X-Git-Url: https://git.dogcows.com/gitweb?a=blobdiff_plain;f=src%2Fcreate.c;h=072732a6195250bd1a9ff436971de709a590bf6f;hb=1bcbbcf1ff2c537ffa970dbf82e3843d4ad110e5;hp=6e7d158b63e07916b143668b687d25d66503067b;hpb=b100f6ae5bd04c76a95567961ba6e033235ff0be;p=chaz%2Ftar diff --git a/src/create.c b/src/create.c index 6e7d158..072732a 100644 --- a/src/create.c +++ b/src/create.c @@ -1,13 +1,13 @@ /* Create a tar archive. Copyright (C) 1985, 1992, 1993, 1994, 1996, 1997, 1999, 2000, 2001, - 2003, 2004, 2005 Free Software Foundation, Inc. + 2003, 2004, 2005, 2006, 2007, 2009 Free Software Foundation, Inc. Written by John Gilmore, on 1985-08-25. 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 @@ -22,7 +22,6 @@ #include #include -#include #include "common.h" #include @@ -34,6 +33,112 @@ struct link size_t nlink; char name[1]; }; + +struct exclusion_tag +{ + const char *name; + size_t length; + enum exclusion_tag_type type; + bool (*predicate) (const char *name); + struct exclusion_tag *next; +}; + +static struct exclusion_tag *exclusion_tags; + +void +add_exclusion_tag (const char *name, enum exclusion_tag_type type, + bool (*predicate) (const char *name)) +{ + struct exclusion_tag *tag = xmalloc (sizeof tag[0]); + tag->next = exclusion_tags; + tag->name = name; + tag->type = type; + tag->predicate = predicate; + tag->length = strlen (name); + exclusion_tags = tag; +} + +void +exclusion_tag_warning (const char *dirname, const char *tagname, + const char *message) +{ + if (verbose_option) + WARNOPT (WARN_CACHEDIR, + (0, 0, + _("%s: contains a cache directory tag %s; %s"), + quotearg_colon (dirname), + quotearg_n (1, tagname), + message)); +} + +enum exclusion_tag_type +check_exclusion_tags (const char *dirname, const char **tag_file_name) +{ + static char *tagname; + static size_t tagsize; + struct exclusion_tag *tag; + size_t dlen = strlen (dirname); + int addslash = !ISSLASH (dirname[dlen-1]); + char *nptr = NULL; + + for (tag = exclusion_tags; tag; tag = tag->next) + { + size_t size = dlen + addslash + tag->length + 1; + if (size > tagsize) + { + tagsize = size; + tagname = xrealloc (tagname, tagsize); + } + + if (!nptr) + { + strcpy (tagname, dirname); + nptr = tagname + dlen; + if (addslash) + *nptr++ = '/'; + } + strcpy (nptr, tag->name); + if (access (tagname, F_OK) == 0 + && (!tag->predicate || tag->predicate (tagname))) + { + if (tag_file_name) + *tag_file_name = tag->name; + return tag->type; + } + } + + return exclusion_tag_none; +} + +/* Exclusion predicate to test if the named file (usually "CACHEDIR.TAG") + contains a valid header, as described at: + http://www.brynosaurus.com/cachedir + Applications can write this file into directories they create + for use as caches containing purely regenerable, non-precious data, + allowing us to avoid archiving them if --exclude-caches is specified. */ + +#define CACHEDIR_SIGNATURE "Signature: 8a477f597d28d172789f06886806bc55" +#define CACHEDIR_SIGNATURE_SIZE (sizeof CACHEDIR_SIGNATURE - 1) + +bool +cachedir_file_p (const char *name) +{ + bool tag_present = false; + int fd = open (name, O_RDONLY); + if (fd >= 0) + { + static char tagbuf[CACHEDIR_SIGNATURE_SIZE]; + + if (read (fd, tagbuf, CACHEDIR_SIGNATURE_SIZE) + == CACHEDIR_SIGNATURE_SIZE + && memcmp (tagbuf, CACHEDIR_SIGNATURE, CACHEDIR_SIGNATURE_SIZE) == 0) + tag_present = true; + + close (fd); + } + return tag_present; +} + /* The maximum uintmax_t value that can be represented with DIGITS digits, assuming that each digit is BITS_PER_DIGIT wide. */ @@ -109,6 +214,73 @@ to_base256 (int negative, uintmax_t value, char *where, size_t size) while (i); } + +static bool +to_chars (int negative, uintmax_t value, size_t valsize, + uintmax_t (*substitute) (int *), + char *where, size_t size, const char *type); + +static bool +to_chars_subst (int negative, int gnu_format, uintmax_t value, size_t valsize, + uintmax_t (*substitute) (int *), + char *where, size_t size, const char *type) +{ + uintmax_t maxval = (gnu_format + ? MAX_VAL_WITH_DIGITS (size - 1, LG_256) + : MAX_VAL_WITH_DIGITS (size - 1, LG_8)); + char valbuf[UINTMAX_STRSIZE_BOUND + 1]; + char maxbuf[UINTMAX_STRSIZE_BOUND]; + char minbuf[UINTMAX_STRSIZE_BOUND + 1]; + char const *minval_string; + char const *maxval_string = STRINGIFY_BIGINT (maxval, maxbuf); + char const *value_string; + + if (gnu_format) + { + uintmax_t m = maxval + 1 ? maxval + 1 : maxval / 2 + 1; + char *p = STRINGIFY_BIGINT (m, minbuf + 1); + *--p = '-'; + minval_string = p; + } + else + minval_string = "0"; + + if (negative) + { + char *p = STRINGIFY_BIGINT (- value, valbuf + 1); + *--p = '-'; + value_string = p; + } + else + value_string = STRINGIFY_BIGINT (value, valbuf); + + if (substitute) + { + int negsub; + uintmax_t sub = substitute (&negsub) & maxval; + /* NOTE: This is one of the few places where GNU_FORMAT differs from + OLDGNU_FORMAT. The actual differences are: + + 1. In OLDGNU_FORMAT all strings in a tar header end in \0 + 2. Incremental archives use oldgnu_header. + + Apart from this they are completely identical. */ + uintmax_t s = (negsub &= archive_format == GNU_FORMAT) ? - sub : sub; + char subbuf[UINTMAX_STRSIZE_BOUND + 1]; + char *sub_string = STRINGIFY_BIGINT (s, subbuf + 1); + if (negsub) + *--sub_string = '-'; + WARN ((0, 0, _("value %s out of %s range %s..