X-Git-Url: https://git.dogcows.com/gitweb?a=blobdiff_plain;f=src%2Fnames.c;h=eaa94d284aa7257c08970ff877155b1e3542e9aa;hb=1bcbbcf1ff2c537ffa970dbf82e3843d4ad110e5;hp=2b24231efe4571abc341a41405ae7a213afed910;hpb=b794387cbaef6374a28e6ffe709e319e0a0647f0;p=chaz%2Ftar diff --git a/src/names.c b/src/names.c index 2b24231..eaa94d2 100644 --- a/src/names.c +++ b/src/names.c @@ -1,11 +1,11 @@ /* Various processing of names. Copyright (C) 1988, 1992, 1994, 1996, 1997, 1998, 1999, 2000, 2001, - 2003, 2004, 2005 Free Software Foundation, Inc. + 2003, 2004, 2005, 2006, 2007, 2009 Free Software Foundation, Inc. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the - Free Software Foundation; either version 2, or (at your option) any later + Free Software Foundation; either version 3, or (at your option) any later version. This program is distributed in the hope that it will be useful, but @@ -178,54 +178,118 @@ gname_to_gid (char const *gname, gid_t *gidp) return 1; } + +struct name * +make_name (const char *file_name) +{ + struct name *p = xzalloc (sizeof (*p)); + if (!file_name) + file_name = ""; + p->name = xstrdup (file_name); + p->length = strlen (p->name); + return p; +} + +void +free_name (struct name *p) +{ + if (p) + { + free (p->name); + free (p->caname); + free (p); + } +} + /* Names from the command call. */ static struct name *namelist; /* first name in list, if any */ static struct name **nametail = &namelist; /* end of name list */ -static const char **name_array; /* store an array of names */ -static int allocated_names; /* how big is the array? */ -static int names; /* how many entries does it have? */ -static int name_index; /* how many of the entries have we scanned? */ -/* Initialize structures. */ -void -init_names (void) +/* File name arguments are processed in two stages: first a + name_array (see below) is filled, then the names from it + are moved into the namelist. + + This awkward process is needed only to implement --same-order option, + which is meant to help process large archives on machines with + limited memory. With this option on, namelist contains at most one + entry, which diminishes the memory consumption. + + However, I very much doubt if we still need this -- Sergey */ + +/* A name_array element contains entries of three types: */ + +#define NELT_NAME 0 /* File name */ +#define NELT_CHDIR 1 /* Change directory request */ +#define NELT_FMASK 2 /* Change fnmatch options request */ + +struct name_elt /* A name_array element. */ +{ + char type; /* Element type, see NELT_* constants above */ + union + { + const char *name; /* File or directory name */ + int matching_flags;/* fnmatch options if type == NELT_FMASK */ + } v; +}; + +static struct name_elt *name_array; /* store an array of names */ +static size_t allocated_names; /* how big is the array? */ +static size_t names; /* how many entries does it have? */ +static size_t name_index; /* how many of the entries have we scanned? */ + +/* Check the size of name_array, reallocating it as necessary. */ +static void +check_name_alloc () { - allocated_names = 10; - name_array = xmalloc (sizeof (const char *) * allocated_names); - names = 0; + if (names == allocated_names) + { + if (allocated_names == 0) + allocated_names = 10; /* Set initial allocation */ + name_array = x2nrealloc (name_array, &allocated_names, + sizeof (name_array[0])); + } } -/* Add NAME at end of name_array, reallocating it as necessary. */ +/* Add to name_array the file NAME with fnmatch options MATCHING_FLAGS */ void -name_add (const char *name) +name_add_name (const char *name, int matching_flags) { - if (names == allocated_names) + static int prev_flags = 0; /* FIXME: Or EXCLUDE_ANCHORED? */ + struct name_elt *ep; + + check_name_alloc (); + ep = &name_array[names++]; + if (prev_flags != matching_flags) { - allocated_names *= 2; - name_array = - xrealloc (name_array, sizeof (const char *) * allocated_names); + ep->type = NELT_FMASK; + ep->v.matching_flags = matching_flags; + prev_flags = matching_flags; + check_name_alloc (); + ep = &name_array[names++]; } - name_array[names++] = name; + ep->type = NELT_NAME; + ep->v.name = name; } +/* Add to name_array a chdir request for the directory NAME */ +void +name_add_dir (const char *name) +{ + struct name_elt *ep; + check_name_alloc (); + ep = &name_array[names++]; + ep->type = NELT_CHDIR; + ep->v.name = name; +} + /* Names from external name file. */ static char *name_buffer; /* buffer to hold the current file name */ static size_t name_buffer_length; /* allocated length of name_buffer */ -/* FIXME: I should better check more closely. It seems at first glance that - is_pattern is only used when reading a file, and ignored for all - command line arguments. */ - -static inline int -is_pattern (const char *string) -{ - return strchr (string, '*') || strchr (string, '[') || strchr (string, '?'); -} - /* Set up to gather file names for tar. They can either come from a file or were saved from decoding arguments. */ void @@ -242,27 +306,36 @@ name_term (void) free (name_array); } -/* Get the next name from ARGV or the file of names. Result is in +static int matching_flags; /* exclude_fnmatch options */ + +/* Get the next NELT_NAME element from name_array. Result is in static storage and can't be relied upon across two calls. - If CHANGE_DIRS is true, treat a filename of the form "-C" as - meaning that the next filename is the name of a directory to change - to. If filename_terminator is NUL, CHANGE_DIRS is effectively - always false. */ -char * -name_next (int change_dirs) + If CHANGE_DIRS is true, treat any entries of type NELT_CHDIR as + the request to change to the given directory. + + Entries of type NELT_FMASK cause updates of the matching_flags + value. */ +struct name_elt * +name_next_elt (int change_dirs) { + static struct name_elt entry; const char *source; char *cursor; - int chdir_flag = 0; - - if (filename_terminator == '\0') - change_dirs = 0; while (name_index != names) { + struct name_elt *ep; size_t source_len; - source = name_array[name_index++]; + + ep = &name_array[name_index++]; + if (ep->type == NELT_FMASK) + { + matching_flags = ep->v.matching_flags; + continue; + } + + source = ep->v.name; source_len = strlen (source); if (name_buffer_length < source_len) { @@ -285,25 +358,31 @@ name_next (int change_dirs) while (cursor > name_buffer && ISSLASH (*cursor)) *cursor-- = '\0'; - if (chdir_flag) + if (change_dirs && ep->type == NELT_CHDIR) { if (chdir (name_buffer) < 0) chdir_fatal (name_buffer); - chdir_flag = 0; } - else if (change_dirs && strcmp (name_buffer, "-C") == 0) - chdir_flag = 1; else { if (unquote_option) unquote_string (name_buffer); if (incremental_option) register_individual_file (name_buffer); - return name_buffer; + entry.type = ep->type; + entry.v.name = name_buffer; + return &entry; } } - return 0; + return NULL; +} + +const char * +name_next (int change_dirs) +{ + struct name_elt *nelt = name_next_elt (change_dirs); + return nelt ? nelt->v.name : NULL; } /* Gather names in a list for scanning. Could hash them later if we @@ -320,58 +399,33 @@ void name_gather (void) { /* Buffer able to hold a single name. */ - static struct name *buffer; - static size_t allocated_size; + static struct name *buffer = NULL; - char const *name; + struct name_elt *ep; if (same_order_option) { static int change_dir; - if (allocated_size == 0) - { - allocated_size = offsetof (struct name, name) + NAME_FIELD_SIZE + 1; - buffer = xmalloc (allocated_size); - /* FIXME: This memset is overkill, and ugly... */ - memset (buffer, 0, allocated_size); - } - - while ((name = name_next (0)) && strcmp (name, "-C") == 0) - { - char const *dir = name_next (0); - if (! dir) - FATAL_ERROR ((0, 0, _("Missing file name after -C"))); - change_dir = chdir_arg (xstrdup (dir)); - } + while ((ep = name_next_elt (0)) && ep->type == NELT_CHDIR) + change_dir = chdir_arg (xstrdup (ep->v.name)); - if (name) + if (ep) { - size_t needed_size; - buffer->length = strlen (name); - needed_size = offsetof (struct name, name) + buffer->length + 1; - if (allocated_size < needed_size) - { - do - { - allocated_size *= 2; - if (! allocated_size) - xalloc_die (); - } - while (allocated_size < needed_size); - - buffer = xrealloc (buffer, allocated_size); - } + free_name (buffer); + buffer = make_name (ep->v.name); buffer->change_dir = change_dir; - strcpy (buffer->name, name); buffer->next = 0; buffer->found_count = 0; - + buffer->matching_flags = matching_flags; + buffer->dir_contents = NULL; + buffer->parent = NULL; + namelist = buffer; nametail = &namelist->next; } else if (change_dir) - addname (0, change_dir); + addname (0, change_dir, NULL); } else { @@ -381,19 +435,15 @@ name_gather (void) for (;;) { int change_dir0 = change_dir; - while ((name = name_next (0)) && strcmp (name, "-C") == 0) - { - char const *dir = name_next (0); - if (! dir) - FATAL_ERROR ((0, 0, _("Missing file name after -C"))); - change_dir = chdir_arg (xstrdup (dir)); - } - if (name) - addname (name, change_dir); + while ((ep = name_next_elt (0)) && ep->type == NELT_CHDIR) + change_dir = chdir_arg (xstrdup (ep->v.name)); + + if (ep) + addname (ep->v.name, change_dir, NULL); else { if (change_dir != change_dir0) - addname (0, change_dir); + addname (0, change_dir, NULL); break; } } @@ -402,42 +452,18 @@ name_gather (void) /* Add a name to the namelist. */ struct name * -addname (char const *string, int change_dir) +addname (char const *string, int change_dir, struct name *parent) { - size_t length = string ? strlen (string) : 0; - struct name *name = xmalloc (offsetof (struct name, name) + length + 1); - - if (string) - { - name->fake = 0; - strcpy (name->name, string); - } - else - { - name->fake = 1; - - /* FIXME: This initialization (and the byte of memory that it - initializes) is probably not needed, but we are currently in - bug-fix mode so we'll leave it in for now. */ - name->name[0] = 0; - } + struct name *name = make_name (string); - name->next = 0; - name->length = length; + name->prev = *nametail; + name->next = NULL; name->found_count = 0; - name->regexp = 0; /* assume not a regular expression */ - name->firstch = 1; /* assume first char is literal */ + name->matching_flags = matching_flags; name->change_dir = change_dir; - name->dir_contents = 0; - name->explicit = 1; + name->dir_contents = NULL; + name->parent = parent; - if (string && is_pattern (string)) - { - name->regexp = 1; - if (string[0] == '*' || string[0] == '[' || string[0] == '?') - name->firstch = 0; - } - *nametail = name; nametail = &name->next; return name; @@ -452,26 +478,33 @@ namelist_match (char const *file_name, size_t length) for (p = namelist; p; p = p->next) { - /* If first chars don't match, quick skip. */ - - if (p->firstch && p->name[0] != file_name[0]) - continue; - - if (p->regexp - ? fnmatch (p->name, file_name, recursion_option) == 0 - : (p->length <= length - && (file_name[p->length] == '\0' - || (ISSLASH (file_name[p->length]) && recursion_option)) - && memcmp (file_name, p->name, p->length) == 0)) + if (p->name[0] + && exclude_fnmatch (p->name, file_name, p->matching_flags)) return p; } - return 0; + return NULL; +} + +void +remname (struct name *name) +{ + struct name *p; + + if ((p = name->prev) != NULL) + p->next = name->next; + else + namelist = name->next; + + if ((p = name->next) != NULL) + p->prev = name->prev; + else + nametail = &name->prev; } /* Return true if and only if name FILE_NAME (from an archive) matches any name from the namelist. */ -int +bool name_match (const char *file_name) { size_t length = strlen (file_name); @@ -481,14 +514,14 @@ name_match (const char *file_name) struct name *cursor = namelist; if (!cursor) - return 1; - - if (cursor->fake) + return true; + + if (cursor->name[0] == 0) { chdir_do (cursor->change_dir); namelist = 0; nametail = &namelist; - return 1; + return true; } cursor = namelist_match (file_name, length); @@ -518,10 +551,10 @@ name_match (const char *file_name) { name_gather (); /* read one more */ if (namelist->found_count) - return 0; + return false; } else - return 0; + return false; } } @@ -547,14 +580,38 @@ all_names_found (struct tar_stat_info *p) len = strlen (p->file_name); for (cursor = namelist; cursor; cursor = cursor->next) { - if (cursor->regexp - || (!WASFOUND(cursor) && !cursor->fake) + if ((cursor->name[0] && !WASFOUND (cursor)) || (len >= cursor->length && ISSLASH (p->file_name[cursor->length]))) return false; } return true; } +static inline int +is_pattern (const char *string) +{ + return strchr (string, '*') || strchr (string, '[') || strchr (string, '?'); +} + +static void +regex_usage_warning (const char *name) +{ + static int warned_once = 0; + + if (warn_regex_usage && is_pattern (name)) + { + warned_once = 1; + WARN ((0, 0, + /* TRANSLATORS: The following three msgids form a single sentence. + */ + _("Pattern matching characters used in file names. Please,"))); + WARN ((0, 0, + _("use --wildcards to enable pattern matching, or --no-wildcards to"))); + WARN ((0, 0, + _("suppress this warning."))); + } +} + /* Print the names of things in the namelist that were not matched. */ void names_notfound (void) @@ -562,14 +619,15 @@ names_notfound (void) struct name const *cursor; for (cursor = namelist; cursor; cursor = cursor->next) - if (!WASFOUND(cursor) && !cursor->fake) + if (!WASFOUND (cursor) && cursor->name[0]) { + regex_usage_warning (cursor->name); if (cursor->found_count == 0) ERROR ((0, 0, _("%s: Not found in archive"), quotearg_colon (cursor->name))); else ERROR ((0, 0, _("%s: Required occurrence not found in archive"), - quotearg_colon (cursor->name))); + quotearg_colon (cursor->name))); } /* Don't bother freeing the name list; we're about to exit. */ @@ -578,25 +636,31 @@ names_notfound (void) if (same_order_option) { - char *name; + const char *name; while ((name = name_next (1)) != NULL) - ERROR ((0, 0, _("%s: Not found in archive"), - quotearg_colon (name))); + { + regex_usage_warning (name); + ERROR ((0, 0, _("%s: Not found in archive"), + quotearg_colon (name))); + } } } /* Sorting name lists. */ -/* Sort linked LIST of names, of given LENGTH, using COMPARE to order - names. Return the sorted list. Apart from the type `struct name' - and the definition of SUCCESSOR, this is a generic list-sorting - function, but it's too painful to make it both generic and portable +/* Sort *singly* linked LIST of names, of given LENGTH, using COMPARE + to order names. Return the sorted list. Note that after calling + this function, the `prev' links in list elements are messed up. + + Apart from the type `struct name' and the definition of SUCCESSOR, + this is a generic list-sorting function, but it's too painful to + make it both generic and portable in C. */ static struct name * -merge_sort (struct name *list, int length, - int (*compare) (struct name const*, struct name const*)) +merge_sort_sll (struct name *list, int length, + int (*compare) (struct name const*, struct name const*)) { struct name *first_list; struct name *second_list; @@ -634,8 +698,8 @@ merge_sort (struct name *list, int length, second_list = SUCCESSOR (cursor); SUCCESSOR (cursor) = 0; - first_list = merge_sort (first_list, first_length, compare); - second_list = merge_sort (second_list, second_length, compare); + first_list = merge_sort_sll (first_list, first_length, compare); + second_list = merge_sort_sll (second_list, second_length, compare); merge_point = &result; while (first_list && second_list) @@ -663,37 +727,61 @@ merge_sort (struct name *list, int length, #undef SUCCESSOR } +/* Sort doubly linked LIST of names, of given LENGTH, using COMPARE + to order names. Return the sorted list. */ +static struct name * +merge_sort (struct name *list, int length, + int (*compare) (struct name const*, struct name const*)) +{ + struct name *head, *p, *prev; + head = merge_sort_sll (list, length, compare); + /* Fixup prev pointers */ + for (prev = NULL, p = head; p; prev = p, p = p->next) + p->prev = prev; + return head; +} + /* A comparison function for sorting names. Put found names last; break ties by string comparison. */ static int -compare_names (struct name const *n1, struct name const *n2) +compare_names_found (struct name const *n1, struct name const *n2) { - int found_diff = WASFOUND(n2) - WASFOUND(n1); + int found_diff = WASFOUND (n2) - WASFOUND (n1); return found_diff ? found_diff : strcmp (n1->name, n2->name); } + +/* Simple comparison by names. */ +static int +compare_names (struct name const *n1, struct name const *n2) +{ + return strcmp (n1->name, n2->name); +} + /* Add all the dirs under NAME, which names a directory, to the namelist. If any of the files is a directory, recurse on the subdirectory. - DEVICE is the device not to leave, if the -l option is specified. */ + DEVICE is the device not to leave, if the -l option is specified. + CMDLINE is true, if the NAME appeared on the command line. */ static void -add_hierarchy_to_namelist (struct name *name, dev_t device) +add_hierarchy_to_namelist (struct name *name, dev_t device, bool cmdline) { char *file_name = name->name; - char *buffer = get_directory_contents (file_name, device); - + const char *buffer = scan_directory (file_name, device, cmdline); + if (! buffer) name->dir_contents = "\0\0\0\0"; else { + struct name *child_head = NULL, *child_tail = NULL; size_t name_length = name->length; size_t allocated_length = (name_length >= NAME_FIELD_SIZE ? name_length + NAME_FIELD_SIZE : NAME_FIELD_SIZE); char *namebuf = xmalloc (allocated_length + 1); /* FIXME: + 2 above? */ - char *string; + const char *string; size_t string_length; int change_dir = name->change_dir; @@ -711,7 +799,7 @@ add_hierarchy_to_namelist (struct name *name, dev_t device) if (*string == 'D') { struct name *np; - + if (allocated_length <= name_length + string_length) { do @@ -725,16 +813,64 @@ add_hierarchy_to_namelist (struct name *name, dev_t device) namebuf = xrealloc (namebuf, allocated_length + 1); } strcpy (namebuf + name_length, string + 1); - np = addname (namebuf, change_dir); - np->explicit = 0; - add_hierarchy_to_namelist (np, device); + np = addname (namebuf, change_dir, name); + if (!child_head) + child_head = np; + else + child_tail->sibling = np; + child_tail = np; + add_hierarchy_to_namelist (np, device, false); } } free (namebuf); + name->child = child_head; } } +/* Auxiliary functions for hashed table of struct name's. */ + +static size_t +name_hash (void const *entry, size_t n_buckets) +{ + struct name const *name = entry; + return hash_string (name->caname, n_buckets); +} + +/* Compare two directories for equality of their names. */ +static bool +name_compare (void const *entry1, void const *entry2) +{ + struct name const *name1 = entry1; + struct name const *name2 = entry2; + return strcmp (name1->caname, name2->caname) == 0; +} + + +/* Rebase `name' member of CHILD and all its siblings to + the new PARENT. */ +static void +rebase_child_list (struct name *child, struct name *parent) +{ + size_t old_prefix_len = child->parent->length; + size_t new_prefix_len = parent->length; + char *new_prefix = parent->name; + + for (; child; child = child->sibling) + { + size_t size = child->length - old_prefix_len + new_prefix_len; + char *newp = xmalloc (size + 1); + strcpy (newp, new_prefix); + strcat (newp, child->name + old_prefix_len); + free (child->name); + child->name = newp; + child->length = size; + + rebase_directory (child->name, old_prefix_len, child->parent->name, + new_prefix); + } +} + /* Collect all the names from argv[] (or whatever), expand them into a directory tree, and sort them. This gets only subdirectories, not all files. */ @@ -743,27 +879,51 @@ void collect_and_sort_names (void) { struct name *name; - struct name *next_name; + struct name *next_name, *prev_name; int num_names; struct stat statbuf; - + Hash_table *nametab; + name_gather (); - if (listed_incremental_option) - read_directory_file (); - if (!namelist) - addname (".", 0); + addname (".", 0, NULL); + if (listed_incremental_option) + { + switch (chdir_count ()) + { + case 0: + break; + + case 1: + if (namelist->change_dir == 0) + USAGE_ERROR ((0, 0, + _("Using -C option inside file list is not " + "allowed with --listed-incremental"))); + break; + + default: + USAGE_ERROR ((0, 0, + _("Only one -C option is allowed with " + "--listed-incremental"))); + } + chdir_do (namelist->change_dir); + read_directory_file (); + } + for (name = namelist; name; name = next_name) { next_name = name->next; if (name->found_count || name->dir_contents) continue; - if (name->regexp) /* FIXME: just skip regexps for now */ + if (name->matching_flags & EXCLUDE_WILDCARDS) + /* NOTE: EXCLUDE_ANCHORED is not relevant here */ + /* FIXME: just skip regexps for now */ continue; chdir_do (name->change_dir); - if (name->fake) + + if (name->name[0] == 0) continue; if (deref_stat (dereference_option, name->name, &statbuf) != 0) @@ -774,23 +934,75 @@ collect_and_sort_names (void) if (S_ISDIR (statbuf.st_mode)) { name->found_count++; - add_hierarchy_to_namelist (name, statbuf.st_dev); + if (name->found_count == 1) + add_hierarchy_to_namelist (name, statbuf.st_dev, true); } } num_names = 0; for (name = namelist; name; name = name->next) num_names++; + namelist = merge_sort (namelist, num_names, compare_names); - for (name = namelist; name; name = name->next) - name->found_count = 0; + num_names = 0; + nametab = hash_initialize (0, 0, + name_hash, + name_compare, NULL); + for (name = namelist; name; name = next_name) + { + next_name = name->next; + name->caname = normalize_filename (name->name); + if (prev_name) + { + struct name *p = hash_lookup (nametab, name); + if (p) + { + /* Keep the one listed in the command line */ + if (!name->parent) + { + if (p->child) + rebase_child_list (p->child, name); + /* FIXME: remove_directory (p->caname); ? */ + remname (p); + free_name (p); + num_names--; + } + else + { + if (name->child) + rebase_child_list (name->child, p); + /* FIXME: remove_directory (name->caname); ? */ + remname (name); + free_name (name); + continue; + } + } + } + name->found_count = 0; + hash_insert (nametab, name); + prev_name = name; + num_names++; + } + nametail = &prev_name; + hash_free (nametab); + + namelist = merge_sort (namelist, num_names, compare_names_found); + + if (listed_incremental_option) + { + for (name = namelist; name && name->name[0] == 0; name++) + ; + if (name) + name->dir_contents = append_incremental_renames (name->dir_contents); + } } -/* This is like name_match, except that it returns a pointer to the - name it matched, and doesn't set FOUND in structure. The caller - will have to do that if it wants to. Oh, and if the namelist is - empty, it returns null, unlike name_match, which returns TRUE. */ +/* This is like name_match, except that + 1. It returns a pointer to the name it matched, and doesn't set FOUND + in structure. The caller will have to do that if it wants to. + 2. If the namelist is empty, it returns null, unlike name_match, which + returns TRUE. */ struct name * name_scan (const char *file_name) { @@ -828,7 +1040,8 @@ name_from_list (void) { if (!gnu_list_name) gnu_list_name = namelist; - while (gnu_list_name && (gnu_list_name->found_count || gnu_list_name->fake)) + while (gnu_list_name + && (gnu_list_name->found_count || gnu_list_name->name[0] == 0)) gnu_list_name = gnu_list_name->next; if (gnu_list_name) { @@ -850,7 +1063,7 @@ blank_name_list (void) } /* Yield a newly allocated file name consisting of FILE_NAME concatenated to - NAME, with an intervening slash if FILE_NAME does not already end in one. */ + NAME, with an intervening slash if FILE_NAME does not already end in one. */ char * new_name (const char *file_name, const char *name) { @@ -895,12 +1108,12 @@ static void register_individual_file (char const *name) { struct stat st; - + if (deref_stat (dereference_option, name, &st) != 0) return; /* Will be complained about later */ if (S_ISDIR (st.st_mode)) return; - + hash_string_insert (&individual_file_table, name); } @@ -948,11 +1161,10 @@ contains_dot_dot (char const *name) if (p[0] == '.' && p[1] == '.' && (ISSLASH (p[2]) || !p[2])) return 1; - do + while (! ISSLASH (*p)) { if (! *p++) return 0; } - while (! ISSLASH (*p)); } }