X-Git-Url: https://git.dogcows.com/gitweb?a=blobdiff_plain;f=src%2Fcreate.c;h=3ba645869fe0cc985eddcf1d1311bf9dfe644eec;hb=ed7e2ec8a334cc534db981fe2fd0e02b3d39bf62;hp=7c2b49f13f588e82f445f72ec5d846953a33e3c7;hpb=6290150c47daea1d034f2293756b623169ad1b9b;p=chaz%2Ftar diff --git a/src/create.c b/src/create.c index 7c2b49f..3ba6458 100644 --- a/src/create.c +++ b/src/create.c @@ -33,6 +33,8 @@ struct utimbuf }; #endif +#include + #include "common.h" #ifndef MSDOS @@ -49,127 +51,286 @@ struct link struct link *next; dev_t dev; ino_t ino; - short linkcount; char name[1]; }; -struct link *linklist = NULL; /* points to first link in list */ +static struct link *linklist; /* points to first link in list */ - -/*------------------------------------------------------------------------. -| Convert VALUE (with substitute SUBSTITUTE if VALUE is out of range) | -| into a size-SIZE field at WHERE, including a | -| trailing space. For example, 3 for SIZE means two digits and a space. | -| | -| We assume the trailing NUL is already there and don't fill it in. This | -| fact is used by start_header and finish_header, so don't change it! | -`------------------------------------------------------------------------*/ - -/* Output VALUE in octal, using SUBSTITUTE if value won't fit. +/* The maximum uintmax_t value that can be represented with DIGITS digits, + assuming that each digit is BITS_PER_DIGIT wide. */ +#define MAX_VAL_WITH_DIGITS(digits, bits_per_digit) \ + ((digits) * (bits_per_digit) < sizeof (uintmax_t) * CHAR_BIT \ + ? ((uintmax_t) 1 << ((digits) * (bits_per_digit))) - 1 \ + : (uintmax_t) -1) + +/* Convert VALUE to an octal representation suitable for tar headers. Output to buffer WHERE with size SIZE. - TYPE is the kind of value being output (useful for diagnostics). - Prefer SIZE - 1 octal digits (with leading '0's), followed by '\0'; - but if SIZE octal digits would fit, omit the '\0'. */ + The result is undefined if SIZE is 0 or if VALUE is too large to fit. */ static void -to_oct (uintmax_t value, uintmax_t substitute, char *where, size_t size, const char *type) +to_octal (uintmax_t value, char *where, size_t size) { uintmax_t v = value; size_t i = size; -# define MAX_OCTAL_VAL_WITH_DIGITS(digits) \ - ((digits) * 3 < sizeof (uintmax_t) * CHAR_BIT \ - ? ((uintmax_t) 1 << ((digits) * 3)) - 1 \ - : (uintmax_t) -1) + do + { + where[--i] = '0' + (v & ((1 << LG_8) - 1)); + v >>= LG_8; + } + while (i); +} - /* Output a trailing NUL unless the value is too large. */ - if (value <= MAX_OCTAL_VAL_WITH_DIGITS (size - 1)) - where[--i] = '\0'; +/* Convert NEGATIVE VALUE to a base-256 representation suitable for + tar headers. NEGATIVE is 1 if VALUE was negative before being cast + to uintmax_t, 0 otherwise. Output to buffer WHERE with size SIZE. + The result is undefined if SIZE is 0 or if VALUE is too large to + fit. */ - /* Produce the digits -- at least one. */ +static void +to_base256 (int negative, uintmax_t value, char *where, size_t size) +{ + uintmax_t v = value; + uintmax_t propagated_sign_bits = + ((uintmax_t) - negative << (CHAR_BIT * sizeof v - LG_256)); + size_t i = size; do { - where[--i] = '0' + (int) (v & 7); /* one octal digit */ - v >>= 3; + where[--i] = v & ((1 << LG_256) - 1); + v = propagated_sign_bits | (v >> LG_256); + } + while (i); +} + +/* 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 + negative before being cast to uintmax_t; its original bitpattern + can be deduced from VALSIZE, its original size before casting. + TYPE is the kind of value being output (useful for diagnostics). + Prefer the POSIX format of SIZE - 1 octal digits (with leading zero + digits), followed by '\0'. If this won't work, and if GNU or + OLDGNU format is allowed, use '\200' followed