%s; substituting %s"), + value_string, type, minval_string, maxval_string, + sub_string)); + return to_chars (negsub, s, valsize, 0, where, size, type); + } + else + ERROR ((0, 0, _("value %s out of %s range %s..%s"), + value_string, type, minval_string, maxval_string)); + return false; +} + /* Convert NEGATIVE VALUE (which was originally of size VALSIZE) to external form, using SUBSTITUTE (...) if VALUE won't fit. Output to buffer WHERE with size SIZE. NEGATIVE is 1 iff VALUE was @@ -123,103 +295,61 @@ to_base256 (int negative, uintmax_t value, char *where, size_t size) SUBSTITUTE the address of an 0-or-1 flag recording whether the substitute value is negative. */ -static void +static bool to_chars (int negative, uintmax_t value, size_t valsize, uintmax_t (*substitute) (int *), char *where, size_t size, const char *type) { - int base256_allowed = (archive_format == GNU_FORMAT - || archive_format == OLDGNU_FORMAT); + int gnu_format = (archive_format == GNU_FORMAT + || archive_format == OLDGNU_FORMAT); /* Generate the POSIX octal representation if the number fits. */ if (! negative && value <= MAX_VAL_WITH_DIGITS (size - 1, LG_8)) { where[size - 1] = '\0'; to_octal (value, where, size - 1); + return true; } - - /* Otherwise, generate the base-256 representation if we are - generating an old or new GNU format and if the number fits. */ - else if (((negative ? -1 - value : value) - <= MAX_VAL_WITH_DIGITS (size - 1, LG_256)) - && base256_allowed) - { - where[0] = negative ? -1 : 1 << (LG_256 - 1); - to_base256 (negative, value, where + 1, size - 1); - } - - /* Otherwise, if the number is negative, and if it would not cause - ambiguity on this host by confusing positive with negative - values, then generate the POSIX octal representation of the value - modulo 2**(field bits). The resulting tar file is - machine-dependent, since it depends on the host word size. Yuck! - But this is the traditional behavior. */ - else if (negative && valsize * CHAR_BIT <= (size - 1) * LG_8) - { - static int warned_once; - if (! warned_once) - { - warned_once = 1; - WARN ((0, 0, _("Generating negative octal headers"))); - } - where[size - 1] = '\0'; - to_octal (value & MAX_VAL_WITH_DIGITS (valsize * CHAR_BIT, 1), - where, size - 1); - } - - /* Otherwise, output a substitute value if possible (with a - warning), and an error message if not. */ - else + else if (gnu_format) { - uintmax_t maxval = (base256_allowed - ? MAX_VAL_WITH_DIGITS (size - 1, LG_256) - : MAX_VAL_WITH_DIGITS (size - 1, LG_8)); - char valbuf[UINTMAX_STRSIZE_BOUND + 1]; - char maxbuf[UINTMAX_STRSIZE_BOUND]; - char minbuf[UINTMAX_STRSIZE_BOUND + 1]; - char const *minval_string; - char const *maxval_string = STRINGIFY_BIGINT (maxval, maxbuf); - char const *value_string; - - if (base256_allowed) - { - uintmax_t m = maxval + 1 ? maxval + 1 : maxval / 2 + 1; - char *p = STRINGIFY_BIGINT (m, minbuf + 1); - *--p = '-'; - minval_string = p; - } - else - minval_string = "0"; + /* Try to cope with the number by using traditional GNU format + methods */ - if (negative) + /* Generate the base-256 representation if the number fits. */ + if (((negative ? -1 - value : value) + <= MAX_VAL_WITH_DIGITS (size - 1, LG_256))) { - char *p = STRINGIFY_BIGINT (- value, valbuf + 1); - *--p = '-'; - value_string = p; + where[0] = negative ? -1 : 1 << (LG_256 - 1); + to_base256 (negative, value, where + 1, size - 1); + return true; } - else - value_string = STRINGIFY_BIGINT (value, valbuf); - if (substitute) + /* Otherwise, if the number is negative, and if it would not cause + ambiguity on this host by confusing positive with negative + values, then generate the POSIX octal representation of the value + modulo 2**(field bits). The resulting tar file is + machine-dependent, since it depends on the host word size. Yuck! + But this is the traditional behavior. */ + else if (negative && valsize * CHAR_BIT <= (size - 1) * LG_8) { - int negsub; - uintmax_t sub = substitute (&negsub) & maxval; - /* FIXME: This is the only place where GNU_FORMAT differs from - OLDGNU_FORMAT. Apart from this they are completely identical. */ - uintmax_t s = (negsub &= archive_format == GNU_FORMAT) ? - sub : sub; - char subbuf[UINTMAX_STRSIZE_BOUND + 1]; - char *sub_string = STRINGIFY_BIGINT (s, subbuf + 1); - if (negsub) - *--sub_string = '-'; - WARN ((0, 0, _("value %s out of %s range %s..