ttoklen + 2,
4495 token_buffer_size,
4496 TOKEN_DEFAULT_GROW_SIZE);
4497 token[token_index++] = character;
4498 strcpy (token + token_index, ttok);
4499 token_index += ttoklen;
4500 FREE (ttok);
4501 all_digit_token = 0;
4502 goto next_character;
4503 }
4504 /* Identify possible compound array variable assignment. */
4505 else if MBTEST(character == '=' && token_index > 0 && (assignment_acceptable (last_read_token) || (parser_state & PST_ASSIGNOK)) && token_is_assignment (token, token_index))
4506 {
4507 peek_char = shell_getc (1);
4508 if MBTEST(peek_char == '(') /* ) */
4509 {
4510 ttok = parse_compound_assignment (&ttoklen);
4511
4512 RESIZE_MALLOCED_BUFFER (token, token_index, ttoklen + 4,
4513 token_buffer_size,
4514 TOKEN_DEFAULT_GROW_SIZE);
4515
4516 token[token_index++] = '=';
4517 token[token_index++] = '(';
4518 if (ttok)
4519 {
4520 strcpy (token + token_index, ttok);
4521 token_index += ttoklen;
4522 }
4523 token[token_index++] = ')';
4524 FREE (ttok);
4525 all_digit_token = 0;
4526 compound_assignment = 1;
4527 #if 1
4528 goto next_character;
4529 #else
4530 goto got_token; /* ksh93 seems to do this */
4531 #endif
4532 }
4533 else
4534 shell_ungetc (peek_char);
4535 }
4536 #endif
4537
4538 /* When not parsing a multi-character word construct, shell meta-
4539 characters break words. */
4540 if MBTEST(shellbreak (character))
4541 {
4542 shell_ungetc (character);
4543 goto got_token;
4544 }
4545
4546 got_character:
4547
4548 if (character == CTLESC || character == CTLNUL)
4549 token[token_index++] = CTLESC;
4550
4551 got_escaped_character:
4552
4553 all_digit_token &= DIGIT (character);
4554 dollar_present |= character == '$';
4555
4556 token[token_index++] = character;
4557
4558 RESIZE_MALLOCED_BUFFER (token, token_index, 1, token_buffer_size,
4559 TOKEN_DEFAULT_GROW_SIZE);
4560
4561 next_character:
4562 if (character == '\n' && SHOULD_PROMPT ())
4563 prompt_again ();
4564
4565 /* We want to remove quoted newlines (that is, a \<newline> pair)
4566 unless we are within single quotes or pass_next_character is
4567 set (the shell equivalent of literal-next). */
4568 cd = current_delimiter (dstack);
4569 character = shell_getc (cd != '\'' && pass_next_character == 0);
4570 } /* end for (;;) */
4571
4572 got_token:
4573
4574 token[token_index] = '\0';
4575
4576 /* Check to see what thing we should return. If the last_read_token
4577 is a `<', or a `&', or the character which ended this token is
4578 a '>' or '<', then, and ONLY then, is this input token a NUMBER.
4579 Otherwise, it is just a word, and should be returned as such. */
4580 if MBTEST(all_digit_token && (character == '<' || character == '>' || \
4581 last_read_token == LESS_AND || \
4582 last_read_token == GREATER_AND))
4583 {
4584 if (legal_number (token, &lvalue) && (int)lvalue == lvalue)
4585 yylval.number = lvalue;
4586 else
4587 yylval.number = -1;
4588 return (NUMBER);
4589 }
4590
4591 /* Check for special case tokens. */
4592 result = (last_shell_getc_is_singlebyte) ? special_case_tokens (token) : -1;
4593 if (result >= 0)
4594 return result;
4595
4596 #if defined (ALIAS)
4597 /* Posix.2 does not allow reserved words to be aliased, so check for all
4598 of them, including special cases, before expanding the current token
4599 as an alias. */
4600 if MBTEST(posixly_correct)
4601 CHECK_FOR_RESERVED_WORD (token);
4602
4603 /* Aliases are expanded iff EXPAND_ALIASES is non-zero, and quoting
4604 inhibits alias expansion. */
4605 if (expand_aliases && quoted == 0)
4606 {
4607 result = alias_expand_token (token);
4608 if (result == RE_READ_TOKEN)
4609 return (RE_READ_TOKEN);
4610 else if (result == NO_EXPANSION)
4611 parser_state &= ~PST_ALEXPNEXT;
4612 }
4613
4614 /* If not in Posix.2 mode, check for reserved words after alias
4615 expansion. */
4616 if MBTEST(posixly_correct == 0)
4617 #endif
4618 CHECK_FOR_RESERVED_WORD (token);
4619
4620 the_word = (WORD_DESC *)xmalloc (sizeof (WORD_DESC));
4621 the_word->word = (char *)xmalloc (1 + token_index);
4622 the_word->flags = 0;
4623 strcpy (the_word->word, token);
4624 if (dollar_present)
4625 the_word->flags |= W_HASDOLLAR;
4626 if (quoted)
4627 the_word->flags |= W_QUOTED; /*(*/
4628 if (compound_assignment && token[token_index-1] == ')')
4629 the_word->flags |= W_COMPASSIGN;
4630 /* A word is an assignment if it appears at the beginning of a
4631 simple command, or after another assignment word. This is
4632 context-dependent, so it cannot be handled in the grammar. */
4633 if (assignment (token, (parser_state & PST_COMPASSIGN) != 0))
4634 {
4635 the_word->flags |= W_ASSIGNMENT;
4636 /* Don't perform word splitting on assignment statements. */
