glibc/posix/wordexp.c
Ulrich Drepper 76fbcfdd66 Update.
1998-03-14 00:52  Tim Waugh  <tim@cyberelk.demon.co.uk>

	* posix/wordexp.c (parse_param): Perform field-splitting after
	expanding positional parameter.

	* posix/wordexp-tst.sh: Test that field-splitting is performed
	after expanding positional parameter.

	* posix/wordexp.c (parse_param): Fixed memory leak in
	field-splitting after parameter expansion.

1998-03-14  Ulrich Drepper  <drepper@cygnus.com>

	* locale/programs/linereader.c (lr_token): Return EOF token at
	EOF.
	(get_toplvl_escape): Correctly terminate loop at EOF.
	Patch by Cristian Gafton <gafton@redhat.com>.
1998-03-14 09:27:24 +00:00

2227 lines
46 KiB
C

/* POSIX.2 wordexp implementation.
Copyright (C) 1997, 1998 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Tim Waugh <tim@cyberelk.demon.co.uk>.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Library General Public License as
published by the Free Software Foundation; either version 2 of the
License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Library General Public License for more details.
You should have received a copy of the GNU Library General Public
License along with the GNU C Library; see the file COPYING.LIB. If not,
write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
Boston, MA 02111-1307, USA. */
#include <wordexp.h>
#include <signal.h>
#include <stdlib.h>
#include <pwd.h>
#include <sys/types.h>
#include <string.h>
#include <glob.h>
#include <ctype.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/stat.h>
#include <paths.h>
#include <errno.h>
#include <sys/param.h>
#include <stdio.h>
#include <fnmatch.h>
#include <stdio-common/_itoa.h>
/* Undefine the following line for the production version. */
/* #define NDEBUG 1 */
#include <assert.h>
/*
* This is a recursive-descent-style word expansion routine.
*/
/* These variables are defined and initialized in the startup code. */
extern int __libc_argc;
extern char **__libc_argv;
/* Some forward declarations */
static int parse_dollars (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags,
wordexp_t *pwordexp, const char *ifs,
const char *ifs_white, int quoted)
internal_function;
static int parse_backtick (char **word, size_t *word_length,
size_t *max_length, const char *words,
size_t *offset, int flags, wordexp_t *pwordexp,
const char *ifs, const char *ifs_white)
internal_function;
static int eval_expr (char *expr, long int *result) internal_function;
/* The w_*() functions manipulate word lists. */
#define W_CHUNK (100)
static inline char *
w_addchar (char *buffer, size_t *actlen, size_t *maxlen, char ch)
/* (lengths exclude trailing zero) */
{
/* Add a character to the buffer, allocating room for it if needed.
*/
if (*actlen == *maxlen)
{
char *old_buffer = buffer;
assert (buffer == NULL || *maxlen != 0);
*maxlen += W_CHUNK;
buffer = realloc (buffer, 1 + *maxlen);
if (buffer == NULL)
free (old_buffer);
}
if (buffer != NULL)
{
buffer[*actlen] = ch;
buffer[++(*actlen)] = '\0';
}
return buffer;
}
static char *
internal_function
w_addmem (char *buffer, size_t *actlen, size_t *maxlen, const char *str,
size_t len)
{
/* Add a string to the buffer, allocating room for it if needed.
*/
if (*actlen + len > *maxlen)
{
char *old_buffer = buffer;
assert (buffer == NULL || *maxlen != 0);
*maxlen += MAX (2 * len, W_CHUNK);
buffer = realloc (old_buffer, 1 + *maxlen);
if (buffer == NULL)
free (old_buffer);
}
if (buffer != NULL)
{
*((char *) __mempcpy (&buffer[*actlen], str, len)) = '\0';
*actlen += len;
}
return buffer;
}
static char *
internal_function
w_addstr (char *buffer, size_t *actlen, size_t *maxlen, const char *str)
/* (lengths exclude trailing zero) */
{
/* Add a string to the buffer, allocating room for it if needed.
*/
size_t len;
assert (str != NULL); /* w_addstr only called from this file */
len = strlen (str);
return w_addmem (buffer, actlen, maxlen, str, len);
}
static int
internal_function
w_addword (wordexp_t *pwordexp, char *word)
{
/* Add a word to the wordlist */
size_t num_p;
num_p = 2 + pwordexp->we_wordc + pwordexp->we_offs;
pwordexp->we_wordv = realloc (pwordexp->we_wordv, sizeof (char *) * num_p);
if (pwordexp->we_wordv != NULL)
{
pwordexp->we_wordv[pwordexp->we_wordc++] = word;
pwordexp->we_wordv[pwordexp->we_wordc] = NULL;
return 0;
}
return WRDE_NOSPACE;
}
/* The parse_*() functions should leave *offset being the offset in 'words'
* to the last character processed.