%s; substituting %s"), - value_string, type, minval_string, maxval_string, - sub_string)); - to_chars (negsub, s, valsize, 0, where, size, type); + static int warned_once; + if (! warned_once) + { + warned_once = 1; + WARN ((0, 0, _("Generating negative octal headers"))); + } + where[size - 1] = '\0'; + to_octal (value & MAX_VAL_WITH_DIGITS (valsize * CHAR_BIT, 1), + where, size - 1); + return true; } - else - ERROR ((0, 0, _("value %s out of %s range %s..%s"), - value_string, type, minval_string, maxval_string)); + /* Otherwise fall back to substitution, if possible: */ } + else + substitute = NULL; /* No substitution for formats, other than GNU */ + + return to_chars_subst (negative, gnu_format, value, valsize, substitute, + where, size, type); } static uintmax_t @@ -238,25 +368,25 @@ gid_substitute (int *negative) return r; } -void +bool gid_to_chars (gid_t v, char *p, size_t s) { - to_chars (v < 0, (uintmax_t) v, sizeof v, gid_substitute, p, s, "gid_t"); + return to_chars (v < 0, (uintmax_t) v, sizeof v, gid_substitute, p, s, "gid_t"); } -void +bool major_to_chars (major_t v, char *p, size_t s) { - to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "major_t"); + return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "major_t"); } -void +bool minor_to_chars (minor_t v, char *p, size_t s) { - to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "minor_t"); + return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "minor_t"); } -void +bool mode_to_chars (mode_t v, char *p, size_t s) { /* In the common case where the internal and external mode bits are the same, @@ -272,7 +402,8 @@ mode_to_chars (mode_t v, char *p, size_t s) && S_IROTH == TOREAD && S_IWOTH == TOWRITE && S_IXOTH == TOEXEC && archive_format != POSIX_FORMAT && archive_format != USTAR_FORMAT - && archive_format != GNU_FORMAT) + && archive_format != GNU_FORMAT + && archive_format != OLDGNU_FORMAT) { negative = v < 0; u = v; @@ -293,25 +424,25 @@ mode_to_chars (mode_t v, char *p, size_t s) | (v & S_IWOTH ? TOWRITE : 0) | (v & S_IXOTH ? TOEXEC : 0)); } - to_chars (negative, u, sizeof v, 0, p, s, "mode_t"); + return to_chars (negative, u, sizeof v, 0, p, s, "mode_t"); } -void +bool off_to_chars (off_t v, char *p, size_t s) { - to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "off_t"); + return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "off_t"); } -void +bool size_to_chars (size_t v, char *p, size_t s) { - to_chars (0, (uintmax_t) v, sizeof v, 0, p, s, "size_t"); + return to_chars (0, (uintmax_t) v, sizeof v, 0, p, s, "size_t"); } -void +bool time_to_chars (time_t v, char *p, size_t s) { - to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "time_t"); + return to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "time_t"); } static uintmax_t @@ -330,16 +461,16 @@ uid_substitute (int *negative) return r; } -void +bool uid_to_chars (uid_t v, char *p, size_t s) { - to_chars (v < 0, (uintmax_t) v, sizeof v, uid_substitute, p, s, "uid_t"); + return to_chars (v < 0, (uintmax_t) v, sizeof v, uid_substitute, p, s, "uid_t"); } -void +bool uintmax_to_chars (uintmax_t v, char *p, size_t s) { - to_chars (0, v, sizeof v, 0, p, s, "uintmax_t"); + return to_chars (0, v, sizeof v, 0, p, s, "uintmax_t"); } void @@ -361,7 +492,7 @@ bool file_dumpable_p (struct tar_stat_info *st) { if (dev_null_output) - return totals_option && sparse_option && sparse_file_p (st); + return totals_option && sparse_option && ST_IS_SPARSE (st->stat); return !(st->archive_file_size == 0 && (st->stat.st_mode & MODE_R) == MODE_R); } @@ -475,7 +606,7 @@ split_long_name (const char *name, size_t length) size_t i; if (length > PREFIX_FIELD_SIZE) - length = PREFIX_FIELD_SIZE+2; + length = PREFIX_FIELD_SIZE + 1; for (i = length - 1; i > 0; i--) if (ISSLASH (name[i])) break; @@ -576,19 +707,29 @@ write_long_name (struct tar_stat_info *st) return write_short_name (st); } -static union block * -write_extended (struct tar_stat_info *st, union block *old_header) +union block * +write_extended (bool global, struct tar_stat_info *st, union block *old_header) { union block *header, hp; char *p; + int type; - if (extended_header.buffer || extended_header.stk == NULL) + if (st->xhdr.buffer || st->xhdr.stk == NULL) return old_header; - xheader_finish (&extended_header); + xheader_finish (&st->xhdr); memcpy (hp.buffer, old_header, sizeof (hp)); - p = xheader_xhdr_name (st); - xheader_write (XHDTYPE, p, &extended_header); + if (global) + { + type = XGLTYPE; + p = xheader_ghdr_name (); + } + else + { + type = XHDTYPE; + p = xheader_xhdr_name (st); + } + xheader_write (type, p, &st->xhdr); free (p); header = find_next_block (); memcpy (header, &hp.buffer, sizeof (hp.buffer)); @@ -634,7 +775,8 @@ start_header (struct