From a8125d855218832508af812df585c5fac5093bc3 Mon Sep 17 00:00:00 2001 From: Ulrich Drepper Date: Tue, 15 Sep 1998 21:24:37 +0000 Subject: [PATCH] Update 1998-09-16 00:47 Tim Waugh * posix/wordexp-test.c: If expansion or substitution occurs anywhere in a word, the entire word is subject to field-splitting. * posix/wordexp.c (parse_glob): Look for end of word instead of end of field when deciding what to glob. (field_split_word): New function, now the only place where field-splitting is performed. (parse_dollars): New parameter - tell the caller if field-splitting should be performed on this word. * posix/wordexp-test.c (testit): Only call wordfree if wordexp succeeded (or failed with WRDE_NOSPACE). 1998-09-15 19:53 1998 Tim Waugh * posix/wordexp.c (wordexp): Don't convert IFS characters to blanks. * posix/wordexp-test.c: Words not the result of expansion or substitution should remain unchanged. --- ChangeLog | 23 ++++ posix/wordexp-test.c | 10 +- posix/wordexp.c | 348 ++++++++++++++++++++++----------------------------- 3 files changed, 181 insertions(+), 200 deletions(-) diff --git a/ChangeLog b/ChangeLog index 4893627..ab24a49 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,26 @@ +1998-09-16 00:47 Tim Waugh + + * posix/wordexp-test.c: If expansion or substitution occurs + anywhere in a word, the entire word is subject to field-splitting. + + * posix/wordexp.c (parse_glob): Look for end of word instead of + end of field when deciding what to glob. + (field_split_word): New function, now the only place where + field-splitting is performed. + (parse_dollars): New parameter - tell the caller if + field-splitting should be performed on this word. + + * posix/wordexp-test.c (testit): Only call wordfree if wordexp + succeeded (or failed with WRDE_NOSPACE). + +1998-09-15 19:53 1998 Tim Waugh + + * posix/wordexp.c (wordexp): Don't convert IFS characters to + blanks. + + * posix/wordexp-test.c: Words not the result of expansion or + substitution should remain unchanged. + 1998-09-15 Ulrich Drepper * elf/dynamic-link.h [!ELF_MACHINE_PLTREL_OVERLAP] diff --git a/posix/wordexp-test.c b/posix/wordexp-test.c index be81c4b..746b817 100644 --- a/posix/wordexp-test.c +++ b/posix/wordexp-test.c @@ -43,7 +43,7 @@ struct test_case_struct { 0, NULL, "one two", 0, 2, { "one", "two", }, IFS }, { 0, NULL, "one two three", 0, 3, { "one", "two", "three", }, IFS }, { 0, NULL, " \tfoo\t\tbar ", 0, 2, { "foo", "bar", }, IFS }, - { 0, NULL, "red , white blue", 0, 4, { "red", " ", "white", "blue", }, " ," }, + { 0, NULL, "red , white blue", 0, 4, { "red", ",", "white", "blue", }, " ," }, { 0, NULL, "one two three", 0, 3, { "one", "two", "three", }, "" }, { 0, NULL, "one \"two three\"", 0, 2, { "one", "two three", }, IFS }, { 0, NULL, "one \"two three\"", 0, 2, { "one", "two three", }, "" }, @@ -60,10 +60,11 @@ struct test_case_struct { 0, NULL, "$(echo :abc:\\ )", 0, 2, { "", "abc", }, ": " }, { 0, NULL, "$(echo :abc\\ )", 0, 2, { "", "abc", }, ": " }, { 0, ":abc:", "$(echo $var)", 0, 2, { "", "abc", }, ":" }, - { 0, NULL, ":abc:", 0, 1, { " abc ", }, ":" }, + { 0, NULL, ":abc:", 0, 1, { ":abc:", }, ":" }, { 0, NULL, "$(echo :abc:)def", 0, 3, { "", "abc", "def", }, ":" }, { 0, NULL, "$(echo abc:de)f", 0, 2, { "abc", "def", }, ":" }, - { 0, NULL, "$(echo abc:de)f:ghi", 0, 2, { "abc", "def ghi", }, ":" }, + { 0, NULL, "$(echo abc:de)f:ghi", 0, 3, { "abc", "def", "ghi", }, ":" }, + { 0, NULL, "abc:d$(echo ef:ghi)", 0, 3, { "abc", "def", "ghi", }, ":" }, { 0, "abc:", "$var$(echo def:ghi)", 0, 3, { "abc", "def", "ghi", }, ":" }, { 0, "abc:d", "$var$(echo ef:ghi)", 0, 3, { "abc", "def", "ghi", }, ":" }, { 