*/
static int
internal_function
parse_backslash (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset)
{
/* We are poised _at_ a backslash, not in quotes */
switch (words[1 + *offset])
{
case 0:
/* Backslash is last character of input words */
return WRDE_SYNTAX;
case '\n':
++(*offset);
break;
default:
*word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
if (*word == NULL)
return WRDE_NOSPACE;
++(*offset);
break;
}
return 0;
}
static int
internal_function
parse_qtd_backslash (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset)
{
/* We are poised _at_ a backslash, inside quotes */
switch (words[1 + *offset])
{
case 0:
/* Backslash is last character of input words */
return WRDE_SYNTAX;
case '\n':
++(*offset);
break;
case '$':
case '`':
case '"':
case '\\':
*word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
if (*word == NULL)
return WRDE_NOSPACE;
++(*offset);
break;
default:
*word = w_addchar (*word, word_length, max_length, words[*offset]);
if (*word != NULL)
*word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
if (*word == NULL)
return WRDE_NOSPACE;
++(*offset);
break;
}
return 0;
}
static int
internal_function
parse_tilde (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, size_t wordc)
{
/* We are poised _at_ a tilde */
size_t i;
if (*word_length != 0)
{
if (!((*word)[*word_length - 1] == '=' && wordc == 0))
{
if (!((*word)[*word_length - 1] == ':'
&& strchr (*word, '=') && wordc == 0))
{
*word = w_addchar (*word, word_length, max_length, '~');
return *word ? 0 : WRDE_NOSPACE;
}
}
}
for (i = 1 + *offset; words[i]; i++)
{
if (words[i] == ':' || words[i] == '/' || words[i] == ' ' ||
words[i] == '\t' || words[i] == 0 )
break;
if (words[i] == '\\')
{
*word = w_addchar (*word, word_length, max_length, '~');
return *word ? 0 : WRDE_NOSPACE;
}
}
if (i == 1 + *offset)
{
/* Tilde appears on its own */
uid_t uid;
struct passwd pwd, *tpwd;
int buflen = 1000;
char* buffer = __alloca (buflen);
int result;
uid = getuid ();
while ((result = __getpwuid_r (uid, &pwd, buffer, buflen, &tpwd)) != 0
&& errno == ERANGE)
{
buflen += 1000;
buffer = __alloca (buflen);
}
if (result == 0 && pwd.pw_dir != NULL)
{
*word = w_addstr (*word, word_length, max_length, pwd.pw_dir);
if (*word == NULL)
return WRDE_NOSPACE;
}
else
{
*word = w_addchar (*word, word_length, max_length, '~');
if (*word == NULL)
return WRDE_NOSPACE;
}
}
else
{
/* Look up user name in database to get home directory */
char *user = __strndup (&words[1 + *offset], i - *offset);
struct passwd pwd, *tpwd;
int buflen = 1000;
char* buffer = __alloca (buflen);
int result;
while ((result = __getpwnam_r (user, &pwd, buffer, buflen, &tpwd)) != 0
&& errno == ERANGE)
{
buflen += 1000;
buffer = __alloca (buflen);
}
if (result == 0 && pwd.pw_dir)
*word = w_addstr (*word, word_length, max_length, pwd.pw_dir);
else
{
/* (invalid login name) */
*word = w_addchar (*word, word_length, max_length, '~');
if (*word != NULL)
*word = w_addstr (*word, word_length, max_length, user);
}
*offset = i - 1;
}
return *word ? 0 : WRDE_NOSPACE;
}
static int
internal_function
parse_glob (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags,
wordexp_t *pwordexp, const char *ifs, const char *ifs_white)
{
/* We are poised just after a '*', a '[' or a '?'. */
int error;
glob_t globbuf;
int match;
char *matching_word;
int quoted = 0; /* 1 if singly-quoted, 2 if doubly */
for (; words[*offset]; (*offset)++)
{
if ((ifs && strchr (ifs, words[*offset])) ||
(!ifs && strchr (" \t\n", words[*offset])))
/* Reached IFS */
break;
/* Sort out quoting */
if (words[*offset] == '\'')
if (quoted == 0)
{
quoted = 1;
continue;
}
else if (quoted == 1)
{
quoted = 0;
continue;
}
else if (words[*offset] == '"')
if (quoted == 0)
{
quoted = 2;
continue;
}
else if (quoted == 2)
{
quoted = 0;
continue;
}
/* Sort out other special characters */
if (quoted != 1 && words[*offset] == '$')
{
error = parse_dollars (word, word_length, max_length, words, offset,
flags, pwordexp, ifs, ifs_white, quoted == 2);
if (error)
return error;
continue;
}
else if (words[*offset] == '\\')
{
if (quoted)
error = parse_qtd_backslash (word, word_length, max_length, words,
offset);
else
error = parse_backslash (word, word_length, max_length, words,
offset);
if (error)
return error;
continue;
}
*word = w_addchar (*word, word_length, max_length, words[*offset]);
if (*word == NULL)
return WRDE_NOSPACE;
}
error = glob (*word, GLOB_NOCHECK, NULL, &globbuf);
if (error != 0)
{
/* We can only run into memory problems. */
assert (error == GLOB_NOSPACE);
return WRDE_NOSPACE;
}
if (ifs && !*ifs)
{
/* No field splitting allowed */
size_t length = strlen (globbuf.gl_pathv[0]);
*word = realloc (*word, length + 1);
if (*word == NULL)
goto no_space;
memcpy (*word, globbuf.gl_pathv[0], length + 1);
*word_length = length;
for (match = 1; match < globbuf.gl_pathc && *word != NULL; ++match)
{
*word = w_addchar (*word, word_length, max_length, ' ');
if (*word != NULL)