tar_stat_info *st) if (mode_option) st->stat.st_mode = ((st->stat.st_mode & ~MODE_ALL) - | mode_adjust (st->stat.st_mode, mode_option, initial_umask)); + | mode_adjust (st->stat.st_mode, S_ISDIR (st->stat.st_mode) != 0, + initial_umask, mode_option, NULL)); /* Paul Eggert tried the trivial test ($WRITER cf a b; $READER tvf a) for a few tars and came up with the following interoperability @@ -672,7 +814,8 @@ start_header (struct tar_stat_info *st) xheader_store ("uid", st, NULL); uid = 0; } - UID_TO_CHARS (uid, header->header.uid); + if (!UID_TO_CHARS (uid, header->header.uid)) + return NULL; } { @@ -683,7 +826,8 @@ start_header (struct tar_stat_info *st) xheader_store ("gid", st, NULL); gid = 0; } - GID_TO_CHARS (gid, header->header.gid); + if (!GID_TO_CHARS (gid, header->header.gid)) + return NULL; } { @@ -694,20 +838,22 @@ start_header (struct tar_stat_info *st) xheader_store ("size", st, NULL); size = 0; } - OFF_TO_CHARS (size, header->header.size); + if (!OFF_TO_CHARS (size, header->header.size)) + return NULL; } { - struct timespec mtime = st->mtime; + struct timespec mtime = set_mtime_option ? mtime_option : st->mtime; if (archive_format == POSIX_FORMAT) { if (MAX_OCTAL_VAL (header->header.mtime) < mtime.tv_sec || mtime.tv_nsec != 0) - xheader_store ("mtime", st, NULL); + xheader_store ("mtime", st, &mtime); if (MAX_OCTAL_VAL (header->header.mtime) < mtime.tv_sec) mtime.tv_sec = 0; } - TIME_TO_CHARS (mtime.tv_sec, header->header.mtime); + if (!TIME_TO_CHARS (mtime.tv_sec, header->header.mtime)) + return NULL; } /* FIXME */ @@ -723,7 +869,8 @@ start_header (struct tar_stat_info *st) xheader_store ("devmajor", st, NULL); devmajor = 0; } - MAJOR_TO_CHARS (devmajor, header->header.devmajor); + if (!MAJOR_TO_CHARS (devmajor, header->header.devmajor)) + return NULL; if (archive_format == POSIX_FORMAT && MAX_OCTAL_VAL (header->header.devminor) < devminor) @@ -731,12 +878,14 @@ start_header (struct tar_stat_info *st) xheader_store ("devminor", st, NULL); devminor = 0; } - MINOR_TO_CHARS (devminor, header->header.devminor); + if (!MINOR_TO_CHARS (devminor, header->header.devminor)) + return NULL; } else if (archive_format != GNU_FORMAT && archive_format != OLDGNU_FORMAT) { - MAJOR_TO_CHARS (0, header->header.devmajor); - MINOR_TO_CHARS (0, header->header.devminor); + if (!(MAJOR_TO_CHARS (0, header->header.devmajor) + && MINOR_TO_CHARS (0, header->header.devminor))) + return NULL; } if (archive_format == POSIX_FORMAT) @@ -852,7 +1001,7 @@ finish_header (struct tar_stat_info *st, print_header (st, block_ordinal); } - header = write_extended (st, header); + header = write_extended (false, st, header); simple_finish_header (header); } @@ -863,7 +1012,7 @@ pad_archive (off_t size_left) union block *blk; while (size_left > 0) { - save_sizeleft = size_left; + mv_size_left (size_left); blk = find_next_block (); memset (blk->buffer, 0, BLOCKSIZE); set_next_block_after (blk); @@ -889,16 +1038,13 @@ dump_regular_file (int fd, struct tar_stat_info *st) finish_header (st, blk, block_ordinal); + mv_begin (st); while (size_left > 0) { size_t bufsize, count; + + mv_size_left (size_left); - if (multi_volume_option) - { - assign_string (&save_name, st->orig_file_name); - save_sizeleft = size_left; - save_totsize = st->stat.st_size; - } blk = find_next_block (); bufsize = available_space_after (blk); @@ -921,104 +1067,36 @@ dump_regular_file (int fd, struct tar_stat_info *st) return dump_status_short; } size_left -= count; - if (count) - set_next_block_after (blk + (bufsize - 1) / BLOCKSIZE); + set_next_block_after (blk + (bufsize - 1) / BLOCKSIZE); if (count != bufsize) { char buf[UINTMAX_STRSIZE_BOUND]; memset (blk->buffer + count, 0, bufsize - count); - WARN ((0, 0, - ngettext ("%s: File shrank by %s byte; padding with zeros", - "%s: File shrank by %s bytes; padding with zeros", - size_left), - quotearg_colon (st->orig_file_name), - STRINGIFY_BIGINT (size_left, buf))); - if (! ignore_failed_read_option) - exit_status = TAREXIT_FAILURE; - pad_archive (size_left - (bufsize-count)); + WARNOPT (WARN_FILE_SHRANK, + (0, 0, + ngettext ("%s: File shrank by %s byte; padding with zeros", + "%s: File shrank by %s bytes; padding with zeros", + size_left), + quotearg_colon (st->orig_file_name), + STRINGIFY_BIGINT (size_left, buf))); + if (! ignore_failed_read_option) + exit_status = TAREXIT_DIFFERS; + pad_archive (size_left - (bufsize - count)); return dump_status_short; } } return dump_status_ok; } -static void -dump_regular_finish (int fd, struct tar_stat_info *st, - struct timespec original_ctime) -{ - if (fd >= 0) - { - struct stat final_stat; - if (fstat (fd, &final_stat) != 0) - { - stat_diag (st->orig_file_name); - } - else if (timespec_cmp (get_stat_ctime (&final_stat), original_ctime) - != 0) - { - WARN ((0, 0, _("%s: file changed as