4637 if (assignment_acceptable (last_read_token) || (parser_state & PST_COMPASSIGN) != 0)
4638 the_word->flags |= W_NOSPLIT;
4639 }
4640
4641 if (command_token_position (last_read_token))
4642 {
4643 struct builtin *b;
4644 b = builtin_address_internal (token, 0);
4645 if (b && (b->flags & ASSIGNMENT_BUILTIN))
4646 parser_state |= PST_ASSIGNOK;
4647 else if (STREQ (token, "eval") || STREQ (token, "let"))
4648 parser_state |= PST_ASSIGNOK;
4649 }
4650
4651 yylval.word = the_word;
4652
4653 if (token[0] == '{' && token[token_index-1] == '}' &&
4654 (character == '<' || character == '>'))
4655 {
4656 /* can use token; already copied to the_word */
4657 token[token_index-1] = '\0';
4658 if (legal_identifier (token+1))
4659 {
4660 strcpy (the_word->word, token+1);
4661 /*itrace("read_token_word: returning REDIR_WORD for %s", the_word->word);*/
4662 return (REDIR_WORD);
4663 }
4664 }
4665
4666 result = ((the_word->flags & (W_ASSIGNMENT|W_NOSPLIT)) == (W_ASSIGNMENT|W_NOSPLIT))
4667 ? ASSIGNMENT_WORD : WORD;
4668
4669 switch (last_read_token)
4670 {
4671 case FUNCTION:
4672 parser_state |= PST_ALLOWOPNBRC;
4673 function_dstart = line_number;
4674 break;
4675 case CASE:
4676 case SELECT:
4677 case FOR:
4678 if (word_top < MAX_CASE_NEST)
4679 word_top++;
4680 word_lineno[word_top] = line_number;
4681 break;
4682 }
4683
4684 return (result);
4685 }
4686
4687 /* Return 1 if TOKSYM is a token that after being read would allow
4688 a reserved word to be seen, else 0. */
4689 static int
4690 reserved_word_acceptable (toksym)
4691 int toksym;
4692 {
4693 switch (toksym)
4694 {
4695 case '\n':
4696 case ';':
4697 case '(':
4698 case ')':
4699 case '|':
4700 case '&':
4701 case '{':
4702 case '}': /* XXX */
4703 case AND_AND:
4704 case BANG:
4705 case BAR_AND:
4706 case DO:
4707 case DONE:
4708 case ELIF:
4709 case ELSE:
4710 case ESAC:
4711 case FI:
4712 case IF:
4713 case OR_OR:
4714 case SEMI_SEMI:
4715 case SEMI_AND:
4716 case SEMI_SEMI_AND:
4717 case THEN:
4718 case TIME:
4719 case TIMEOPT:
4720 case COPROC:
4721 case UNTIL:
4722 case WHILE:
4723 case 0:
4724 return 1;
4725 default:
4726 #if defined (COPROCESS_SUPPORT)
4727 if (last_read_token == WORD && token_before_that == COPROC)
4728 return 1;
4729 #endif
4730 return 0;
4731 }
4732 }
4733
4734 /* Return the index of TOKEN in the alist of reserved words, or -1 if
4735 TOKEN is not a shell reserved word. */
4736 int
4737 find_reserved_word (tokstr)
4738 char *tokstr;
4739 {
4740 int i;
4741 for (i = 0; word_token_alist[i].word; i++)
4742 if (STREQ (tokstr, word_token_alist[i].word))
4743 return i;
4744 return -1;
4745 }
4746
4747 #if 0
4748 #if defined (READLINE)
4749 /* Called after each time readline is called. This insures that whatever
4750 the new prompt string is gets propagated to readline's local prompt
4751 variable. */
4752 static void
4753 reset_readline_prompt ()
4754 {
4755 char *temp_prompt;
4756
4757 if (prompt_string_pointer)
4758 {
4759 temp_prompt = (*prompt_string_pointer)
4760 ? decode_prompt_string (*prompt_string_pointer)
4761 : (char *)NULL;
4762
4763 if (temp_prompt == 0)
4764 {
4765 temp_prompt = (char *)xmalloc (1);
4766 temp_prompt[0] = '\0';
4767 }
4768
4769 FREE (current_readline_prompt);
4770 current_readline_prompt = temp_prompt;
4771 }
4772 }
4773 #endif /* READLINE */
4774 #endif /* 0 */
4775
4776 #if defined (HISTORY)
4777 /* A list of tokens which can be followed by newlines, but not by
4778 semi-colons. When concatenating multiple lines of history, the
4779 newline separator for such tokens is replaced with a space. */
4780 static const int no_semi_successors[] = {
4781 '\n', '{', '(', ')', ';', '&', '|',
4782 CASE, DO, ELSE, IF, SEMI_SEMI, SEMI_AND, SEMI_SEMI_AND, THEN, UNTIL,
4783 WHILE, AND_AND, OR_OR, IN,
4784 0
4785 };
4786
4787 /* If we are not within a delimited expression, try to be smart
4788 about which separators can be semi-colons and which must be
4789 newlines. Returns the string that should be added into the
4790 history entry. */
4791 char *
4792 history_delimiting_chars ()
4793 {
4794 register int i;
4795
4796 if (dstack.delimiter_depth != 0)
4797 return ("\n");
4798
4799 /* We look for current_command_line_count == 2 because we are looking to
4800 add the first line of the body of the here document (the second line
4801 of the command). */
4802 if (parser_state & PST_HEREDOC)
4803 return (current_command_line_count == 2 ? "\n" : "");
4804
4805 /* First, handle some special cases. */
4806 /*(*/
4807 /* If we just read `()', assume it's a function definition, and don't
4808 add a semicolon. If the token before the `)' was not `(', and we're
4809 not in the midst of parsing a case statement, assume it's a