*word = w_addstr (*word, word_length, max_length,
globbuf.gl_pathv[match]);
}
/* Re-parse white space on return */
globfree (&globbuf);
--(*offset);
return *word ? 0 : WRDE_NOSPACE;
}
/* here ifs != "" */
free (*word);
*word = NULL;
*word_length = 0;
matching_word = __strdup (globbuf.gl_pathv[0]);
if (matching_word == NULL)
goto no_space;
if (w_addword (pwordexp, matching_word) == WRDE_NOSPACE)
goto no_space;
for (match = 1; match < globbuf.gl_pathc; ++match)
{
matching_word = __strdup (globbuf.gl_pathv[match]);
if (matching_word == NULL)
goto no_space;
if (w_addword (pwordexp, matching_word) == WRDE_NOSPACE)
goto no_space;
}
globfree (&globbuf);
/* Re-parse white space on return */
--(*offset);
return 0;
no_space:
globfree (&globbuf);
return WRDE_NOSPACE;
}
static int
internal_function
parse_squote (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset)
{
/* We are poised just after a single quote */
for (; words[*offset]; ++(*offset))
{
if (words[*offset] != '\'')
{
*word = w_addchar (*word, word_length, max_length, words[*offset]);
if (*word == NULL)
return WRDE_NOSPACE;
}
else return 0;
}
/* Unterminated string */
return WRDE_SYNTAX;
}
/* Functions to evaluate an arithmetic expression */
static int
internal_function
eval_expr_val (char **expr, long int *result)
{
int sgn = +1;
char *digit;
/* Skip white space */
for (digit = *expr; digit && *digit && isspace (*digit); ++digit);
switch (*digit)
{
case '(':
/* Scan for closing paren */
for (++digit; **expr && **expr != ')'; ++(*expr));
/* Is there one? */
if (!**expr)
return WRDE_SYNTAX;
*(*expr)++ = 0;
if (eval_expr (digit, result))
return WRDE_SYNTAX;
return 0;
case '+': /* Positive value */
++digit;
break;
case '-': /* Negative value */
++digit;
sgn = -1;
break;
default:
if (!isdigit (*digit))
return WRDE_SYNTAX;
}
*result = 0;
for (; *digit && isdigit (*digit); ++digit)
*result = (*result * 10) + (*digit - '0');
*expr = digit;
*result *= sgn;
return 0;
}
static int
internal_function
eval_expr_multdiv (char **expr, long int *result)
{
long int arg;
/* Read a Value */
if (eval_expr_val (expr, result) != 0)
return WRDE_SYNTAX;
while (**expr)
{
/* Skip white space */
for (; *expr && **expr && isspace (**expr); ++(*expr));
if (**expr == '*')
{
++(*expr);
if (eval_expr_val (expr, &arg) != 0)
return WRDE_SYNTAX;
*result *= arg;
}
else if (**expr == '/')
{
++(*expr);
if (eval_expr_val (expr, &arg) != 0)
return WRDE_SYNTAX;
*result /= arg;
}
else break;
}
return 0;
}
static int
internal_function
eval_expr (char *expr, long int *result)
{
long int arg;
/* Read a Multdiv */
if (eval_expr_multdiv (&expr, result) != 0)
return WRDE_SYNTAX;
while (*expr)
{
/* Skip white space */
for (; expr && *expr && isspace (*expr); ++expr);
if (*expr == '+')
{
++expr;
if (eval_expr_multdiv (&expr, &arg) != 0)
return WRDE_SYNTAX;
*result += arg;
}
else if (*expr == '-')
{
++expr;
if (eval_expr_multdiv (&expr, &arg) != 0)
return WRDE_SYNTAX;
*result -= arg;
}
else break;
}
return 0;
}
static int
internal_function
parse_arith (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags, int bracket)
{
/* We are poised just after "$((" or "$[" */
int error;
int paren_depth = 1;
size_t expr_length = 0;
size_t expr_maxlen = 0;
char *expr = NULL;
for (; words[*offset]; ++(*offset))
{
switch (words[*offset])
{
case '$':
error = parse_dollars (&expr, &expr_length, &expr_maxlen,
words, offset, flags, NULL, NULL, NULL, 1);
/* The ``1'' here is to tell parse_dollars not to
* split the fields.
*/
if (error)
{
free (expr);
return error;
}
break;
case '`':
(*offset)++;
error = parse_backtick (&expr, &expr_length, &expr_maxlen,
words, offset, flags, NULL, NULL, NULL);
/* The first NULL here is to tell parse_backtick not to
* split the fields.
*/
if (error)
{
free (expr);
return error;
}
break;
case '\\':
error = parse_qtd_backslash (&expr, &expr_length, &expr_maxlen,
words, offset);
if (error)
{
free (expr);
return error;
}
/* I think that a backslash within an
* arithmetic expansion is bound to
* cause an error sooner or later anyway though.
*/
break;
case ')':
if (--paren_depth == 0)
{
char result[21]; /* 21 = ceil(log10(2^64)) + 1 */
long int numresult = 0;
long long int convertme;
if (bracket || words[1 + *offset] != ')')
return WRDE_SYNTAX;
++(*offset);
/* Go - evaluate. */
if (*expr && eval_expr (expr, &numresult) != 0)
return WRDE_SYNTAX;
if (numresult < 0)
{
convertme = -numresult;
*word = w_addchar (*word, word_length, max_length, '-');
if (!*word)
{
free (expr);
return WRDE_NOSPACE;
}
}
else
convertme = numresult;
result[20] = '\0';
*word = w_addstr (*word, word_length, max_length,
_itoa (convertme, &result[20], 10, 0));
free (expr);
return *word ? 0 : WRDE_NOSPACE;
}
expr = w_addchar (expr, &expr_length, &expr_maxlen, words[*offset]);
if (expr == NULL)
return WRDE_NOSPACE;
break;
case ']':
if (bracket && paren_depth == 1)
{
char result[21]; /* 21 = ceil(log10(2^64)) + 1 */
long int numresult = 0;
/* Go - evaluate. */
if (*expr && eval_expr (expr, &numresult) != 0)