by base-256, or (if + NEGATIVE is nonzero) '\377' followed by two's complement base-256. + If neither format works, use SUBSTITUTE (...) instead. Pass to + SUBSTITUTE the address of an 0-or-1 flag recording whether the + substitute value is negative. */ + +static void +to_chars (int negative, uintmax_t value, size_t valsize, + uintmax_t (*substitute) PARAMS ((int *)), + char *where, size_t size, const char *type) +{ + int base256_allowed = (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); + } + + /* 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); } - while (i != 0 && v != 0); - /* Leading zeros, if necessary. */ - while (i != 0) - where[--i] = '0'; + /* 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); + } - if (v != 0) + /* Otherwise, output a substitute value if possible (with a + warning), and an error message if not. */ + else { - uintmax_t maxval = MAX_OCTAL_VAL_WITH_DIGITS (size); - char buf1[UINTMAX_STRSIZE_BOUND]; - char buf2[UINTMAX_STRSIZE_BOUND]; - char buf3[UINTMAX_STRSIZE_BOUND]; - char *value_string = STRINGIFY_BIGINT (value, buf1); - char *maxval_string = STRINGIFY_BIGINT (maxval, buf2); + 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"; + + if (negative) + { + char *p = STRINGIFY_BIGINT (- value, valbuf + 1); + *--p = '-'; + value_string = p; + } + else + value_string = STRINGIFY_BIGINT (value, valbuf); + if (substitute) { - substitute &= maxval; - WARN ((0, 0, _("%s value %s too large (max=%s); substituting %s"), - type, value_string, maxval_string, - STRINGIFY_BIGINT (substitute, buf3))); - to_oct (substitute, (uintmax_t) 0, where, size, type); + int negsub; + uintmax_t sub = substitute (&negsub) & maxval; + 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); } else - ERROR ((0, 0, _("%s value %s too large (max=%s)"), - type, value_string, maxval_string)); + ERROR ((0, 0, _("value %s out of %s range %s..%s"), + value_string, type, minval_string, maxval_string)); } } -#ifndef GID_NOBODY -#define GID_NOBODY 0 + +static uintmax_t +gid_substitute (int *negative) +{ + gid_t r; +#ifdef GID_NOBODY + r = GID_NOBODY; +#else + static gid_t gid_nobody; + if (!gid_nobody && !gname_to_gid ("nobody", &gid_nobody)) + gid_nobody = -2; + r = gid_nobody; #endif + *negative = r < 0; + return r; +} + void -gid_to_oct (gid_t v, char *p, size_t s) +gid_to_chars (gid_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) GID_NOBODY, p, s, "gid_t"); + to_chars (v < 0, (uintmax_t) v, sizeof v, gid_substitute, p, s, "gid_t"); } + void -major_to_oct (major_t v, char *p, size_t s) +major_to_chars (major_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) 0, p, s, "major_t"); + to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "major_t"); } + void -minor_to_oct (minor_t v, char *p, size_t s) +minor_to_chars (minor_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) 0, p, s, "minor_t"); + to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "minor_t"); } + void -mode_to_oct (mode_t v, char *p, size_t s) +mode_to_chars (mode_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) 0, p, s, "mode_t"); + /* In the common case where the internal and external mode bits are the same, + and we are not using POSIX or GNU format, + propagate all unknown bits to the external mode. + This matches historical practice. + Otherwise, just copy the bits we know about. */ + int negative; + uintmax_t u; + if (S_ISUID == TSUID && S_ISGID == TSGID && S_ISVTX == TSVTX + && S_IRUSR == TUREAD && S_IWUSR == TUWRITE && S_IXUSR == TUEXEC + && S_IRGRP == TGREAD && S_IWGRP == TGWRITE && S_IXGRP == TGEXEC + && S_IROTH == TOREAD && S_IWOTH == TOWRITE && S_IXOTH == TOEXEC + && archive_format != POSIX_FORMAT + && archive_format != GNU_FORMAT) + { + negative = v < 0; + u = v; + } + else + { + negative = 0; + u = ((v & S_ISUID ? TSUID : 0) + | (v & S_ISGID ? TSGID : 0) + | (v & S_ISVTX ? TSVTX : 0) + | (v & S_IRUSR ? TUREAD : 0) + | (v & S_IWUSR ? TUWRITE : 0) + | (v & S_IXUSR ? TUEXEC : 0) + | (v & S_IRGRP ? TGREAD : 0) + | (v & S_IWGRP ? TGWRITE : 0) + | (v & S_IXGRP ? TGEXEC : 0) + | (v & S_IROTH ? TOREAD : 0) + | (v & S_IWOTH ? TOWRITE : 0) + | (v & S_IXOTH ? TOEXEC : 0)); + } + to_chars (negative, u, sizeof v, 0, p, s, "mode_t"); } + void -off_to_oct (off_t v, char *p, size_t s) +off_to_chars (off_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) 0, p, s, "off_t"); + to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "off_t"); } + void -size_to_oct (size_t v, char *p, size_t s) +size_to_chars (size_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) 0, p, s, "size_t"); + to_chars (0, (uintmax_t) v, sizeof v, 0, p, s, "size_t"); } + void -time_to_oct (time_t v, char *p, size_t s) +time_to_chars (time_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) 0, p, s, "time_t"); + to_chars (v < 0, (uintmax_t) v, sizeof v, 0, p, s, "time_t"); } -#ifndef UID_NOBODY -#define UID_NOBODY 0 + +static uintmax_t +uid_substitute (int *negative) +{ + uid_t r; +#ifdef UID_NOBODY + r = UID_NOBODY; +#else + static uid_t uid_nobody; + if (!uid_nobody && !uname_to_uid ("nobody", &uid_nobody)) + uid_nobody = -2; + r = uid_nobody; #endif + *negative = r < 0; + return r; +} + void -uid_to_oct (uid_t v, char *p, size_t s) +uid_to_chars (uid_t v, char *p, size_t s) { - to_oct ((uintmax_t) v, (uintmax_t) UID_NOBODY, p, s, "uid_t"); + to_chars (v < 0, (uintmax_t) v, sizeof v, uid_substitute, p, s, "uid_t"); } + void -uintmax_to_oct (uintmax_t v, char *p, size_t s) +uintmax_to_chars (uintmax_t v, char *p, size_t s) { - to_oct (v, (uintmax_t) 0, p, s, "uintmax_t"); + to_chars (0, v, sizeof v, 0, p, s, "uintmax_t"); } /* Writing routines. */ @@ -186,7 +347,7 @@ clear_buffer (char *buffer) } /*-------------------------------------------------------------------------. -| Write the EOT block(s). We actually zero at least one block, through | +| Write the EOT block(s). We zero at least two blocks, through | | the end of the record. Old tar, as previous versions of GNU tar, writes | | garbage after two zeroed blocks. | `-------------------------------------------------------------------------*/ @@ -195,14 +356,11 @@ void write_eot (void) { union block *pointer = find_next_block (); - - if (pointer) - { - size_t space = available_space_after (pointer); - - memset (pointer->buffer, 0, space); - set_next_block_after (pointer); - } + memset (pointer->buffer, 0, BLOCKSIZE); + set_next_block_after (pointer); + pointer = find_next_block (); + memset (pointer->buffer, 0, available_space_after (pointer)); + set_next_block_after (pointer); } /*-----------------------------------------------------. @@ -248,10 +406,8 @@ write_long (const char *p, char type) /* Header handling. */ -/*---------------------------------------------------------------------. -| Make a header block for the file name whose stat info is st. Return | -| header pointer for success, NULL if the name is too long. | -`---------------------------------------------------------------------*/ +/* Make a header block for the file whose stat info is st, + and return its address. */ static union block * start_header (const char *name, struct stat *st) @@ -260,35 +416,42 @@ start_header (const char *name, struct stat *st) if (!absolute_names_option) { - static int warned_once = 0; + size_t prefix_len = FILESYSTEM_PREFIX_LEN (name); -#if MSDOS - if (name[1] == ':') + if (prefix_len) { - name += 2; + static int warned_once; if (!warned_once) { warned_once = 1; - WARN ((0, 0, _("Removing drive spec from names in the archive"))); + WARN ((0, 0, _("Removing `%.