4810 parenthesized command and add the semicolon. */
4811 /*)(*/
4812 if (token_before_that == ')')
4813 {
4814 if (two_tokens_ago == '(') /*)*/ /* function def */
4815 return " ";
4816 /* This does not work for subshells inside case statement
4817 command lists. It's a suboptimal solution. */
4818 else if (parser_state & PST_CASESTMT) /* case statement pattern */
4819 return " ";
4820 else
4821 return "; "; /* (...) subshell */
4822 }
4823 else if (token_before_that == WORD && two_tokens_ago == FUNCTION)
4824 return " "; /* function def using `function name' without `()' */
4825
4826 else if (token_before_that == WORD && two_tokens_ago == FOR)
4827 {
4828 /* Tricky. `for i\nin ...' should not have a semicolon, but
4829 `for i\ndo ...' should. We do what we can. */
4830 for (i = shell_input_line_index; whitespace (shell_input_line[i]); i++)
4831 ;
4832 if (shell_input_line[i] && shell_input_line[i] == 'i' && shell_input_line[i+1] == 'n')
4833 return " ";
4834 return ";";
4835 }
4836 else if (two_tokens_ago == CASE && token_before_that == WORD && (parser_state & PST_CASESTMT))
4837 return " ";
4838
4839 for (i = 0; no_semi_successors[i]; i++)
4840 {
4841 if (token_before_that == no_semi_successors[i])
4842 return (" ");
4843 }
4844
4845 return ("; ");
4846 }
4847 #endif /* HISTORY */
4848
4849 /* Issue a prompt, or prepare to issue a prompt when the next character
4850 is read. */
4851 static void
4852 prompt_again ()
4853 {
4854 char *temp_prompt;
4855
4856 if (interactive == 0 || expanding_alias ()) /* XXX */
4857 return;
4858
4859 ps1_prompt = get_string_value ("PS1");
4860 ps2_prompt = get_string_value ("PS2");
4861
4862 if (!prompt_string_pointer)
4863 prompt_string_pointer = &ps1_prompt;
4864
4865 temp_prompt = *prompt_string_pointer
4866 ? decode_prompt_string (*prompt_string_pointer)
4867 : (char *)NULL;
4868
4869 if (temp_prompt == 0)
4870 {
4871 temp_prompt = (char *)xmalloc (1);
4872 temp_prompt[0] = '\0';
4873 }
4874
4875 current_prompt_string = *prompt_string_pointer;
4876 prompt_string_pointer = &ps2_prompt;
4877
4878 #if defined (READLINE)
4879 if (!no_line_editing)
4880 {
4881 FREE (current_readline_prompt);
4882 current_readline_prompt = temp_prompt;
4883 }
4884 else
4885 #endif /* READLINE */
4886 {
4887 FREE (current_decoded_prompt);
4888 current_decoded_prompt = temp_prompt;
4889 }
4890 }
4891
4892 int
4893 get_current_prompt_level ()
4894 {
4895 return ((current_prompt_string && current_prompt_string == ps2_prompt) ? 2 : 1);
4896 }
4897
4898 void
4899 set_current_prompt_level (x)
4900 int x;
4901 {
4902 prompt_string_pointer = (x == 2) ? &ps2_prompt : &ps1_prompt;
4903 current_prompt_string = *prompt_string_pointer;
4904 }
4905
4906 static void
4907 print_prompt ()
4908 {
4909 fprintf (stderr, "%s", current_decoded_prompt);
4910 fflush (stderr);
4911 }
4912
4913 /* Return a string which will be printed as a prompt. The string
4914 may contain special characters which are decoded as follows:
4915
4916 \a bell (ascii 07)
4917 \d the date in Day Mon Date format
4918 \e escape (ascii 033)
4919 \h the hostname up to the first `.'
4920 \H the hostname
4921 \j the number of active jobs
4922 \l the basename of the shell's tty device name
4923 \n CRLF
4924 \r CR
4925 \s the name of the shell
4926 \t the time in 24-hour hh:mm:ss format
4927 \T the time in 12-hour hh:mm:ss format
4928 \@ the time in 12-hour hh:mm am/pm format
4929 \A the time in 24-hour hh:mm format
4930 \D{fmt} the result of passing FMT to strftime(3)
4931 \u your username
4932 \v the version of bash (e.g., 2.00)
4933 \V the release of bash, version + patchlevel (e.g., 2.00.0)
4934 \w the current working directory
4935 \W the last element of $PWD
4936 \! the history number of this command
4937 \# the command number of this command
4938 \$ a $ or a # if you are root
4939 \nnn character code nnn in octal
4940 \\ a backslash
4941 \[ begin a sequence of non-printing chars
4942 \] end a sequence of non-printing chars
4943 */
4944 #define PROMPT_GROWTH 48
4945 char *
4946 decode_prompt_string (string)
4947 char *string;
4948 {
4949 WORD_LIST *list;
4950 char *result, *t;
4951 struct dstack save_dstack;
4952 int last_exit_value, last_comsub_pid;
4953 #if defined (PROMPT_STRING_DECODE)
4954 int result_size, result_index;
4955 int c, n, i;
4956 char *temp, octal_string[4];
4957 struct tm *tm;
4958 time_t the_time;
4959 char timebuf[128];
4960 char *timefmt;
4961
4962 result = (char *)xmalloc (result_size = PROMPT_GROWTH);
4963 result[result_index = 0] = 0;
4964 temp = (char *)NULL;
4965
4966 while (c = *string++)
4967 {
4968 if (posixly_correct && c == '!')
4969 {
4970 if (*string == '!')