return WRDE_SYNTAX;
result[20] = '\0';
*word = w_addstr (*word, word_length, max_length,
_itoa_word (numresult, &result[20], 10, 0));
free (expr);
return *word ? 0 : WRDE_NOSPACE;
}
free (expr);
return WRDE_SYNTAX;
case '\n':
case ';':
case '{':
case '}':
free (expr);
return WRDE_BADCHAR;
case '(':
++paren_depth;
default:
expr = w_addchar (expr, &expr_length, &expr_maxlen, words[*offset]);
if (expr == NULL)
return WRDE_NOSPACE;
}
}
/* Premature end */
free (expr);
return WRDE_SYNTAX;
}
/* Function to execute a command and retrieve the results */
/* pwordexp contains NULL if field-splitting is forbidden */
static int
internal_function
exec_comm (char *comm, char **word, size_t *word_length, size_t *max_length,
int flags, wordexp_t *pwordexp, const char *ifs,
const char *ifs_white)
{
int fildes[2];
int bufsize = 128;
int buflen;
int i;
char *buffer;
pid_t pid;
/* Don't fork() unless necessary */
if (!comm || !*comm)
return 0;
if (pipe (fildes))
/* Bad */
return WRDE_NOSPACE;
if ((pid = fork ()) < 0)
{
/* Bad */
return WRDE_NOSPACE;
}
if (pid == 0)
{
/* Child */
const char *args[4] = { _PATH_BSHELL, "-c", comm, NULL };
/* Redirect input and output */
dup2 (fildes[1], 1);
/* Close stderr if we have to */
if ((flags & WRDE_SHOWERR) == 0)
close (2);
__execve (_PATH_BSHELL, (char *const *) args, __environ);
/* Bad. What now? */
abort ();
}
/* Parent */
close (fildes[1]);
buffer = __alloca (bufsize);
if (!pwordexp)
{ /* Quoted - no field splitting */
while (1)
{
if ((buflen = read (fildes[0], buffer, bufsize)) < 1)
{
if (__waitpid (pid, NULL, WNOHANG) == 0)
continue;
if ((buflen = read (fildes[0], buffer, bufsize)) < 1)
break;
}
*word = w_addmem (*word, word_length, max_length, buffer, buflen);
if (*word == NULL)
{
close (fildes[0]);
return WRDE_NOSPACE;
}
}
close (fildes[0]);
/* bash chops off a terminating linefeed, which seems sensible */
if ((*word)[*word_length - 1] == '\n')
(*word)[--*word_length] = '\0';
return 0;
}
else
/* Not quoted - split fields */
{
int copying = 0;
/* 'copying' is:
* 0 when searching for first character in a field not IFS white space
* 1 when copying the text of a field
* 2 when searching for possible non-whitespace IFS
*/
while (1)
{
if ((buflen = read (fildes[0], buffer, bufsize)) < 1)
{
if (__waitpid (pid, NULL, WNOHANG) == 0)
continue;
if ((read (fildes[0], buffer, bufsize)) < 1)
break;
}
for (i = 0; i < buflen; ++i)
{
if (strchr (ifs, buffer[i]) != NULL)
{
/* Current character is IFS */
if (strchr (ifs_white, buffer[i]) == NULL)
{
/* Current character is IFS but not whitespace */
if (copying == 2)
{
/* current character
* |
* V
* eg: text<space><comma><space>moretext
*
* So, strip whitespace IFS (like at the start)
*/
copying = 0;
continue;
}
copying = 0;
/* fall through and delimit field.. */
}
else
{
/* Current character is IFS white space */
/* If not copying a field, ignore it */
if (copying != 1)
continue;
/* End of field (search for non-IFS afterwards) */
copying = 2;
}
/* First IFS white space, or IFS non-whitespace.
* Delimit the field. */
if (!*word)
{
/* This field is null, so make it an empty string */
*word = w_addchar (*word, word_length, max_length, 0);
if (*word == NULL)
{
close (fildes[0]);
return WRDE_NOSPACE;
}
}
if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
{
/* Should do __waitpid? */
close (fildes[0]);
return WRDE_NOSPACE;
}
*word = NULL;
*word_length = 0;
*max_length = 0;
/* fall back round the loop.. */
}
else
{
/* Not IFS character */
copying = 1;
*word = w_addchar (*word, word_length, max_length,
buffer[i]);
if (*word == NULL)
{
close (fildes[0]);
return WRDE_NOSPACE;
}
}
}
}
}
close (fildes[0]);
return 0;
}
static int
internal_function
parse_comm (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags, wordexp_t *pwordexp,
const char *ifs, const char *ifs_white)
{
/* We are poised just after "$(" */
int paren_depth = 1;
int error;
size_t comm_length = 0;
size_t comm_maxlen = 0;
char *comm = NULL;
for (; words[*offset]; ++(*offset))
{
switch (words[*offset])
{
case ')':
if (--paren_depth == 0)
{
/* Go -- give script to the shell */
error = exec_comm (comm, word, word_length, max_length, flags,
pwordexp, ifs, ifs_white);
free (comm);
return error;
}
/* This is just part of the script */
comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
if (comm == NULL)
return WRDE_NOSPACE;
break;
case '(':
++paren_depth;
default:
comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
if (comm == NULL)
return WRDE_NOSPACE;
break;
}
}
/* Premature end */
free (comm);
return WRDE_SYNTAX;
}
static int
internal_function
parse_param (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags, wordexp_t *pwordexp,
const char *ifs, const char *ifs_white, int quoted)
{
/* We are poised just after "$" */
enum remove_pattern_enum
{
RP_NONE = 0,
RP_SHORT_LEFT,
RP_LONG_LEFT,
RP_SHORT_RIGHT,
RP_LONG_RIGHT
};
size_t start = *offset;
size_t env_length = 0;
size_t env_maxlen = 0;
size_t pat_length = 0;
size_t pat_maxlen = 0;