0, "def:ghi", "$(echo abc:)$var", 0, 3, { "abc", "def", "ghi", }, ":" }, @@ -308,7 +309,8 @@ testit (struct test_case_struct *tc) else printf ("OK\n"); - wordfree (&we); + if (retval == 0 || retval == WRDE_NOSPACE) + wordfree (&we); return bzzzt; } diff --git a/posix/wordexp.c b/posix/wordexp.c index 0139a99..e4bf414 100644 --- a/posix/wordexp.c +++ b/posix/wordexp.c @@ -56,7 +56,7 @@ extern char **__libc_argv; 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) + const char *ifs_white, int quoted, int *fsplit) internal_function; static int parse_backtick (char **word, size_t *word_length, size_t *max_length, const char *words, @@ -183,6 +183,78 @@ no_space: return WRDE_NOSPACE; } +static int +internal_function +field_split_word (char *word, wordexp_t *pwordexp, const char *ifs, + const char *ifs_white) +{ + size_t field_length; + size_t field_maxlen; + char *field = w_newword (&field_length, &field_maxlen); + char *field_begin = word; + int seen_nonws_ifs = 0; + + if (!word) + return 0; + + do + { + char *field_end = field_begin; + char *next_field; + + /* If this isn't the first field, start a new word */ + if (field_begin != word) + { + if (w_addword (pwordexp, field) == WRDE_NOSPACE) + goto no_space; + + field = w_newword (&field_length, &field_maxlen); + } + + /* Skip IFS whitespace before the field */ + field_begin += strspn (field_begin, ifs_white); + + if (!seen_nonws_ifs && *field_begin == 0) + /* Nothing but whitespace */ + break; + + /* Search for the end of the field */ + field_end = field_begin + strcspn (field_begin, ifs); + + /* Set up pointer to the character after end of field and + skip whitespace IFS after it. */ + next_field = field_end + strspn (field_end, ifs_white); + + /* Skip at most one non-whitespace IFS character after the field */ + seen_nonws_ifs = 0; + if (*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 */ + field = w_addstr (field, &field_length, &field_maxlen, field_begin); + + if (field == NULL && *field_begin != '\0') + goto no_space; + + field_begin = next_field; + } + while (seen_nonws_ifs || *field_begin); + + if (field && w_addword (pwordexp, field)) + goto no_space; + + return 0; + +no_space: + return WRDE_NOSPACE; +} + /* The parse_*() functions should leave *offset being the offset in 'words' * to the last character processed. */ @@ -424,15 +496,15 @@ parse_glob (char **word, size_t *word_length, size_t *max_length, int quoted = 0; /* 1 if singly-quoted, 2 if doubly */ int i; wordexp_t glob_list; /* List of words to glob */ + int fieldsplit = 0; glob_list.we_wordc = 0; glob_list.we_wordv = NULL; glob_list.we_offs = 0; for (; words[*offset] != '\0'; ++*offset) { - if ((ifs && strchr (ifs, words[*offset])) || - (!ifs && strchr (" \t\n", words[*offset]))) - /* Reached IFS */ + if (strchr (" \t\n", words[*offset])) + /* Reached end of word */ break; /* Sort out quoting */ @@ -468,7 +540,7 @@ parse_glob (char **word, size_t *word_length, size_t *max_length, { error = parse_dollars (word, word_length, max_length, words, offset, flags, &glob_list, ifs, ifs_white, - quoted == 2); + quoted == 2, &fieldsplit); if (error) goto tidy_up; @@ -497,8 +569,16 @@ parse_glob (char **word, size_t *word_length, size_t *max_length, /* Don't forget to re-parse the character we stopped at. */ --*offset; + if (fieldsplit) + { + error = field_split_word (*word, &glob_list, ifs, ifs_white); + if (*word) + free (*word); + } + else + error = w_addword (&glob_list, *word); + /* Glob the words */ - error = w_addword (&glob_list, *word); *word = w_newword (word_length, max_length); for (i = 0; error == 0 && i < glob_list.we_wordc; i++) error = do_parse_glob (glob_list.we_wordv[i], word, word_length, @@ -676,7 +756,8 @@ parse_arith (char **word, size_t *word_length, size_t *max_length, { case '$': error = parse_dollars (&expr, &expr_length, &expr_maxlen, - words, offset, flags, NULL, NULL, NULL, 1); + words, offset, flags, NULL, NULL, NULL, 1, + NULL); /* The ``1'' here is to tell parse_dollars not to * split the fields. */ @@ -871,100 +952,20 @@ exec_comm (char *comm, char **word, size_t *word_length, size_t *max_length, __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) - goto no_space; - } - } - else - /* Not quoted - split fields */ + /* Read fildes[0] and put it into a word. */ + while (1) { - 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 ((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: textmoretext - * - * 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-ws IFS afterwards) */ - copying = 2; - } - - /* First IFS white space, or IFS non-whitespace. - * Delimit the field. Nulls are converted by w_addword. */ - if (w_addword (pwordexp, *word) == WRDE_NOSPACE) - goto no_space; - - *word = w_newword (word_length, max_length); - /* fall back round the loop.. */ - } - else - { - /* Not IFS character */ - copying = 1; - - *word = w_addchar (*word, word_length, max_length, - buffer[i]); - if (*word == NULL) - goto no_space; - } - } + break; } + + *word = w_addmem (*word, word_length, max_length, buffer, buflen); + if (*word == NULL) + goto no_space; } /* Bash chops off trailing newlines, which seems sensible. */ @@ -1701,87 +1702,11 @@ envsubst: if (value == NULL) return 0; - 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) - goto no_space; - - 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); - goto no_space; - } - - *word = w_newword (word_length, max_length); - } - - /* Skip IFS whitespace before the field */ - field_begin += strspn (field_begin, ifs_white); - - if (!seen_nonws_ifs && *field_begin == 0) - /* Nothing but whitespace */ - break; - - /* Search for the end of the field */ - field_end = field_begin + strcspn (field_begin, ifs); - - /* Set up pointer to the character after end of field and - skip whitespace IFS after it. */ - next_field = field_end + strspn (field_end, ifs_white); - - /* Skip at most one non-whitespace IFS character after the field */ - seen_nonws_ifs = 0; - if (*next_field && strchr (ifs, *next_field)) - { - seen_nonws_ifs = 1; - next_field++; - } - - /* Null-terminate it */ - *field_end = 0; + *word = w_addstr (*word, word_length, max_length, value); + if (free_value) + free (value); - /* Tag a copy onto the current word */ - *word = w_addstr (*word, word_length, max_length, field_begin); - - if (*word == NULL && *field_begin != '\0') - { - free (value_copy); - goto no_space; - } - - field_begin = next_field; - } - while (seen_nonws_ifs || *field_begin); - - free (value_copy); - } - - return 0; + return *word ? 0 : WRDE_NOSPACE; success: error = 0; @@ -1809,7 +1734,7 @@ 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) + int quoted, int *fsplit) { /* We are poised _at_ "$" */ switch (words[1 + *offset]) @@ -1839,6 +1764,12 @@ parse_dollars (char **word, size_t *word_length, size_t *max_length, if (words[i] == ')' && words[i + 1] == ')') { (*offset) += 3; + + /* This word is subject to field-splitting as long as + * it isn't quoted. */ + if (fsplit) + *fsplit = !quoted; + /* Call parse_arith -- 0 is for "no brackets" */ return parse_arith (word, word_length, max_length, words, offset, flags, 0); @@ -1849,11 +1780,23 @@ parse_dollars (char **word, size_t *word_length, size_t *max_length, return WRDE_CMDSUB; (*offset) += 