4971 {
4972 temp = savestring ("!");
4973 goto add_string;
4974 }
4975 else
4976 {
4977 #if !defined (HISTORY)
4978 temp = savestring ("1");
4979 #else /* HISTORY */
4980 temp = itos (history_number ());
4981 #endif /* HISTORY */
4982 string--; /* add_string increments string again. */
4983 goto add_string;
4984 }
4985 }
4986 if (c == '\\')
4987 {
4988 c = *string;
4989
4990 switch (c)
4991 {
4992 case '0':
4993 case '1':
4994 case '2':
4995 case '3':
4996 case '4':
4997 case '5':
4998 case '6':
4999 case '7':
5000 strncpy (octal_string, string, 3);
5001 octal_string[3] = '\0';
5002
5003 n = read_octal (octal_string);
5004 temp = (char *)xmalloc (3);
5005
5006 if (n == CTLESC || n == CTLNUL)
5007 {
5008 temp[0] = CTLESC;
5009 temp[1] = n;
5010 temp[2] = '\0';
5011 }
5012 else if (n == -1)
5013 {
5014 temp[0] = '\\';
5015 temp[1] = '\0';
5016 }
5017 else
5018 {
5019 temp[0] = n;
5020 temp[1] = '\0';
5021 }
5022
5023 for (c = 0; n != -1 && c < 3 && ISOCTAL (*string); c++)
5024 string++;
5025
5026 c = 0; /* tested at add_string: */
5027 goto add_string;
5028
5029 case 'd':
5030 case 't':
5031 case 'T':
5032 case '@':
5033 case 'A':
5034 /* Make the current time/date into a string. */
5035 (void) time (&the_time);
5036 tm = localtime (&the_time);
5037
5038 if (c == 'd')
5039 n = strftime (timebuf, sizeof (timebuf), "%a %b %d", tm);
5040 else if (c == 't')
5041 n = strftime (timebuf, sizeof (timebuf), "%H:%M:%S", tm);
5042 else if (c == 'T')
5043 n = strftime (timebuf, sizeof (timebuf), "%I:%M:%S", tm);
5044 else if (c == '@')
5045 n = strftime (timebuf, sizeof (timebuf), "%I:%M %p", tm);
5046 else if (c == 'A')
5047 n = strftime (timebuf, sizeof (timebuf), "%H:%M", tm);
5048
5049 if (n == 0)
5050 timebuf[0] = '\0';
5051 else
5052 timebuf[sizeof(timebuf) - 1] = '\0';
5053
5054 temp = savestring (timebuf);
5055 goto add_string;
5056
5057 case 'D': /* strftime format */
5058 if (string[1] != '{') /* } */
5059 goto not_escape;
5060
5061 (void) time (&the_time);
5062 tm = localtime (&the_time);
5063 string += 2; /* skip { */
5064 timefmt = xmalloc (strlen (string) + 3);
5065 for (t = timefmt; *string && *string != '}'; )
5066 *t++ = *string++;
5067 *t = '\0';
5068 c = *string; /* tested at add_string */
5069 if (timefmt[0] == '\0')
5070 {
5071 timefmt[0] = '%';
5072 timefmt[1] = 'X'; /* locale-specific current time */
5073 timefmt[2] = '\0';
5074 }
5075 n = strftime (timebuf, sizeof (timebuf), timefmt, tm);
5076 free (timefmt);
5077
5078 if (n == 0)
5079 timebuf[0] = '\0';
5080 else
5081 timebuf[sizeof(timebuf) - 1] = '\0';
5082
5083 if (promptvars || posixly_correct)
5084 /* Make sure that expand_prompt_string is called with a
5085 second argument of Q_DOUBLE_QUOTES if we use this
5086 function here. */
5087 temp = sh_backslash_quote_for_double_quotes (timebuf);
5088 else
5089 temp = savestring (timebuf);
5090 goto add_string;
5091
5092 case 'n':
5093 temp = (char *)xmalloc (3);
5094 temp[0] = no_line_editing ? '\n' : '\r';
5095 temp[1] = no_line_editing ? '\0' : '\n';
5096 temp[2] = '\0';
5097 goto add_string;
5098
5099 case 's':
5100 temp = base_pathname (shell_name);
5101 temp = savestring (temp);
5102 goto add_string;
5103
5104 case 'v':
5105 case 'V':
5106 temp = (char *)xmalloc (16);
5107 if (c == 'v')
5108 strcpy (temp, dist_version);
5109 else
5110 sprintf (temp, "%s.%d", dist_version, patch_level);
5111 goto add_string;
5112
5113 case 'w':
5114 case 'W':
5115 {
5116 /* Use the value of PWD because it is much more efficient. */
5117 char t_string[PATH_MAX];
5118 int tlen;
5119
5120 temp = get_string_value ("PWD");
5121
5122 if (temp == 0)
5123 {
5124 if (getcwd (t_string, sizeof(t_string)) == 0)
5125 {
5126 t_string[0] = '.';
5127 tlen = 1;
5128 }
5129 else
5130 tlen = strlen (t_string);
5131 }
5132 else
5133 {
5134 tlen = sizeof (t_string) - 1;
5135 strncpy (t_string, temp, tlen);
5136 }
5137 t_string[tlen] = '\0';
5138
5139 #if defined (MACOSX)
5140 /* Convert from "fs" format to "input" format */
5141 temp = fnx_fromfs (t_string, strlen (t_string));
5142 if (temp != t_string)
5143 strcpy (t_string, temp);
5144 #endif
5145
5146 #define ROOT_PATH(x) ((x)[0] == '/' && (x)[1] == 0)
5147 #define DOUBLE_SLASH_ROOT(x) ((x)[0] == '/' && (x)[1] == '/' && (x)[2] == 0)
5148 /* Abbreviate \W as ~ if $PWD == $HOME */
5149 if (c == 'W' && (((t = get_string_value ("HOME")) == 0) || STREQ (t, t_string) == 0))
5150 {
5151 if (ROOT_PATH (t_string) == 0 && DOUBLE_SLASH_ROOT (t_string) == 0)
5152 {
5153 t = strrchr (t_string, '/');
5154 if (t)
5155 strcpy (t_string, t + 1);
5156 }
5157 }
5158 #undef ROOT_PATH
5159 #undef DOUBLE_SLASH_ROOT
5160 else
5161 /* polite_directory_format is guaranteed to return a string
5162 no longer than PATH_MAX - 1 characters. */
5163 strcpy (t_string, polite_directory_format (t_string));
5164
5165 temp = trim_pathname (t_string, PATH_MAX - 1);
5166 /* If we're going to be expanding the prompt string later,
5167 quote the directory name. */
5168 if (promptvars || posixly_correct)
5169 /* Make sure that expand_prompt_string is called with a
5170 second argument of Q_DOUBLE_QUOTES if we use this
5171 function here. */
5172 temp = sh_backslash_quote_for_double_quotes (t_string);
5173 else
5174 temp = savestring (t_string);
5175
5176 goto add_string;
5177 }
5178
5179 case 'u':
5180 if (current_user.user_name == 0)
5181 get_current_user_info ();
5182 temp = savestring (current_user.user_name);
5183 goto add_string;
5184
5185 case 'h':
5186 case 'H':
5187 temp = savestring (current_host_name);
5188 if (c == 'h' && (t = (char *)strchr (temp, '.')))