char *env = NULL;
char *pattern = NULL;
char *value = NULL;
char action = '\0';
enum remove_pattern_enum remove = RP_NONE;
int colon_seen = 0;
int depth = 0;
int seen_hash = 0;
int free_value = 0;
int error;
for (; words[*offset]; ++(*offset))
{
switch (words[*offset])
{
case '{':
++depth;
if (action != '\0' || remove != RP_NONE)
{
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
}
if (*offset == start)
break;
/* Otherwise evaluate */
/* (and re-parse this character) */
--(*offset);
goto envsubst;
case '}':
if (words[start] != '{')
--(*offset);
if (action != '\0' || remove != RP_NONE)
{
if (--depth)
{
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
}
}
/* Evaluate */
goto envsubst;
case '#':
/* '#' only has special meaning inside braces or as the very
* first character after $ */
if (*offset == start)
{
seen_hash = 1;
goto envsubst;
}
if (words[start] != '{')
{
/* Evaluate */
/* (and re-parse this character) */
--(*offset);
goto envsubst;
}
/* At the start? (i.e. 'string length') */
if (*offset == start + 1)
{
seen_hash = 1;
break;
}
else if (seen_hash)
goto syntax;
/* Separating variable name from prefix pattern? */
if (remove == RP_NONE)
{
remove = RP_SHORT_LEFT;
break;
}
else if (remove == RP_SHORT_LEFT)
{
remove = RP_LONG_LEFT;
break;
}
/* Must be part of prefix/suffix pattern. */
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
case '%':
if (!env || !*env)
goto syntax;
/* Separating variable name from suffix pattern? */
if (remove == RP_NONE)
{
remove = RP_SHORT_RIGHT;
break;
}
else if (remove == RP_SHORT_RIGHT)
{
remove = RP_LONG_RIGHT;
break;
}
/* Must be part of prefix/suffix pattern. */
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
case ':':
if (!env || !*env)
goto syntax;
if (action != '\0' || remove != RP_NONE)
{
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
}
if ((words[1 + *offset] == '-') || (words[1 + *offset] == '=')
|| (words[1 + *offset] == '?') || (words[1 + *offset] == '+'))
{
colon_seen = 1;
break;
}
goto syntax;
case '-':
case '=':
case '?':
case '+':
if (!env || !*env)
goto syntax;
if (seen_hash)
goto syntax;
if (action != '\0' || remove != RP_NONE)
{
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
}
action = words[*offset];
break;
case '\\':
if (action != '\0' || remove != RP_NONE)
{
/* Um. Is this right? */
error = parse_qtd_backslash (word, word_length, max_length,
words, offset);
if (error == 0)
break;
}
else
{
error = WRDE_SYNTAX;
}
if (env)
free (env);
if (pattern != NULL)
free (pattern);
return error;
default:
if (action != '\0' || remove != RP_NONE)
{
pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
words[*offset]);
if (pattern == NULL)
goto no_space;
break;
}
else
{
int special = (strchr ("*@$", words[*offset]) != NULL
|| isdigit (words[*offset]));
if (isalpha (words[*offset]) || special)
{
env = w_addchar (env, &env_length, &env_maxlen,
words[*offset]);
if (env == NULL)
goto no_space;
if (special && words[start] != '{')
goto envsubst;
/* Keep going (get next char) */
break;
}
/* Stop and evaluate, remembering char we stopped at */
--(*offset);
goto envsubst;
}
}
}
/* End of input string -- remember to reparse the character that we stopped
* at. */
--(*offset);
envsubst:
if (words[start] == '{' && words[*offset] != '}')
goto syntax;
if (!env || !*env)
{
if (seen_hash)
{
/* $# expands to the number of positional parameters */
char buffer[21];
buffer[20] = '\0';
*word = w_addstr (*word, word_length, max_length,
_itoa_word (__libc_argc - 1, &buffer[20], 10, 0));
}
else
{
/* Just $ on its own */
*offset = start - 1;
*word = w_addchar (*word, word_length, max_length, '$');
}
if (env)
free (env);
return *word ? 0 : WRDE_NOSPACE;
}
/* Is it a special parameter? */
if (strpbrk (env, "0123456789*@$"))
{
if (env[1])
{
/* Bad substitution if there is more than one character */
free (env);
fprintf (stderr, "${%s}: bad substitution\n", env);
return WRDE_SYNTAX;
}
/* Is it a digit? */
if (isdigit(*env))
{
int n = *env - '0';
free (env);
if (n >= __libc_argc)
/* Substitute NULL */
return 0;
/* Replace with the appropriate positional parameter */
value = __libc_argv[n];
goto maybe_fieldsplit;
}
/* Is it `$$' ? */
else if (*env == '$')
{
char pidstr[21];
free (env);
pidstr[20] = '\0';
*word = w_addstr (*word, word_length, max_length,
_itoa_word (getpid(), &pidstr[20], 10, 0));
return *word ? 0 : WRDE_NOSPACE;
}
/* Is it `$*' or `$@' (unquoted) ? */
else if (*env == '*' || (*env == '@' && !quoted))
{
size_t plist_len = 1;
int p;
/* Build up value parameter by parameter (copy them) */
free (env);
for (p = 1; __libc_argv[p]; ++p)
{
char *old_pointer = value;
size_t argv_len = strlen (__libc_argv[p]);
size_t old_plist_len = plist_len;
if (value)
value[plist_len - 1] = 0;
plist_len += 1 + argv_len;
/* First realloc will act as malloc because value is
* initialised to NULL. */
value = realloc (value, plist_len);
if (value == NULL)
{
free (old_pointer);
return WRDE_NOSPACE;
}