we read it"), - quotearg_colon (st->orig_file_name))); - } - if (close (fd) != 0) - { - close_diag (st->orig_file_name); - } - } - if (remove_files_option) - { - if (unlink (st->orig_file_name) == -1) - unlink_error (st->orig_file_name); - } -} - -/* Look in directory DIRNAME for a cache directory tag file - with the magic name "CACHEDIR.TAG" and a standard header, - as described at: - http://www.brynosaurus.com/cachedir - Applications can write this file into directories they create - for use as caches containing purely regenerable, non-precious data, - allowing us to avoid archiving them if --exclude-caches is specified. */ - -#define CACHEDIR_SIGNATURE "Signature: 8a477f597d28d172789f06886806bc55" -#define CACHEDIR_SIGNATURE_SIZE (sizeof CACHEDIR_SIGNATURE - 1) - -static bool -check_cache_directory (char *dirname) -{ - static char tagname[] = "CACHEDIR.TAG"; - char *tagpath; - int fd; - int tag_present = false; - - tagpath = xmalloc (strlen (dirname) + strlen (tagname) + 1); - strcpy (tagpath, dirname); - strcat (tagpath, tagname); - - fd = open (tagpath, O_RDONLY); - if (fd >= 0) - { - static char tagbuf[CACHEDIR_SIGNATURE_SIZE]; - - if (read (fd, tagbuf, CACHEDIR_SIGNATURE_SIZE) - == CACHEDIR_SIGNATURE_SIZE - && memcmp (tagbuf, CACHEDIR_SIGNATURE, CACHEDIR_SIGNATURE_SIZE) == 0) - tag_present = true; - - close (fd); - } - - free (tagpath); - - return tag_present; -} - + static void dump_dir0 (char *directory, struct tar_stat_info *st, int top_level, dev_t parent_device) { dev_t our_device = st->stat.st_dev; - + const char *tag_file_name; + if (!is_avoided_name (st->orig_file_name)) { union block *blk = NULL; @@ -1029,7 +1107,7 @@ dump_dir0 (char *directory, if (!blk) return; - if (incremental_option) + if (incremental_option && archive_format != POSIX_FORMAT) blk->header.typeflag = GNUTYPE_DUMPDIR; else /* if (standard_option) */ blk->header.typeflag = DIRTYPE; @@ -1040,51 +1118,51 @@ dump_dir0 (char *directory, finish_header (st, blk, block_ordinal); else if (gnu_list_name->dir_contents) { - off_t size_left; - off_t totsize; - size_t bufsize; - ssize_t count; - const char *buffer, *p_buffer; - - block_ordinal = current_block_ordinal (); - buffer = gnu_list_name->dir_contents; /* FOO */ - totsize = 0; - if (buffer) - for (p_buffer = buffer; *p_buffer; ) - { - size_t size = strlen (p_buffer) + 1; - totsize += size; - p_buffer += size; - } - totsize++; - OFF_TO_CHARS (totsize, blk->header.size); - finish_header (st, blk, block_ordinal); - p_buffer = buffer; - size_left = totsize; - while (size_left > 0) + if (archive_format == POSIX_FORMAT) { - if (multi_volume_option) - { - assign_string (&save_name, st->orig_file_name); - save_sizeleft = size_left; - save_totsize = totsize; - } - blk = find_next_block (); - bufsize = available_space_after (blk); - if (size_left < bufsize) + xheader_store ("GNU.dumpdir", st, gnu_list_name->dir_contents); + finish_header (st, blk, block_ordinal); + } + else + { + off_t size_left; + off_t totsize; + size_t bufsize; + ssize_t count; + const char *buffer, *p_buffer; + + block_ordinal = current_block_ordinal (); + buffer = gnu_list_name->dir_contents; + if (buffer) + totsize = dumpdir_size (buffer); + else + totsize = 0; + OFF_TO_CHARS (totsize, blk->header.size); + finish_header (st, blk, block_ordinal); + p_buffer = buffer; + size_left = totsize; + + mv_begin (st); + mv_total_size (totsize); + while (size_left > 0) { - bufsize = size_left; - count = bufsize % BLOCKSIZE; - if (count) - memset (blk->buffer + size_left, 0, BLOCKSIZE - count); + mv_size_left (size_left); + blk = find_next_block (); + bufsize = available_space_after (blk); + if (size_left < bufsize) + { + bufsize = size_left; + count = bufsize % BLOCKSIZE; + if (count) + memset (blk->buffer + size_left, 0, BLOCKSIZE - count); + } + memcpy (blk->buffer, p_buffer, bufsize); + size_left -= bufsize; + p_buffer += bufsize; + set_next_block_after (blk + (bufsize - 1) / BLOCKSIZE); } - memcpy (blk->buffer, p_buffer, bufsize); - size_left -= bufsize; - p_buffer += bufsize; - set_next_block_after (blk + (bufsize - 1) / BLOCKSIZE); + mv_end (); } - if (multi_volume_option) - assign_string (&save_name, 0); return; } } @@ -1097,47 +1175,67 @@ dump_dir0 (char *directory, && parent_device != st->stat.st_dev) { if (verbose_option) - WARN ((0, 0, - _("%s: file is on a different filesystem; not dumped"), - quotearg_colon (st->orig_file_name))); - return; + WARNOPT (WARN_XDEV, + (0, 0, + _("%s: file is on a different filesystem; not dumped"), + quotearg_colon (st->orig_file_name))); } - - if (exclude_caches_option - && check_cache_directory(st->orig_file_name)) + else { - if (verbose_option) - WARN ((0, 0, - _("%s: contains a cache directory tag; not dumped"), - quotearg_colon (st->orig_file_name))); - return; - } - - { - char const *entry; - size_t entry_len; - char *name_buf = xstrdup (st->orig_file_name); - size_t name_size = strlen (name_buf); - size_t name_len = name_size; + char *name_buf; + size_t name_size; + + switch (check_exclusion_tags (st->orig_file_name, &tag_file_name)) + { + case exclusion_tag_all: + /* Handled in dump_file0 */ + break; + + case exclusion_tag_none: + { + char const *entry; + size_t entry_len; + size_t name_len; - /* Now output all the files in the directory. */ - /* FIXME: Should speed this up by cd-ing into the dir. */ + name_buf = xstrdup (st->orig_file_name); + name_size = name_len = strlen (name_buf); - for (entry = directory; (entry_len = strlen (entry)) != 0; - entry += entry_len + 1) - { - if (name_size < name_len + entry_len) - { - name_size = name_len + entry_len; - name_buf = xrealloc (name_buf, name_size + 1); + /* Now output all the files in the directory. */ + /* FIXME: Should speed this up by cd-ing into the dir. */ + for (entry = directory; (entry_len = strlen (entry)) != 0; + entry += entry_len + 1) + { + if (name_size < name_len + entry_len) + { + name_size = name_len + entry_len; + name_buf = xrealloc (name_buf, name_size + 1); + } + strcpy (name_buf + name_len, entry); + if (!excluded_name (name_buf)) + dump_file (name_buf, 0, our_device); + } + + free (name_buf); } - strcpy (name_buf + name_len, entry); - if (!excluded_name (name_buf)) + break; + + case exclusion_tag_contents: + exclusion_tag_warning (st->orig_file_name, tag_file_name, + _("contents not dumped")); + name_size = strlen (st->orig_file_name) + strlen (tag_file_name) + 1; + name_buf = xmalloc (name_size); + strcpy (name_buf, st->orig_file_name); + strcat (name_buf, tag_file_name); dump_file (name_buf, 0, our_device); - } - - free (name_buf); - } + free (name_buf); + break; + + case exclusion_tag_under: + exclusion_tag_warning (st->orig_file_name, tag_file_name, + _("contents not dumped")); + break; + } + } } /* Ensure exactly one trailing slash. */ @@ -1154,20 +1252,15 @@ ensure_slash (char **pstr) } static bool -dump_dir (struct tar_stat_info *st, int top_level, dev_t parent_device) +dump_dir (int fd, struct tar_stat_info *st, int top_level, dev_t parent_device) { - char *directory; - - directory = savedir (st->orig_file_name); + char *directory = fdsavedir (fd); if (!directory) { savedir_diag (st->orig_file_name); return false; } - ensure_slash (&st->orig_file_name); - ensure_slash (&st->file_name); - dump_dir0 (directory, st, top_level, parent_device); free (directory); @@ -1180,10 +1273,10 @@ dump_dir (struct tar_stat_info *st, int top_level, dev_t parent_device) void create_archive (void) { - char *p; + const char *p; open_archive (ACCESS_WRITE); - xheader_write_global (); + buffer_write_global_xheader (); if (incremental_option) { @@ -1210,7 +1303,7 @@ create_archive (void) } memcpy (buffer, p, plen); if (! ISSLASH (buffer[plen - 1])) - buffer[plen++] = '/'; + buffer[plen++] = DIRECTORY_SEPARATOR; q = gnu_list_name->dir_contents; if (q) while (*q) @@ -1266,10 +1359,11 @@ compare_links (void const *entry1, void const *entry2) } static void -unknown_file_error (char *p) +unknown_file_error (char const *p) { - WARN ((0, 0, _("%s: Unknown file type; file ignored"), - quotearg_colon (p))); + WARNOPT (WARN_FILE_IGNORED, + (0, 0, _("%s: Unknown file type; file ignored"), + quotearg_colon (p))); if (!ignore_failed_read_option) exit_status = TAREXIT_FAILURE; } @@ -1282,12 +1376,12 @@ unknown_file_error (char *p) again if we've done it once already. */ static Hash_table *link_table; -/* Try to dump stat as a hard link to another file in the archive. If - succeeded returns true */ +/* Try to dump stat as a hard link to another file in the archive. + Return true if successful. */ static bool dump_hard_link (struct tar_stat_info *st) { - if (link_table && st->stat.st_nlink > 1) + if (link_table && (st->stat.st_nlink > 1 || remove_files_option)) { struct link lp; struct link *duplicate; @@ -1314,7 +1408,7 @@ dump_hard_link (struct tar_stat_info *st) st->stat.st_size = 0; blk = start_header (st); if (!blk) - return true; + return false; tar_copy_str (blk->header.linkname, link_name, NAME_FIELD_SIZE); blk->header.typeflag = LNKTYPE; @@ -1332,22 +1426,31 @@ dump_hard_link (struct tar_stat_info *st) static void file_count_links (struct tar_stat_info *st) { + if (hard_dereference_option) + return; if (st->stat.st_nlink > 1) { struct link *duplicate; - struct link *lp = xmalloc (offsetof (struct link, name) - + strlen (st->orig_file_name) + 1); + char *linkname = NULL; + struct link *lp; + + assign_string (&linkname, st->orig_file_name); + transform_name (&linkname, XFORM_LINK); + + lp = xmalloc (offsetof (struct link, name) + + strlen (linkname) + 1); lp->ino = st->stat.st_ino; lp->dev = st->stat.st_dev; lp->nlink = st->stat.st_nlink; - strcpy (lp->name, st->orig_file_name); - + strcpy (lp->name, linkname); + free (linkname); + if (! ((link_table || (link_table = hash_initialize (0, 0, hash_link, compare_links, 0))) && (duplicate = hash_insert (link_table, lp)))) xalloc_die (); - + if (duplicate != lp) abort (); lp->nlink--; @@ -1369,7 +1472,7 @@ check_links (void) { if (lp->nlink) { - WARN ((0, 0, _("Missing links to %s.