5189 *t = '\0';
5190 goto add_string;
5191
5192 case '#':
5193 temp = itos (current_command_number);
5194 goto add_string;
5195
5196 case '!':
5197 #if !defined (HISTORY)
5198 temp = savestring ("1");
5199 #else /* HISTORY */
5200 temp = itos (history_number ());
5201 #endif /* HISTORY */
5202 goto add_string;
5203
5204 case '$':
5205 t = temp = (char *)xmalloc (3);
5206 if ((promptvars || posixly_correct) && (current_user.euid != 0))
5207 *t++ = '\\';
5208 *t++ = current_user.euid == 0 ? '#' : '$';
5209 *t = '\0';
5210 goto add_string;
5211
5212 case 'j':
5213 temp = itos (count_all_jobs ());
5214 goto add_string;
5215
5216 case 'l':
5217 #if defined (HAVE_TTYNAME)
5218 temp = (char *)ttyname (fileno (stdin));
5219 t = temp ? base_pathname (temp) : "tty";
5220 temp = savestring (t);
5221 #else
5222 temp = savestring ("tty");
5223 #endif /* !HAVE_TTYNAME */
5224 goto add_string;
5225
5226 #if defined (READLINE)
5227 case '[':
5228 case ']':
5229 if (no_line_editing)
5230 {
5231 string++;
5232 break;
5233 }
5234 temp = (char *)xmalloc (3);
5235 n = (c == '[') ? RL_PROMPT_START_IGNORE : RL_PROMPT_END_IGNORE;
5236 i = 0;
5237 if (n == CTLESC || n == CTLNUL)
5238 temp[i++] = CTLESC;
5239 temp[i++] = n;
5240 temp[i] = '\0';
5241 goto add_string;
5242 #endif /* READLINE */
5243
5244 case '\\':
5245 case 'a':
5246 case 'e':
5247 case 'r':
5248 temp = (char *)xmalloc (2);
5249 if (c == 'a')
5250 temp[0] = '\07';
5251 else if (c == 'e')
5252 temp[0] = '\033';
5253 else if (c == 'r')
5254 temp[0] = '\r';
5255 else /* (c == '\\') */
5256 temp[0] = c;
5257 temp[1] = '\0';
5258 goto add_string;
5259
5260 default:
5261 not_escape:
5262 temp = (char *)xmalloc (3);
5263 temp[0] = '\\';
5264 temp[1] = c;
5265 temp[2] = '\0';
5266
5267 add_string:
5268 if (c)
5269 string++;
5270 result =
5271 sub_append_string (temp, result, &result_index, &result_size);
5272 temp = (char *)NULL; /* Freed in sub_append_string (). */
5273 result[result_index] = '\0';
5274 break;
5275 }
5276 }
5277 else
5278 {
5279 RESIZE_MALLOCED_BUFFER (result, result_index, 3, result_size, PROMPT_GROWTH);
5280 result[result_index++] = c;
5281 result[result_index] = '\0';
5282 }
5283 }
5284 #else /* !PROMPT_STRING_DECODE */
5285 result = savestring (string);
5286 #endif /* !PROMPT_STRING_DECODE */
5287
5288 /* Save the delimiter stack and point `dstack' to temp space so any
5289 command substitutions in the prompt string won't result in screwing
5290 up the parser's quoting state. */
5291 save_dstack = dstack;
5292 dstack = temp_dstack;
5293 dstack.delimiter_depth = 0;
5294
5295 /* Perform variable and parameter expansion and command substitution on
5296 the prompt string. */
5297 if (promptvars || posixly_correct)
5298 {
5299 last_exit_value = last_command_exit_value;
5300 last_comsub_pid = last_command_subst_pid;
5301 list = expand_prompt_string (result, Q_DOUBLE_QUOTES, 0);
5302 free (result);
5303 result = string_list (list);
5304 dispose_words (list);
5305 last_command_exit_value = last_exit_value;
5306 last_command_subst_pid = last_comsub_pid;
5307 }
5308 else
5309 {
5310 t = dequote_string (result);
5311 free (result);
5312 result = t;
5313 }
5314
5315 dstack = save_dstack;
5316
5317 return (result);
5318 }
5319
5320 /************************************************
5321 * *
5322 * ERROR HANDLING *
5323 * *
5324 ************************************************/
5325
5326 /* Report a syntax error, and restart the parser. Call here for fatal
5327 errors. */
5328 int
5329 yyerror (msg)
5330 const char *msg;
5331 {
5332 report_syntax_error ((char *)NULL);
5333 reset_parser ();
5334 return (0);
5335 }
5336
5337 static char *
5338 error_token_from_token (tok)
5339 int tok;
5340 {
5341 char *t;
5342
5343 if (t = find_token_in_alist (tok, word_token_alist, 0))
5344 return t;
5345
5346 if (t = find_token_in_alist (tok, other_token_alist, 0))
5347 return t;
5348
5349 t = (char *)NULL;
5350 /* This stuff is dicy and needs closer inspection */
5351 switch (current_token)
5352 {
5353 case WORD:
5354 case ASSIGNMENT_WORD:
5355 if (yylval.word)
5356 t = savestring (yylval.word->word);
5357 break;
5358 case NUMBER:
5359 t = itos (yylval.number);
5360 break;
5361 case ARITH_CMD:
5362 if (yylval.word_list)
5363 t = string_list (yylval.word_list);
5364 break;
5365 case ARITH_FOR_EXPRS:
5366 if (yylval.word_list)
5367 t = string_list_internal (yylval.word_list, " ; ");
5368 break;
5369 case COND_CMD:
5370 t = (char *)NULL; /* punt */
5371 break;
5372 }
5373
5374 return t;
5375 }
5376
5377 static char *
5378 error_token_from_text ()
5379 {
5380 char *msg, *t;
5381 int token_end, i;
5382
5383 t = shell_input_line;
5384 i = shell_input_line_index;
5385 token_end = 0;
5386 msg = (char *)NULL;
5387
5388 if (i && t[i] == '\0')
5389 i--;
5390
5391 while (i && (whitespace (t[i]) || t[i] == '\n'))
5392 i--;
5393
5394 if (i)
5395 token_end = i + 1;
5396
5397 while (i && (member (t[i], " \n\t;|&") == 0))
5398 i--;
5399
5400 while (i != token_end && (whitespace (t[i]) || t[i] == '\n'))
5401 i++;
5402
5403 /* Return our idea of the offending token. */
5404 if (token_end || (i == 0 && token_end == 0))
5405 {
5406 if (token_end)
5407 msg = substring (t, i, token_end);
5408 else /* one-character token */
5409 {
5410 msg = (char *)xmalloc (2);
5411 msg[0] = t[i];
5412 msg[1] = '\0';
5413 }
5414 }
5415
5416 return (msg);
5417 }
5418
5419 static void
5420 print_offending_line ()
5421 {
5422 char *msg;
5423 int token_end;
5424
5425 msg = savestring (shell_input_line);
5426 token_end = strlen (msg);
5427 while (token_end && msg[token_end - 1] == '\n')
5428 msg[--token_end] = '\0';
5429
5430 parser_error (line_number, "`%s'", msg);
5431 free (msg);
5432 }
5433
5434 /* Report a syntax error with line numbers, etc.