memcpy (&value[old_plist_len - 1], __libc_argv[p], argv_len + 1);
if (__libc_argv[p + 1])
{
value[plist_len - 1] = '\0';
value[plist_len - 2] = ' ';
}
}
free_value = 1;
if (value)
goto maybe_fieldsplit;
return 0;
}
/* Must be a quoted `$@' */
assert (*env == '@');
assert (quoted);
free (env);
/* Each parameter is a separate word ("$@") */
if (__libc_argv[0] == NULL)
{
/* This can happen if the application is started without any
parameter, not even a name. This is legal according to
POSIX since the giving parameters is only a "should" rule. */
*word = __strdup ("");
*max_length = *word_length = 0;
}
else
{
int p;
for (p = 1; __libc_argv[p + 1]; p++)
{
char *copy = __strdup (__libc_argv[p]);
if (copy == NULL)
return WRDE_NOSPACE;
error = w_addword (pwordexp, copy);
if (error)
{
free (copy);
return error;
}
}
/* Last parameter becomes current word */
if (__libc_argv[p])
{
*word = __strdup (__libc_argv[p]);
if (*word == NULL)
return WRDE_NOSPACE;
*max_length = *word_length = strlen (*word);
}
}
return 0;
}
value = getenv (env);
if (action != '\0' || remove != RP_NONE)
{
switch (action)
{
case 0:
{
char *p;
char c;
char *end;
if (!pattern || !*pattern)
break;
end = value + strlen (value);
if (value == NULL)
break;
switch (remove)
{
case RP_SHORT_LEFT:
for (p = value; p <= end; ++p)
{
c = *p;
*p = '\0';
if (fnmatch (pattern, value, 0) != FNM_NOMATCH)
{
*p = c;
value = p;
break;
}
*p = c;
}
break;
case RP_LONG_LEFT:
for (p = end; p >= value; --p)
{
c = *p;
*p = '\0';
if (fnmatch (pattern, value, 0) != FNM_NOMATCH)
{
*p = c;
value = p;
break;
}
*p = c;
}
break;
case RP_SHORT_RIGHT:
for (p = end; p >= value; --p)
{
if (fnmatch (pattern, p, 0) != FNM_NOMATCH)
{
*p = '\0';
break;
}
}
break;
case RP_LONG_RIGHT:
for (p = value; p <= end; ++p)
{
if (fnmatch (pattern, p, 0) != FNM_NOMATCH)
{
*p = '\0';
break;
}
}
break;
default:
assert (! "Unexpected `remove' value\n");
}
break;
}
case '?':
if (value && *value)
/* Substitute parameter */
break;
if (!colon_seen && value)
{
/* Substitute NULL */
free (env);
free (pattern);
return 0;
}
/* Error - exit */
fprintf (stderr, "%s: ", env);
if (*pattern)
{
/* Expand 'pattern' and write it to stderr */
wordexp_t we;
error = wordexp (pattern, &we, flags);
if (error == 0)
{
int i;
for (i = 0; i < we.we_wordc; ++i)
{
fprintf (stderr, "%s%s", i ? " " : "", we.we_wordv[i]);
}
fprintf (stderr, "\n");
error = WRDE_BADVAL;
}
wordfree (&we);
free (env);
free (pattern);
return error;
}
fprintf (stderr, "parameter null or not set\n");
free (env);
free (pattern);
return WRDE_BADVAL;
case '-':
if (value && *value)
/* Substitute parameter */
break;
if (!colon_seen && value)
{
/* Substitute NULL */
free (env);
free (pattern);
return 0;
}
subst_word:
{
/* Substitute word */
wordexp_t we;
int i;
if (quoted)
{
/* No field-splitting is allowed, so imagine
quotes around the word. */
char *qtd_pattern = malloc (3 + strlen (pattern));
sprintf (qtd_pattern, "\"%s\"", pattern);
free (pattern);
pattern = qtd_pattern;
}
error = wordexp (pattern, &we, flags);
if (error)
{
free (env);
free (pattern);
return error;
}
/* Fingers crossed that the quotes worked.. */
assert (!quoted || we.we_wordc == 1);
/* Substitute */
for (i = 0; i < we.we_wordc; i++)
if (w_addword (pwordexp, __strdup(we.we_wordv[i]))
== WRDE_NOSPACE)
break;
if (i < we.we_wordc)
{
/* Ran out of space */
wordfree (&we);
goto no_space;
}
if (action == '=')
{
char *words;
char *cp;
size_t words_size = 0;
for (i = 0; i < we.we_wordc; i++)
words_size += strlen (we.we_wordv[i]) + 1; /* for <space> */
words_size++;
cp = words = __alloca (words_size);
*words = 0;
for (i = 0; i < we.we_wordc - 1; i++)
{
cp = __stpcpy (cp, we.we_wordv[i]);
*cp++ = ' ';
}
__stpcpy (cp, we.we_wordv[i]);
/* Also assign */
setenv (env, words, 1);
}
wordfree (&we);
return 0;
}
case '+':
if (value && *value)
goto subst_word;
if (!colon_seen && value)
goto subst_word;
/* Substitute NULL */
free (env);
free (pattern);
return 0;
case '=':
if (value && *value)
/* Substitute parameter */
break;
if (!colon_seen && value)
{
/* Substitute NULL */
free (env);
free (pattern);
return 0;
}
/* This checks for '=' so it knows to assign */
goto subst_word;
default:
assert (! "Unrecognised action!");
}
}
free (env);
free (pattern);
if (value == NULL)
{
/* Variable not defined */
if (flags & WRDE_UNDEF)
return WRDE_BADVAL;
return 0;
}
if (seen_hash)
{
char param_length[21];
param_length[20] = '\0';
*word = w_addstr (*word, word_length, max_length,
_itoa_word (strlen (value), &param_length[20], 10, 0));
return *word ? 0 : WRDE_NOSPACE;
}
maybe_fieldsplit:
if (quoted || !pwordexp)
{
/* Quoted - no field split */
*word = w_addstr (*word, word_length, max_length, value);
if (free_value)
free (value);
return *word ? 0 : WRDE_NOSPACE;
}
else
{
/* Need to field-split */
char *value_copy = __strdup (value); /* Don't modify value */
char *field_begin = value_copy;
int seen_nonws_ifs = 0;
if (free_value)
free (value);
if (value_copy == NULL)
return WRDE_NOSPACE;
do
{
char *field_end = field_begin;