*s' prefix from member names"), + (int) prefix_len, name)); } + name += prefix_len; } -#endif while (*name == '/') { - name++; /* force relative path */ + static int warned_once; if (!warned_once) { warned_once = 1; - WARN ((0, 0, _("\ -Removing leading `/' from absolute path names in the archive"))); + WARN ((0, 0, _("Removing leading `/' from member names"))); } + name++; } - } - /* Check the file name and put it in the block. */ + { + static int warned_once; + if (! warned_once && contains_dot_dot (name)) + { + warned_once = 1; + WARN ((0, 0, _("Member names contain `..'"))); + } + } + } - if (strlen (name) >= (size_t) NAME_FIELD_SIZE) + if (sizeof header->header.name <= strlen (name)) write_long (name, GNUTYPE_LONGNAME); header = find_next_block (); memset (header->buffer, 0, sizeof (union block)); @@ -305,7 +468,7 @@ Removing leading `/' from absolute path names in the archive"))); if (group_option != (gid_t) -1) st->st_gid = group_option; if (mode_option) - st->st_mode = ((st->st_mode & S_IFMT) + st->st_mode = ((st->st_mode & ~MODE_ALL) | mode_adjust (st->st_mode, mode_option)); /* Paul Eggert tried the trivial test ($WRITER cf a b; $READER tvf a) @@ -332,27 +495,26 @@ Removing leading `/' from absolute path names in the archive"))); acceptor for Paul's test. */ if (archive_format == V7_FORMAT) - MODE_TO_OCT (st->st_mode & 07777, header->header.mode); + MODE_TO_CHARS (st->st_mode & MODE_ALL, header->header.mode); else - MODE_TO_OCT (st->st_mode, header->header.mode); + MODE_TO_CHARS (st->st_mode, header->header.mode); - UID_TO_OCT (st->st_uid, header->header.uid); - GID_TO_OCT (st->st_gid, header->header.gid); - OFF_TO_OCT (st->st_size, header->header.size); - TIME_TO_OCT (st->st_mtime, header->header.mtime); + UID_TO_CHARS (st->st_uid, header->header.uid); + GID_TO_CHARS (st->st_gid, header->header.gid); + OFF_TO_CHARS (st->st_size, header->header.size); + TIME_TO_CHARS (st->st_mtime, header->header.mtime); if (incremental_option) if (archive_format == OLDGNU_FORMAT) { - TIME_TO_OCT (st->st_atime, header->oldgnu_header.atime); - TIME_TO_OCT (st->st_ctime, header->oldgnu_header.ctime); + TIME_TO_CHARS (st->st_atime, header->oldgnu_header.atime); + TIME_TO_CHARS (st->st_ctime, header->oldgnu_header.ctime); } header->header.typeflag = archive_format == V7_FORMAT ? AREGTYPE : REGTYPE; switch (archive_format) { - case DEFAULT_FORMAT: case V7_FORMAT: break; @@ -366,6 +528,9 @@ Removing leading `/' from absolute path names in the archive"))); strncpy (header->header.magic, TMAGIC, TMAGLEN); strncpy (header->header.version, TVERSION, TVERSLEN); break; + + default: + abort (); } if (archive_format == V7_FORMAT || numeric_owner_option) @@ -393,27 +558,25 @@ finish_header (union block *header) int sum; char *p; - memcpy (header->header.chksum, CHKBLANKS, sizeof (header->header.chksum)); + memcpy (header->header.chksum, CHKBLANKS, sizeof header->header.chksum); sum = 0; p = header->buffer; - for (i = sizeof (*header); i-- != 0; ) + for (i = sizeof *header; i-- != 0; ) /* We can't use unsigned char here because of old compilers, e.g. V7. */ sum += 0xFF & *p++; /* Fill in the checksum field. It's formatted differently from the other fields: it has [6] digits, a null, then a space -- rather than - digits, then a null. We use to_oct. - The final space is already there, from checksumming, - and to_oct doesn't modify it. + digits, then a null. We use to_chars. + The final space is already there, from + checksumming, and to_chars doesn't modify it. This is a fast way to do: sprintf(header->header.chksum, "%6o", sum); */ - uintmax_to_oct ((uintmax_t) sum, header->header.chksum, 7); - - set_next_block_after (header); + uintmax_to_chars ((uintmax_t) sum, header->header.chksum, 7); if (verbose_option && header->header.typeflag != GNUTYPE_LONGLINK @@ -426,6 +589,8 @@ finish_header (union block *header) current_format = archive_format; print_header (); } + + set_next_block_after (header); } /* Sparse file processing. */ @@ -460,8 +625,7 @@ init_sparsearray (void) /* Make room for our scratch space -- initially is 10 elts long. */ - sparsearray = (struct sp_array *) - xmalloc (sp_array_size * sizeof (struct sp_array)); + sparsearray = xmalloc (sp_array_size * sizeof (struct sp_array)); for (counter = 0; counter < sp_array_size; counter++) { sparsearray[counter].offset = 0; @@ -518,15 +682,14 @@ deal_with_sparse (char *name, union block *header) init_sparsearray (); clear_buffer (buffer); - while (count = full_read (file, buffer, sizeof buffer), count != 0) + while (count = safe_read (file, buffer, sizeof buffer), count != 0) { /* Realloc the scratch area as necessary. FIXME: should reallocate only at beginning of a new instance of non-zero data. */ if (sparse_index > sp_array_size - 1) { - - sparsearray = (struct sp_array *) + sparsearray = xrealloc (sparsearray, 2 * sp_array_size * sizeof (struct sp_array)); sp_array_size *= 2; @@ -615,17 +778,18 @@ finish_sparse_file (int file, off_t *sizeleft, off_t fullsize, char *name) ERROR ((0, 0, _("Wrote %s of %s bytes to file %s"), STRINGIFY_BIGINT (fullsize - *sizeleft, buf1), - STRINGIFY_BIGINT (fullsize, buf2), - name)); + STRINGIFY_BIGINT (fullsize, buf2), quote (name))); break; } - if (lseek (file, sparsearray[sparse_index++].offset, 0) < 0) + if (lseek (file, sparsearray[sparse_index++].offset, SEEK_SET) < 0) { char buf[UINTMAX_STRSIZE_BOUND]; - ERROR ((0, errno, _("lseek error at byte %s in file %s"), - STRINGIFY_BIGINT (sparsearray[sparse_index - 1].offset, buf), - name)); + int e = errno; + ERROR ((0, e, _("lseek error at byte %s in file %s"), + STRINGIFY_BIGINT (sparsearray[sparse_index - 1].offset, + buf), + quote (name))); break; } @@ -637,7 +801,7 @@ finish_sparse_file (int file, off_t *sizeleft, off_t fullsize, char *name) #if 0 if (amount_read) { - count = full_read (file, start->buffer + amount_read, + count = safe_read (file, start->buffer + amount_read, BLOCKSIZE - amount_read); bufsize -= BLOCKSIZE - amount_read; amount_read = 0; @@ -648,14 +812,15 @@ finish_sparse_file (int file, off_t *sizeleft, off_t fullsize, char *name) #endif /* Store the data. */ - count = full_read (file, start->buffer, BLOCKSIZE); + count = safe_read (file, start->buffer, BLOCKSIZE); if (count < 0) { char buf[UINTMAX_STRSIZE_BOUND]; - ERROR ((0, errno, _("\ -Read error at byte %s, reading %lu bytes, in file %s"), + int e = errno; + ERROR ((0, e, + _("Read error at byte %s, reading %lu bytes, in file %s"), STRINGIFY_BIGINT (fullsize - *sizeleft, buf), - (unsigned long) bufsize, name)); + (unsigned long) bufsize, quote (name))); return 1; } bufsize -= count; @@ -669,18 +834,18 @@ Read error at byte %s, reading %lu bytes, in file %s"), char buffer[BLOCKSIZE]; clear_buffer (buffer); - count = full_read (file, buffer, bufsize); + count = safe_read (file, buffer, bufsize); memcpy (start->buffer, buffer, BLOCKSIZE); } if (count < 0) { char buf[UINTMAX_STRSIZE_BOUND]; - - ERROR ((0, errno, + int e = errno; + ERROR ((0, e, _("Read error at byte %s, reading %lu bytes, in file %s"), STRINGIFY_BIGINT (fullsize - *sizeleft, buf), - (unsigned long) bufsize, name)); + (unsigned long) bufsize, quote (name))); return 1; } #if 0 @@ -691,8 +856,7 @@ Read error at byte %s, reading %lu bytes, in file %s"), if (count != bufsize) { ERROR ((0, 0, - _("File %s shrunk, padding with zeros"), - name)); + _("File %s shrunk, padding with zeros"), quote (name))); return 1; } start = find_next_block (); @@ -733,61 +897,62 @@ create_archive (void) collect_and_sort_names (); while (p = name_from_list (), p) - dump_file (p, (dev_t) -1, 1); + if (!excluded_name (p)) + dump_file (p, -1, (dev_t) 0); blank_name_list (); while (p = name_from_list (), p) - { - strcpy (buffer, p); - if (p[strlen (p) - 1] != '/') - strcat (buffer, "/"); - bufp = buffer + strlen (buffer); - for (q = gnu_list_name->dir_contents; - q && *q; - q += strlen (q) + 1) - { - if (*q == 'Y') - { - strcpy (bufp, q + 1); - dump_file (buffer, (dev_t) -1, 1); - } - } - } + if (!excluded_name (p)) + { + strcpy (buffer, p); + if (p[strlen (p) - 1] != '/') + strcat (buffer, "/"); + bufp = buffer + strlen (buffer); + q = gnu_list_name->dir_contents; + if (q) + for (; *q; q += strlen (q) + 1) + if (*q == 'Y') + { + strcpy (bufp, q + 1); + dump_file (buffer, -1, (dev_t) 0); + } + } free (buffer); } else { while (p = name_next (1), p) - dump_file (p, (dev_t) -1, 1); + if (!excluded_name (p)) + dump_file (p, 1, (dev_t) 0); } write_eot (); close_archive (); if (listed_incremental_option) - write_dir_file (); + write_directory_file (); } -/*----------------------------------------------------------------------. -| Dump a single file. Recurse on directories. Result is nonzero for | -| success. P is file name to dump. PARENT_DEVICE is device our parent | -| directory was on. TOP_LEVEL tells wether we are a toplevel call. | -| | -| Sets global CURRENT_STAT to stat output for this file. | -`----------------------------------------------------------------------*/ +/* Dump a single file, recursing on directories. P is the file name + to dump. TOP_LEVEL tells whether this is a top-level call; zero + means no, positive means yes, and negative means an incremental + dump. PARENT_DEVICE is the device of P's + parent directory; it is examined only if TOP_LEVEL is zero. + + Set global CURRENT_STAT to stat output for this file. */ /* FIXME: One should make sure that for *every* path leading to setting exit_status to failure, a clear diagnostic has been issued. */ void -dump_file (char *p, dev_t parent_device, int top_level) +dump_file (char *p, int top_level, dev_t parent_device) { union block *header; char type; union block *exhdr; char save_typeflag; + time_t original_ctime; struct utimbuf restore_times; - off_t restore_size; /* FIXME: `header' and `upperbound' might be used uninitialized in this function. Reported by Bruno Haible. */ @@ -795,27 +960,18 @@ dump_file (char *p, dev_t parent_device, int top_level) if (interactive_option && !confirm ("add", p)) return; - /* Use stat if following (rather than dumping) 4.2BSD's symbolic links. - Otherwise, use lstat (which falls back to stat if no symbolic links). */ - - if (dereference_option != 0 -#if STX_HIDDEN && !_LARGE_FILES /* AIX */ - ? statx (p, ¤t_stat, STATSIZE, STX_HIDDEN) - : statx (p, ¤t_stat, STATSIZE, STX_HIDDEN | STX_LINK) -#else - ? stat (p, ¤t_stat) : lstat (p, ¤t_stat) -#endif - ) + if (deref_stat (dereference_option, p, ¤t_stat) != 0) { - WARN ((0, errno, _("Cannot add file %s"), p)); - if (!ignore_failed_read_option) - exit_status = TAREXIT_FAILURE; + if (ignore_failed_read_option) + stat_error (p); + else + stat_warn (p); return; } + original_ctime = current_stat.st_ctime; restore_times.actime = current_stat.st_atime; restore_times.modtime = current_stat.st_mtime; - restore_size = current_stat.st_size; #ifdef S_ISHIDDEN if (S_ISHIDDEN (current_stat.st_mode)) @@ -830,15 +986,16 @@ dump_file (char *p, dev_t parent_device, int top_level) } #endif - /* See if we only want new files, and check if this one is too old to + /* See if we want only new files, and check if this one is too old to put in the archive. */ - if (!incremental_option && !S_ISDIR (current_stat.st_mode) + if ((0 < top_level || !incremental_option) + && !S_ISDIR (current_stat.st_mode) && current_stat.st_mtime < newer_mtime_option && (!after_date_option || current_stat.st_ctime < newer_ctime_option)) { - if (parent_device == (dev_t) -1) - WARN ((0, 0, _("%s: is unchanged; not dumped"), p)); + if (0 < top_level) + WARN ((0, 0, _("%s is unchanged; not dumped"), quote (p))); /* FIXME: recheck this return. */ return; } @@ -848,446 +1005,12 @@ dump_file (char *p, dev_t parent_device, int top_level) if (ar_dev && current_stat.st_dev == ar_dev && current_stat.st_ino == ar_ino) { - WARN ((0, 0, _("%s is the archive; not dumped"), p)); + WARN ((0, 0, _("%s is the archive; not dumped"), quote (p))); return; } #endif - /* Check for multiple links. - - We maintain a list of all such files that we've written so far. Any - time we see another, we check the list and avoid dumping the data - again if we've done it once already. */ - - if (current_stat.st_nlink > 1 - && (S_ISREG (current_stat.st_mode) -#ifdef S_ISCTG - || S_ISCTG (current_stat.st_mode) -#endif -#ifdef S_ISCHR - || S_ISCHR (current_stat.st_mode) -#endif -#ifdef S_ISBLK - || S_ISBLK (current_stat.st_mode) -#endif -#ifdef S_ISFIFO - || S_ISFIFO (current_stat.st_mode) -#endif - )) - { - struct link *lp; - - /* FIXME: First quick and dirty. Hashing, etc later. */ - - for (lp = linklist; lp; lp = lp->next) - if (lp->ino == current_stat.st_ino && lp->dev == current_stat.st_dev) - { - char *link_name = lp->name; - - /* We found a link. */ - - while (!absolute_names_option && *link_name == '/') - { - static int warned_once = 0; - - if (!warned_once) - { - warned_once = 1; - WARN ((0, 0, _("\ -Removing leading `/' from absolute links"))); - } - link_name++; - } - if (strlen (link_name) >= NAME_FIELD_SIZE) - write_long (link_name, GNUTYPE_LONGLINK); - assign_string (¤t_link_name, link_name); - - current_stat.st_size = 0; - header = start_header (p, ¤t_stat); - if (header == NULL) - { - exit_status = TAREXIT_FAILURE; - return; - } - strncpy (header->header.linkname, - link_name, NAME_FIELD_SIZE); - - /* Force null truncated. */ - - header->header.linkname[NAME_FIELD_SIZE - 1] = 0; - - header->header.typeflag = LNKTYPE; - finish_header (header); - - /* FIXME: Maybe remove from list after all links found? */ - - if (remove_files_option) - if (unlink (p) == -1) - ERROR ((0, errno, _("Cannot remove %s"), p)); - - /* We dumped it. */ - return; - } - - /* Not found. Add it to the list of possible links. */ - - lp = (struct link *) - xmalloc ((size_t) (sizeof (struct link) + strlen (p))); - lp->ino = current_stat.st_ino; - lp->dev = current_stat.st_dev; - strcpy (lp->name, p); - lp->next = linklist; - linklist = lp; - } - - /* This is not a link to a previously dumped file, so dump it. */ - - if (S_ISREG (current_stat.st_mode) -#ifdef S_ISCTG - || S_ISCTG (current_stat.st_mode) -#endif - ) - { - int f; /* file descriptor */ - size_t bufsize; - ssize_t count; - off_t sizeleft; - union block *start; - int header_moved; - char isextended = 0; - int upperbound; -#if 0 - static int cried_once = 0; -#endif - - header_moved = 0; - - if (sparse_option) - { - /* Check the size of the file against the number of blocks - allocated for it, counting both data and indirect blocks. - If there is a smaller number of blocks that would be - necessary to accommodate a file of this size, this is safe - to say that we have a sparse file: at least one of those - blocks in the file is just a useless hole. For sparse - files not having more hole blocks than indirect blocks, the - sparseness will go undetected. */ - - /* Bruno Haible sent me these statistics for Linux. It seems - that some filesystems count indirect blocks in st_blocks, - while others do not seem