5435 Call here for recoverable errors. If you have a message to print,
5436 then place it in MESSAGE, otherwise pass NULL and this will figure
5437 out an appropriate message for you. */
5438 static void
5439 report_syntax_error (message)
5440 char *message;
5441 {
5442 char *msg;
5443
5444 if (message)
5445 {
5446 parser_error (line_number, "%s", message);
5447 if (interactive && EOF_Reached)
5448 EOF_Reached = 0;
5449 last_command_exit_value = parse_and_execute_level ? EX_BADSYNTAX : EX_BADUSAGE;
5450 return;
5451 }
5452
5453 /* If the line of input we're reading is not null, try to find the
5454 objectionable token. First, try to figure out what token the
5455 parser's complaining about by looking at current_token. */
5456 if (current_token != 0 && EOF_Reached == 0 && (msg = error_token_from_token (current_token)))
5457 {
5458 parser_error (line_number, _("syntax error near unexpected token `%s'"), msg);
5459 free (msg);
5460
5461 if (interactive == 0)
5462 print_offending_line ();
5463
5464 last_command_exit_value = parse_and_execute_level ? EX_BADSYNTAX : EX_BADUSAGE;
5465 return;
5466 }
5467
5468 /* If looking at the current token doesn't prove fruitful, try to find the
5469 offending token by analyzing the text of the input line near the current
5470 input line index and report what we find. */
5471 if (shell_input_line && *shell_input_line)
5472 {
5473 msg = error_token_from_text ();
5474 if (msg)
5475 {
5476 parser_error (line_number, _("syntax error near `%s'"), msg);
5477 free (msg);
5478 }
5479
5480 /* If not interactive, print the line containing the error. */
5481 if (interactive == 0)
5482 print_offending_line ();
5483 }
5484 else
5485 {
5486 msg = EOF_Reached ? _("syntax error: unexpected end of file") : _("syntax error");
5487 parser_error (line_number, "%s", msg);
5488 /* When the shell is interactive, this file uses EOF_Reached
5489 only for error reporting. Other mechanisms are used to
5490 decide whether or not to exit. */
5491 if (interactive && EOF_Reached)
5492 EOF_Reached = 0;
5493 }
5494
5495 last_command_exit_value = parse_and_execute_level ? EX_BADSYNTAX : EX_BADUSAGE;
5496 }
5497
5498 /* ??? Needed function. ??? We have to be able to discard the constructs
5499 created during parsing. In the case of error, we want to return
5500 allocated objects to the memory pool. In the case of no error, we want
5501 to throw away the information about where the allocated objects live.
5502 (dispose_command () will actually free the command.) */
5503 static void
5504 discard_parser_constructs (error_p)
5505 int error_p;
5506 {
5507 }
5508
5509 /************************************************
5510 * *
5511 * EOF HANDLING *
5512 * *
5513 ************************************************/
5514
5515 /* Do that silly `type "bye" to exit' stuff. You know, "ignoreeof". */
5516
5517 /* A flag denoting whether or not ignoreeof is set. */
5518 int ignoreeof = 0;
5519
5520 /* The number of times that we have encountered an EOF character without
5521 another character intervening. When this gets above the limit, the
5522 shell terminates. */
5523 int eof_encountered = 0;
5524
5525 /* The limit for eof_encountered. */
5526 int eof_encountered_limit = 10;
5527
5528 /* If we have EOF as the only input unit, this user wants to leave
5529 the shell. If the shell is not interactive, then just leave.