char *next_field;
/* If this isn't the first field, start a new word */
if (field_begin != value_copy)
{
if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
{
free (value_copy);
return WRDE_NOSPACE;
}
*word = NULL;
*word_length = *max_length = 0;
}
/* Skip IFS whitespace before the field */
while (*field_begin && strchr (ifs_white, *field_begin) != NULL)
field_begin++;
if (!seen_nonws_ifs && *field_begin == 0)
/* Nothing but whitespace */
break;
/* Search for the end of the field */
field_end = field_begin;
while (*field_end && strchr (ifs, *field_end) == NULL)
field_end++;
/* Set up pointer to the character after end of field */
next_field = *field_end ? field_end : NULL;
/* Skip whitespace IFS after the field */
while (next_field && *next_field && strchr (ifs_white, *next_field))
next_field++;
/* Skip at most one non-whitespace IFS character after the field */
seen_nonws_ifs = 0;
if (next_field && *next_field && strchr (ifs, *next_field))
{
seen_nonws_ifs = 1;
next_field++;
}
/* Null-terminate it */
*field_end = 0;
/* Tag a copy onto the current word */
*word = w_addstr (*word, word_length, max_length, field_begin);
if (*word == NULL)
{
free (value_copy);
return WRDE_NOSPACE;
}
field_begin = next_field;
}
while (seen_nonws_ifs || (field_begin != NULL && *field_begin));
free (value_copy);
}
return 0;
no_space:
if (env)
free (env);
if (pattern)
free (pattern);
return WRDE_NOSPACE;
syntax:
if (env)
free (env);
if (pattern)
free (pattern);
return WRDE_SYNTAX;
}
static int
internal_function
parse_dollars (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags,
wordexp_t *pwordexp, const char *ifs, const char *ifs_white,
int quoted)
{
/* We are poised _at_ "$" */
switch (words[1 + *offset])
{
case '"':
case '\'':
case 0:
*word = w_addchar (*word, word_length, max_length, '$');
return *word ? 0 : WRDE_NOSPACE;
case '(':
if (words[2 + *offset] == '(')
{
/* Differentiate between $((1+3)) and $((echo);(ls)) */
int i = 3 + *offset;
while (words[i] && words[i] != ')')
++i;
if (words[i] == ')' && words[i + 1] == ')')
{
(*offset) += 3;
/* Call parse_arith -- 0 is for "no brackets" */
return parse_arith (word, word_length, max_length, words, offset,
flags, 0);
}
}
if (flags & WRDE_NOCMD)
return WRDE_CMDSUB;
(*offset) += 2;
return parse_comm (word, word_length, max_length, words, offset, flags,
quoted? NULL : pwordexp, ifs, ifs_white);
case '[':
(*offset) += 2;
/* Call parse_arith -- 1 is for "brackets" */
return parse_arith (word, word_length, max_length, words, offset, flags,
1);
case '{':
default:
++(*offset); /* parse_param needs to know if "{" is there */
return parse_param (word, word_length, max_length, words, offset, flags,
pwordexp, ifs, ifs_white, quoted);
}
}
static int
parse_backtick (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags,
wordexp_t *pwordexp, const char *ifs, const char *ifs_white)
{
/* We are poised just after "`" */
int error;
size_t comm_length = 0;
size_t comm_maxlen = 0;
char *comm = NULL;
int squoting = 0;
for (; words[*offset]; ++(*offset))
{
switch (words[*offset])
{
case '`':
/* Go -- give the script to the shell */
error = exec_comm (comm, word, word_length, max_length, flags,
pwordexp, ifs, ifs_white);
free (comm);
return error;
case '\\':
if (squoting)
{
error = parse_qtd_backslash (&comm, &comm_length, &comm_maxlen,
words, offset);
if (error)
{
free (comm);
return error;
}
break;
}
++(*offset);
error = parse_backslash (&comm, &comm_length, &comm_maxlen, words,
offset);
if (error)
{
free (comm);
return error;
}
break;
case '\'':
squoting = 1 - squoting;
default:
comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
if (comm == NULL)
return WRDE_NOSPACE;
}
}
/* Premature end */
free (comm);
return WRDE_SYNTAX;
}
static int
internal_function
parse_dquote (char **word, size_t *word_length, size_t *max_length,
const char *words, size_t *offset, int flags,
wordexp_t *pwordexp, const char * ifs, const char * ifs_white)
{
/* We are poised just after a double-quote */
int error;
for (; words[*offset]; ++(*offset))
{
switch (words[*offset])
{
case '"':
return 0;
case '$':
error = parse_dollars (word, word_length, max_length, words, offset,
flags, pwordexp, ifs, ifs_white, 1);
/* The ``1'' here is to tell parse_dollars not to
* split the fields. It may need to, however ("$@").
*/
if (error)
return error;
break;
case '`':
if (flags & WRDE_NOCMD)
return WRDE_CMDSUB;
++(*offset);
error = parse_backtick (word, word_length, max_length, words,
offset, flags, NULL, NULL, NULL);
/* The first NULL here is to tell parse_backtick not to
* split the fields.
*/
if (error)
return error;
break;
case '\\':
error = parse_qtd_backslash (word, word_length, max_length, words,
offset);
if (error)
return error;
break;
default:
*word = w_addchar (*word, word_length, max_length, words[*offset]);
if (*word == NULL)
return WRDE_NOSPACE;
}
}
/* Unterminated string */
return WRDE_SYNTAX;
}
/*
* wordfree() is to be called after pwordexp is finished with.