2; + + /* This word is subject to field-splitting as long as + * it isn't quoted. */ + if (fsplit) + *fsplit = !quoted; + return parse_comm (word, word_length, max_length, words, offset, flags, quoted? NULL : pwordexp, ifs, ifs_white); case '[': (*offset) += 2; + + /* This word is subject to field-splitting as long as + * it isn't quoted. */ + if (fsplit) + *fsplit = !quoted; + /* Call parse_arith -- 1 is for "brackets" */ return parse_arith (word, word_length, max_length, words, offset, flags, 1); @@ -1861,6 +1804,12 @@ parse_dollars (char **word, size_t *word_length, size_t *max_length, case '{': default: ++(*offset); /* parse_param needs to know if "{" is there */ + + /* This word is subject to field-splitting as long as + * it isn't quoted. */ + if (fsplit) + *fsplit = !quoted; + return parse_param (word, word_length, max_length, words, offset, flags, pwordexp, ifs, ifs_white, quoted); } @@ -1948,7 +1897,7 @@ parse_dquote (char **word, size_t *word_length, size_t *max_length, case '$': error = parse_dollars (word, word_length, max_length, words, offset, - flags, pwordexp, ifs, ifs_white, 1); + flags, pwordexp, ifs, ifs_white, 1, NULL); /* The ``1'' here is to tell parse_dollars not to * split the fields. It may need to, however ("$@"). */ @@ -2030,6 +1979,7 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) char ifs_white[4]; char **old_wordv = pwordexp->we_wordv; size_t old_wordc = (flags & WRDE_REUSE) ? pwordexp->we_wordc : 0; + int fieldsplit_this_word = 0; if (flags & WRDE_REUSE) { @@ -2100,7 +2050,8 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) *whch = '\0'; } - for (words_offset = 0 ; words[words_offset] ; ++words_offset) + fieldsplit_this_word = 0; + for (words_offset = 0 ; ; ++words_offset) switch (words[words_offset]) { case '\\': @@ -2115,7 +2066,7 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) case '$': error = parse_dollars (&word, &word_length, &max_length, words, &words_offset, flags, pwordexp, ifs, ifs_white, - 0); + 0, &fieldsplit_this_word); if (error) goto do_error; @@ -2137,6 +2088,8 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) if (error) goto do_error; + fieldsplit_this_word = 1; + break; case '"': @@ -2181,7 +2134,8 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) default: /* Is it a word separator? */ - if (strchr (" \t", words[words_offset]) == NULL) + if (words[words_offset] != '\0' && + strchr (" \t", words[words_offset]) == NULL) { char ch = words[words_offset]; @@ -2196,11 +2150,6 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) } /* "Ordinary" character -- add it to word */ - - /* Convert IFS chars to blanks -- bash does this */ - if (strchr (ifs, ch)) - ch = ' '; - word = w_addchar (word, &word_length, &max_length, ch); if (word == NULL) @@ -2215,22 +2164,29 @@ wordexp (const char *words, wordexp_t *pwordexp, int flags) /* If a word has been delimited, add it to the list. */ if (word != NULL) { - error = w_addword (pwordexp, word); + if (fieldsplit_this_word) + { + error = field_split_word (word, pwordexp, ifs, ifs_white); + free (word); + } + else + error = w_addword (pwordexp, word); + if (error) goto do_error; } - word = w_newword (&word_length, &max_length); - } + fieldsplit_this_word = 0; - /* End of string */ + if (words[words_offset] == '\0') + /* End of string. */ + goto end_of_string; - /* There was a word separator at the end */ - if (word == NULL) /* i.e. w_newword */ - return 0; + word = w_newword (&word_length, &max_length); + } - /* There was no field separator at the end */ - return w_addword (pwordexp, word); +end_of_string: + return 0; do_error: /* Error: -- 2.7.4