5530 Otherwise, if ignoreeof is set, and we haven't done this the
5531 required number of times in a row, print a message. */
5532 static void
5533 handle_eof_input_unit ()
5534 {
5535 if (interactive)
5536 {
5537 /* shell.c may use this to decide whether or not to write out the
5538 history, among other things. We use it only for error reporting
5539 in this file. */
5540 if (EOF_Reached)
5541 EOF_Reached = 0;
5542
5543 /* If the user wants to "ignore" eof, then let her do so, kind of. */
5544 if (ignoreeof)
5545 {
5546 if (eof_encountered < eof_encountered_limit)
5547 {
5548 fprintf (stderr, _("Use \"%s\" to leave the shell.\n"),
5549 login_shell ? "logout" : "exit");
5550 eof_encountered++;
5551 /* Reset the parsing state. */
5552 last_read_token = current_token = '\n';
5553 /* Reset the prompt string to be $PS1. */
5554 prompt_string_pointer = (char **)NULL;
5555 prompt_again ();
5556 return;
5557 }
5558 }
5559
5560 /* In this case EOF should exit the shell. Do it now. */
5561 reset_parser ();
5562 exit_builtin ((WORD_LIST *)NULL);
5563 }
5564 else
5565 {
5566 /* We don't write history files, etc., for non-interactive shells. */
5567 EOF_Reached = 1;
5568 }
5569 }
5570
5571 /************************************************
5572 * *
5573 * STRING PARSING FUNCTIONS *
5574 * *
5575 ************************************************/
5576
5577 /* It's very important that these two functions treat the characters
5578 between ( and ) identically. */
5579
5580 static WORD_LIST parse_string_error;
5581
5582 /* Take a string and run it through the shell parser, returning the
5583 resultant word list. Used by compound array assignment. */
5584 WORD_LIST *
5585 parse_string_to_word_list (s, flags, whom)
5586 char *s;
5587 int flags;
5588 const char *whom;
5589 {
5590 WORD_LIST *wl;
5591 int tok, orig_current_token, orig_line_number, orig_input_terminator;
5592 int orig_line_count;
5593 int old_echo_input, old_expand_aliases;
5594 #if defined (HISTORY)
5595 int old_remember_on_history, old_history_expansion_inhibited;
5596 #endif
5597
5598 #if defined (HISTORY)
5599 old_remember_on_history = remember_on_history;
5600 # if defined (BANG_HISTORY)
5601 old_history_expansion_inhibited = history_expansion_inhibited;
5602 # endif
5603 bash_history_disable ();
5604 #endif
5605
5606 orig_line_number = line_number;
5607 orig_line_count = current_command_line_count;
5608 orig_input_terminator = shell_input_line_terminator;
5609 old_echo_input = echo_input_at_read;
5610 old_expand_aliases = expand_aliases;
5611
5612 push_stream (1);
5613 last_read_token = WORD; /* WORD to allow reserved words here */
5614 current_command_line_count = 0;
5615 echo_input_at_read = expand_aliases = 0;
5616
5617 with_input_from_string (s, whom);
5618 wl = (WORD_LIST *)NULL;
5619
5620 if (flags & 1)
5621 parser_state |= PST_COMPASSIGN|PST_REPARSE;
5622
5623 while ((tok = read_token (READ)) != yacc_EOF)
5624 {
5625 if (tok == '\n' && *bash_input.location.string == '\0')
5626 break;
5627 if (tok == '\n') /* Allow newlines in compound assignments */
5628 continue;
5629 if (tok != WORD && tok != ASSIGNMENT_WORD)
5630 {
5631 line_number = orig_line_number + line_number - 1;
5632 orig_current_token = current_token;
5633 current_token = tok;
5634 yyerror (NULL); /* does the right thing */
5635 current_token = orig_current_token;
5636 if (wl)
5637 dispose_words (wl);
5638 wl = &parse_string_error;
5639 break;
5640 }
5641 wl = make_word_list (yylval.word, wl);
5642 }
5643
5644 last_read_token = '\n';
5645 pop_stream ();
5646
5647 #if defined (HISTORY)
5648 remember_on_history = old_remember_on_history;
5649 # if defined (BANG_HISTORY)
5650 history_expansion_inhibited = old_history_expansion_inhibited;
5651 # endif /* BANG_HISTORY */
5652 #endif /* HISTORY */
5653
5654 echo_input_at_read = old_echo_input;
5655 expand_aliases = old_expand_aliases;
5656
5657 current_command_line_count = orig_line_count;
5658 shell_input_line_terminator = orig_input_terminator;
5659
5660 if (flags & 1)
5661 parser_state &= ~(PST_COMPASSIGN|PST_REPARSE);
5662
5663 if (wl == &parse_string_error)
5664 {
5665 last_command_exit_value = EXECUTION_FAILURE;
5666 if (interactive_shell == 0 && posixly_correct)
5667 jump_to_top_level (FORCE_EOF);
5668 else
5669 jump_to_top_level (DISCARD);
5670 }
5671
5672 return (REVERSE_LIST (wl, WORD_LIST *));
5673 }
5674
5675 static char *
5676 parse_compound_assignment (retlenp)
5677 int *retlenp;
5678 {
5679 WORD_LIST *wl, *rl;
5680 int tok, orig_line_number, orig_token_size, orig_last_token, assignok;
5681 char *saved_token, *ret;
5682
5683 saved_token = token;
5684 orig_token_size = token_buffer_size;
5685 orig_line_number = line_number;
5686 orig_last_token = last_read_token;
5687
5688 last_read_token = WORD; /* WORD to allow reserved words here */
5689
5690 token = (char *)NULL;
5691 token_buffer_size = 0;
5692
5693 assignok = parser_state&PST_ASSIGNOK; /* XXX */
5694
5695 wl = (WORD_LIST *)NULL; /* ( */
5696 parser_state |= PST_COMPASSIGN;
5697
5698 while ((tok = read_token (READ)) != ')')
5699 {
5700 if (tok == '\n') /* Allow newlines in compound assignments */
5701 {
5702 if (SHOULD_PROMPT ())
5703 prompt_again ();
5704 continue;
5705 }
5706 if (tok != WORD && tok != ASSIGNMENT_WORD)
5707 {
5708 current_token = tok; /* for error reporting */