*/
void
wordfree (wordexp_t *pwordexp)
{
/* wordexp can set pwordexp to NULL */
if (pwordexp && pwordexp->we_wordv)
{
char **wordv = pwordexp->we_wordv;
for (wordv += pwordexp->we_offs; *wordv; ++wordv)
free (*wordv);
free (pwordexp->we_wordv);
pwordexp->we_wordv = NULL;
}
}
/*
* wordexp()
*/
int
wordexp (const char *words, wordexp_t *pwordexp, int flags)
{
size_t wordv_offset;
size_t words_offset;
size_t word_length = 0;
size_t max_length = 0;
char *word = NULL;
int error;
char *ifs;
char ifs_white[4];
char **old_wordv = pwordexp->we_wordv;
size_t old_wordc = (flags & WRDE_REUSE) ? pwordexp->we_wordc : 0;
if (flags & WRDE_REUSE)
/* Minimal implementation of WRDE_REUSE for now */
wordfree (pwordexp);
if (flags & WRDE_DOOFFS)
{
pwordexp->we_wordv = calloc (1 + pwordexp->we_offs, sizeof (char *));
if (pwordexp->we_wordv == NULL)
return WRDE_NOSPACE;
}
else
{
pwordexp->we_wordv = calloc (1, sizeof (char *));
if (pwordexp->we_wordv == NULL)
return WRDE_NOSPACE;
pwordexp->we_offs = 0;
}
if ((flags & WRDE_APPEND) == 0)
pwordexp->we_wordc = 0;
wordv_offset = pwordexp->we_offs + pwordexp->we_wordc;
/* Find out what the field separators are.
* There are two types: whitespace and non-whitespace.
*/
ifs = getenv ("IFS");
if (!ifs)
/* NULL IFS means no field-splitting is to be performed */
ifs = strcpy (ifs_white, "");
else
{
char *ifsch = ifs;
char *whch = ifs_white;
/* Start off with no whitespace IFS characters */
ifs_white[0] = '\0';
while (*ifsch != '\0')
{
if ((*ifsch == ' ') || (*ifsch == '\t') || (*ifsch == '\n'))
{
/* Whitespace IFS. See first whether it is already in our
collection. */
char *runp = ifs_white;
while (runp < whch && *runp != '\0' && *runp != *ifsch)
++runp;
if (runp == whch)
*whch++ = *ifsch;
}
++ifsch;
}
*whch = '\0';
}
for (words_offset = 0 ; words[words_offset] ; ++words_offset)
switch (words[words_offset])
{
case '\n':
case '|':
case '&':
case ';':
case '<':
case '>':
case '(':
case ')':
case '{':
case '}':
/* Fail */
wordfree (pwordexp);
pwordexp->we_wordc = 0;
pwordexp->we_wordv = old_wordv;
return WRDE_BADCHAR;
case '\\':
error = parse_backslash (&word, &word_length, &max_length, words,
&words_offset);
if (error)
goto do_error;
break;
case '$':
error = parse_dollars (&word, &word_length, &max_length, words,
&words_offset, flags, pwordexp, ifs, ifs_white,
0);
if (error)
goto do_error;
break;
case '`':
if (flags & WRDE_NOCMD)
return WRDE_CMDSUB;
++words_offset;
error = parse_backtick (&word, &word_length, &max_length, words,
&words_offset, flags, pwordexp, ifs,
ifs_white);
if (error)
goto do_error;
break;
case '"':
++words_offset;
error = parse_dquote (&word, &word_length, &max_length, words,
&words_offset, flags, pwordexp, ifs, ifs_white);
if (error)
goto do_error;
break;
case '\'':
++words_offset;
error = parse_squote (&word, &word_length, &max_length, words,
&words_offset);
if (error)
goto do_error;
break;
case '~':
error = parse_tilde (&word, &word_length, &max_length, words,
&words_offset, pwordexp->we_wordc);
if (error)
goto do_error;
break;
case '*':
case '[':
case '?':
error = parse_glob (&word, &word_length, &max_length, words,
&words_offset, flags, pwordexp, ifs, ifs_white);
if (error)
goto do_error;
break;
default:
/* Is it a field separator? */
if (strchr (ifs, words[words_offset]) == NULL)
{
/* "Ordinary" character -- add it to word */
word = w_addchar (word, &word_length, &max_length,
words[words_offset]);
if (word == NULL)
{
error = WRDE_NOSPACE;
goto do_error;
}
break;
}
/* Field separator */
if (strchr (ifs_white, words[words_offset]))
{
/* It's a whitespace IFS char. Ignore it at the beginning
of a line and ignore multiple instances. */
if (!word || !*word)
break;
if (w_addword (pwordexp, word) == WRDE_NOSPACE)
{
error = WRDE_NOSPACE;
goto do_error;
}
word = NULL;
word_length = 0;
max_length = 0;
break;
}
/* It's a non-whitespace IFS char */
/* Multiple non-whitespace IFS chars are treated as one;
* IS THIS CORRECT?
*/
if (word != NULL)
{
if (w_addword (pwordexp, word) == WRDE_NOSPACE)
{
error = WRDE_NOSPACE;
goto do_error;
}
}
word = NULL;
word_length = 0;
max_length = 0;
}
/* End of string */
/* There was a field separator at the end */
if (word == NULL)
return 0;
/* There was no field separator at the end */
return w_addword (pwordexp, word);
do_error:
/* Error:
* free memory used (unless error is WRDE_NOSPACE), and
* set we_wordc and wd_wordv back to what they were.
*/
if (error == WRDE_NOSPACE)
return WRDE_NOSPACE;
if (word != NULL)
free (word);
wordfree (pwordexp);
pwordexp->we_wordv = old_wordv;
pwordexp->we_wordc = old_wordc;
return error;
}