\n"), quote (lp->name))); + WARN ((0, 0, _("Missing links to %s."), quote (lp->name))); } } } @@ -1385,14 +1488,16 @@ check_links (void) exit_status to failure, a clear diagnostic has been issued. */ static void -dump_file0 (struct tar_stat_info *st, char *p, +dump_file0 (struct tar_stat_info *st, const char *p, int top_level, dev_t parent_device) { union block *header; char type; + off_t original_size; struct timespec original_ctime; struct timespec restore_times[2]; off_t block_ordinal = -1; + bool is_dir; if (interactive_option && !confirm ("add", p)) return; @@ -1401,12 +1506,14 @@ dump_file0 (struct tar_stat_info *st, char *p, assign_string (&st->file_name, safer_name_suffix (p, false, absolute_names_option)); + transform_name (&st->file_name, XFORM_REGFILE); + if (deref_stat (dereference_option, p, &st->stat) != 0) { stat_diag (p); return; } - st->archive_file_size = st->stat.st_size; + st->archive_file_size = original_size = st->stat.st_size; st->atime = restore_times[0] = get_stat_atime (&st->stat); st->mtime = restore_times[1] = get_stat_mtime (&st->stat); st->ctime = original_ctime = get_stat_ctime (&st->stat); @@ -1426,70 +1533,90 @@ dump_file0 (struct tar_stat_info *st, char *p, /* See if we want only new files, and check if this one is too old to put in the archive. - + This check is omitted if incremental_option is set *and* the requested file is not explicitely listed in the command line. */ - + if (!(incremental_option && !is_individual_file (p)) && !S_ISDIR (st->stat.st_mode) && OLDER_TAR_STAT_TIME (*st, m) && (!after_date_option || OLDER_TAR_STAT_TIME (*st, c))) { if (!incremental_option && verbose_option) - WARN ((0, 0, _("%s: file is unchanged; not dumped"), - quotearg_colon (p))); + WARNOPT (WARN_FILE_UNCHANGED, + (0, 0, _("%s: file is unchanged; not dumped"), + quotearg_colon (p))); return; } /* See if we are trying to dump the archive. */ if (sys_file_is_archive (st)) { - WARN ((0, 0, _("%s: file is the archive; not dumped"), - quotearg_colon (p))); + WARNOPT (WARN_IGNORE_ARCHIVE, + (0, 0, _("%s: file is the archive; not dumped"), + quotearg_colon (p))); return; } if (is_avoided_name (p)) return; - if (S_ISDIR (st->stat.st_mode)) - { - dump_dir (st, top_level, parent_device); - if (atime_preserve_option) - utimens (p, restore_times); - return; - } - else + + is_dir = S_ISDIR (st->stat.st_mode) != 0; + + if (!is_dir && dump_hard_link (st)) + return; + + if (is_dir || S_ISREG (st->stat.st_mode) || S_ISCTG (st->stat.st_mode)) { - /* Check for multiple links. */ - if (dump_hard_link (st)) - return; + bool ok; + int fd = -1; + struct stat final_stat; - /* This is not a link to a previously dumped file, so dump it. */ + if (is_dir || file_dumpable_p (st)) + { + fd = open (p, + (O_RDONLY | O_BINARY + | (is_dir ? O_DIRECTORY | O_NONBLOCK : 0) + | (atime_preserve_option == system_atime_preserve + ? O_NOATIME + : 0))); + if (fd < 0) + { + if (!top_level && errno == ENOENT) + WARNOPT (WARN_FILE_REMOVED, + (0, 0, _("%s: File removed before we read it"), + quotearg_colon (p))); + else + open_diag (p); + return; + } + } - if (S_ISREG (st->stat.st_mode) - || S_ISCTG (st->stat.st_mode)) + if (is_dir) { - int fd; - enum dump_status status; + const char *tag_file_name; + ensure_slash (&st->orig_file_name); + ensure_slash (&st->file_name); - if (file_dumpable_p (st)) + if (check_exclusion_tags (st->orig_file_name, &tag_file_name) + == exclusion_tag_all) { - fd = open (st->orig_file_name, - O_RDONLY | O_BINARY); - if (fd < 0) - { - if (!top_level && errno == ENOENT) - WARN ((0, 0, _("%s: File removed before we read it"), - quotearg_colon (st->orig_file_name))); - else - open_diag (st->orig_file_name); - return; - } + exclusion_tag_warning (st->orig_file_name, tag_file_name, + _("directory not dumped")); + return; } - else + + ok = dump_dir (fd, st, top_level, parent_device); + + /* dump_dir consumes FD if successful. */ + if (ok) fd = -1; + } + else + { + enum dump_status status; - if (fd != -1 && sparse_option && sparse_file_p (st)) + if (fd != -1 && sparse_option && ST_IS_SPARSE (st->stat)) { status = sparse_dump_file (fd, st); if (status == dump_status_not_implemented) @@ -1501,90 +1628,143 @@ dump_file0 (struct tar_stat_info *st, char *p, switch (status) { case dump_status_ok: - if (multi_volume_option) - assign_string (&save_name, 0); - dump_regular_finish (fd, st, original_ctime); - break; - case dump_status_short: - if (multi_volume_option) - assign_string (&save_name, 0); - close (fd); + mv_end (); + file_count_links (st); break; case dump_status_fail: - close (fd); - return; + break; case dump_status_not_implemented: abort (); } - if (atime_preserve_option) - utimens (st->orig_file_name, restore_times); - file_count_links (st); - return; + ok = status == dump_status_ok; } -#ifdef HAVE_READLINK - else if (S_ISLNK (st->stat.st_mode)) + + if (ok) { - char *buffer; - int size; - size_t linklen = st->stat.st_size; - if (linklen != st->stat.st_size || linklen + 1 == 0) - xalloc_die (); - buffer = (char *) alloca (linklen + 1); - size = readlink (p, buffer, linklen + 1); - if (size < 0) + /* If possible, reopen a directory if we are preserving + atimes, so that we can set just the atime on systems with + _FIOSATIME. */ + if (fd < 0 && is_dir + && atime_preserve_option == replace_atime_preserve) + fd = open (p, O_RDONLY | O_BINARY | O_DIRECTORY | O_NONBLOCK); + + if ((fd < 0 + ? deref_stat (dereference_option, p, &final_stat) + : fstat (fd, &final_stat)) + != 0) { - readlink_diag (p); - return; + stat_diag (p); + ok = false; } - buffer[size] = '\0'; - assign_string (&st->link_name, buffer); - if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT) < size) - write_long_link (st); + } - block_ordinal = current_block_ordinal (); - st->stat.st_size = 0; /* force 0 size on symlink */ - header = start_header (st); - if (!header) - return; - tar_copy_str (header->header.linkname, buffer, NAME_FIELD_SIZE); - header->header.typeflag = SYMTYPE; - finish_header (st, header, block_ordinal); - /* nothing more to do to it */ - - if (remove_files_option) + if (ok) + { + if ((timespec_cmp (get_stat_ctime (&final_stat), original_ctime) != 0 + /* Original ctime will change if the file is a directory and + --remove-files is given */ + && !(remove_files_option && is_dir)) + || original_size < final_stat.st_size) { - if (unlink (p) == -1) - unlink_error (p); + WARNOPT (WARN_FILE_CHANGED, + (0, 0, _("%s: file changed as we read it"), + quotearg_colon (p))); + if (exit_status == TAREXIT_SUCCESS) + exit_status = TAREXIT_DIFFERS; } - file_count_links (st); - return; + else if (atime_preserve_option == replace_atime_preserve + && set_file_atime (fd, p, restore_times) != 0) + utime_error (p); } -#endif - else if (S_ISCHR (st->stat.st_mode)) - type = CHRTYPE; - else if (S_ISBLK (st->stat.st_mode)) - type = BLKTYPE; - else if (S_ISFIFO (st->stat.st_mode)) - type = FIFOTYPE; - else if (S_ISSOCK (st->stat.st_mode)) + + if (0 <= fd && close (fd) != 0) { - WARN ((0, 0, _("%s: socket ignored"), quotearg_colon (p))); - return; + close_diag (p); + ok = false; } - else if (S_ISDOOR (st->stat.st_mode)) + + if (ok && remove_files_option) { - WARN ((0, 0, _("%s: door ignored"), quotearg_colon (p))); - return; + if (is_dir) + { + if (rmdir (p) != 0 && errno != ENOTEMPTY) + rmdir_error (p); + } + else + { + if (unlink (p) != 0) + unlink_error (p); + } } - else + + return; + } +#ifdef HAVE_READLINK + else if (S_ISLNK (st->stat.st_mode)) + { + char *buffer; + int size; + size_t linklen = st->stat.st_size; + if (linklen != st->stat.st_size || linklen + 1 == 0) + xalloc_die (); + buffer = (char *) alloca (linklen + 1); + size = readlink (p, buffer, linklen + 1); + if (size < 0) { - unknown_file_error (p); + readlink_diag (p); return; } + buffer[size] = '\0'; + assign_string (&st->link_name, buffer); + transform_name (&st->link_name, XFORM_SYMLINK); + if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT) < size) + write_long_link (st); + + block_ordinal = current_block_ordinal (); + st->stat.st_size = 0; /* force 0 size on symlink */ + header = start_header (st); + if (!header) + return; + tar_copy_str (header->header.linkname, st->link_name, NAME_FIELD_SIZE); + header->header.typeflag = SYMTYPE; + finish_header (st, header, block_ordinal); + /* nothing more to do to it */ + + if (remove_files_option) + { + if (unlink (p) == -1) + unlink_error (p); + } + file_count_links (st); + return; + } +#endif + else if (S_ISCHR (st->stat.st_mode)) + type = CHRTYPE; + else if (S_ISBLK (st->stat.st_mode)) + type = BLKTYPE; + else if (S_ISFIFO (st->stat.st_mode)) + type = FIFOTYPE; + else if (S_ISSOCK (st->stat.st_mode)) + { + WARNOPT (WARN_FILE_IGNORED, + (0, 0, _("%s: socket ignored"), quotearg_colon (p))); + return; + } + else if (S_ISDOOR (st->stat.st_mode)) + { + WARNOPT (WARN_FILE_IGNORED, + (0, 0, _("%s: door ignored"), quotearg_colon (p))); + return; + } + else + { + unknown_file_error (p); + return; } if (archive_format == V7_FORMAT) @@ -1617,7 +1797,7 @@ dump_file0 (struct tar_stat_info *st, char *p, } void -dump_file (char *p, int top_level, dev_t parent_device) +dump_file (const char *p, int top_level, dev_t parent_device) { struct tar_stat_info st; tar_stat_init (&st);