5709 if (tok == yacc_EOF) /* ( */
5710 parser_error (orig_line_number, _("unexpected EOF while looking for matching `)'"));
5711 else
5712 yyerror(NULL); /* does the right thing */
5713 if (wl)
5714 dispose_words (wl);
5715 wl = &parse_string_error;
5716 break;
5717 }
5718 wl = make_word_list (yylval.word, wl);
5719 }
5720
5721 FREE (token);
5722 token = saved_token;
5723 token_buffer_size = orig_token_size;
5724
5725 parser_state &= ~PST_COMPASSIGN;
5726
5727 if (wl == &parse_string_error)
5728 {
5729 last_command_exit_value = EXECUTION_FAILURE;
5730 last_read_token = '\n'; /* XXX */
5731 if (interactive_shell == 0 && posixly_correct)
5732 jump_to_top_level (FORCE_EOF);
5733 else
5734 jump_to_top_level (DISCARD);
5735 }
5736
5737 last_read_token = orig_last_token; /* XXX - was WORD? */
5738
5739 if (wl)
5740 {
5741 rl = REVERSE_LIST (wl, WORD_LIST *);
5742 ret = string_list (rl);
5743 dispose_words (rl);
5744 }
5745 else
5746 ret = (char *)NULL;
5747
5748 if (retlenp)
5749 *retlenp = (ret && *ret) ? strlen (ret) : 0;
5750
5751 if (assignok)
5752 parser_state |= PST_ASSIGNOK;
5753
5754 return ret;
5755 }
5756
5757 /************************************************
5758 * *
5759 * SAVING AND RESTORING PARTIAL PARSE STATE *
5760 * *
5761 ************************************************/
5762
5763 sh_parser_state_t *
5764 save_parser_state (ps)
5765 sh_parser_state_t *ps;
5766 {
5767 #if defined (ARRAY_VARS)
5768 SHELL_VAR *v;
5769 #endif
5770
5771 if (ps == 0)
5772 ps = (sh_parser_state_t *)xmalloc (sizeof (sh_parser_state_t));
5773 if (ps == 0)
5774 return ((sh_parser_state_t *)NULL);
5775
5776 ps->parser_state = parser_state;
5777 ps->token_state = save_token_state ();
5778
5779 ps->input_line_terminator = shell_input_line_terminator;
5780 ps->eof_encountered = eof_encountered;
5781
5782 ps->current_command_line_count = current_command_line_count;
5783
5784 #if defined (HISTORY)
5785 ps->remember_on_history = remember_on_history;
5786 # if defined (BANG_HISTORY)
5787 ps->history_expansion_inhibited = history_expansion_inhibited;
5788 # endif
5789 #endif
5790
5791 ps->last_command_exit_value = last_command_exit_value;
5792 #if defined (ARRAY_VARS)
5793 v = find_variable ("PIPESTATUS");
5794 if (v && array_p (v) && array_cell (v))
5795 ps->pipestatus = array_copy (array_cell (v));
5796 else
5797 ps->pipestatus = (ARRAY *)NULL;
5798 #endif
5799
5800 ps->last_shell_builtin = last_shell_builtin;
5801 ps->this_shell_builtin = this_shell_builtin;
5802
5803 ps->expand_aliases = expand_aliases;
5804 ps->echo_input_at_read = echo_input_at_read;
5805
5806 return (ps);
5807 }
5808
5809 void
5810 restore_parser_state (ps)
5811 sh_parser_state_t *ps;
5812 {
5813 #if defined (ARRAY_VARS)
5814 SHELL_VAR *v;
5815 #endif
5816
5817 if (ps == 0)
5818 return;
5819
5820 parser_state = ps->parser_state;
5821 if (ps->token_state)
5822 {
5823 restore_token_state (ps->token_state);
5824 free (ps->token_state);
5825 }
5826
5827 shell_input_line_terminator = ps->input_line_terminator;
5828 eof_encountered = ps->eof_encountered;
5829
5830 current_command_line_count = ps->current_command_line_count;
5831
5832 #if defined (HISTORY)
5833 remember_on_history = ps->remember_on_history;
5834 # if defined (BANG_HISTORY)
5835 history_expansion_inhibited = ps->history_expansion_inhibited;
5836 # endif
5837 #endif
5838
5839 last_command_exit_value = ps->last_command_exit_value;
5840 #if defined (ARRAY_VARS)
5841 v = find_variable ("PIPESTATUS");
5842 if (v && array_p (v) && array_cell (v))
5843 {
5844 array_dispose (array_cell (v));
5845 var_setarray (v, ps->pipestatus);
5846 }
5847 #endif
5848
5849 last_shell_builtin = ps->last_shell_builtin;
5850 this_shell_builtin = ps->this_shell_builtin;
5851
5852 expand_aliases = ps->expand_aliases;
5853 echo_input_at_read = ps->echo_input_at_read;
5854 }
5855
5856 /************************************************
5857 * *
5858 * MULTIBYTE CHARACTER HANDLING *
5859 * *
5860 ************************************************/
5861
5862 #if defined (HANDLE_MULTIBYTE)
5863 static void
5864 set_line_mbstate ()
5865 {
5866 int i, previ, len, c;
5867 mbstate_t mbs, prevs;
5868 size_t mbclen;
5869
5870 if (shell_input_line == NULL)
5871 return;
5872 len = strlen (shell_input_line); /* XXX - shell_input_line_len ? */
5873 FREE (shell_input_line_property);
5874 shell_input_line_property = (char *)xmalloc (len + 1);
5875
5876 memset (&prevs, '\0', sizeof (mbstate_t));
5877 for (i = previ = 0; i < len; i++)
5878 {
5879 mbs = prevs;
5880
5881 c = shell_input_line[i];
5882 if (c == EOF)
5883 {
5884 int j;
5885 for (j = i; j < len; j++)
5886 shell_input_line_property[j] = 1;
5887 break;
5888 }
5889
5890 mbclen = mbrlen (shell_input_line + previ, i - previ + 1, &mbs);
5891 if (mbclen == 1 || mbclen == (size_t)-1)
5892 {
5893 mbclen = 1;
5894 previ = i + 1;
5895 }
5896 else if (mbclen == (size_t)-2)
5897 mbclen = 0;
5898 else if (mbclen > 1)
5899 {
5900 mbclen = 0;
5901 previ = i + 1;
5902 prevs = mbs;
5903 }
5904 else
5905 {
5906 /* XXX - what to do if mbrlen returns 0? (null wide character) */
5907 int j;
5908 for (j = i; j < len; j++)
5909 shell_input_line_property[j] = 1;
5910 break;
5911 }
5912
5913 shell_input_line_property[i] = mbclen;
5914 }
5915 }
5916 #endif /* HANDLE_MULTIBYTE */