to: - - minix-fs tar: size=7205, st_blocks=18 and ST_NBLOCKS=18 - extfs tar: size=7205, st_blocks=18 and ST_NBLOCKS=18 - ext2fs tar: size=7205, st_blocks=16 and ST_NBLOCKS=16 - msdos-fs tar: size=7205, st_blocks=16 and ST_NBLOCKS=16 - - Dick Streefland reports the previous numbers as misleading, - because ext2fs use 12 direct blocks, while minix-fs uses only - 6 direct blocks. Dick gets: - - ext2 size=20480 ls listed blocks=21 - minix size=20480 ls listed blocks=21 - msdos size=20480 ls listed blocks=20 - - It seems that indirect blocks *are* included in st_blocks. - The minix filesystem does not account for phantom blocks in - st_blocks, so `du' and `ls -s' give wrong results. So, the - --sparse option would not work on a minix filesystem. */ - - if (ST_NBLOCKS (current_stat) - < (current_stat.st_size / ST_NBLOCKSIZE - + (current_stat.st_size % ST_NBLOCKSIZE != 0))) - { - off_t filesize = current_stat.st_size; - int counter; - - header = start_header (p, ¤t_stat); - if (header == NULL) - { - exit_status = TAREXIT_FAILURE; - return; - } - header->header.typeflag = GNUTYPE_SPARSE; - header_moved = 1; - - /* Call the routine that figures out the layout of the - sparse file in question. UPPERBOUND is the index of the - last element of the "sparsearray," i.e., the number of - elements it needed to describe the file. */ - - upperbound = deal_with_sparse (p, header); - - /* See if we'll need an extended header later. */ - - if (upperbound > SPARSES_IN_OLDGNU_HEADER - 1) - header->oldgnu_header.isextended = 1; - - /* We store the "real" file size so we can show that in - case someone wants to list the archive, i.e., tar tvf - . It might be kind of disconcerting if the - shrunken file size was the one that showed up. */ - - OFF_TO_OCT (current_stat.st_size, - header->oldgnu_header.realsize); - - /* This will be the new "size" of the file, i.e., the size - of the file minus the blocks of holes that we're - skipping over. */ - - find_new_file_size (&filesize, upperbound); - current_stat.st_size = filesize; - OFF_TO_OCT (filesize, header->header.size); - - for (counter = 0; counter < SPARSES_IN_OLDGNU_HEADER; counter++) - { - if (!sparsearray[counter].numbytes) - break; - - OFF_TO_OCT (sparsearray[counter].offset, - header->oldgnu_header.sp[counter].offset); - SIZE_TO_OCT (sparsearray[counter].numbytes, - header->oldgnu_header.sp[counter].numbytes); - } - - } - } - else - upperbound = SPARSES_IN_OLDGNU_HEADER - 1; - - sizeleft = current_stat.st_size; - - /* Don't bother opening empty, world readable files. Also do not open - files when archive is meant for /dev/null. */ - - if (dev_null_output - || (sizeleft == 0 && 0444 == (0444 & current_stat.st_mode))) - f = -1; - else - { - f = open (p, O_RDONLY | O_BINARY); - if (f < 0) - { - WARN ((0, errno, _("Cannot add file %s"), p)); - if (!ignore_failed_read_option) - exit_status = TAREXIT_FAILURE; - return; - } - } - - /* If the file is sparse, we've already taken care of this. */ - - if (!header_moved) - { - header = start_header (p, ¤t_stat); - if (header == NULL) - { - if (f >= 0) - close (f); - exit_status = TAREXIT_FAILURE; - return; - } - } -#ifdef S_ISCTG - /* Mark contiguous files, if we support them. */ - - if (archive_format != V7_FORMAT && S_ISCTG (current_stat.st_mode)) - header->header.typeflag = CONTTYPE; -#endif - isextended = header->oldgnu_header.isextended; - save_typeflag = header->header.typeflag; - finish_header (header); - if (isextended) - { -#if 0 - int sum = 0; -#endif - int counter; -#if 0 - union block *exhdr; - int arraybound = SPARSES_IN_SPARSE_HEADER; -#endif - /* static */ int index_offset = SPARSES_IN_OLDGNU_HEADER; - - extend: - exhdr = find_next_block (); - - if (exhdr == NULL) - { - exit_status = TAREXIT_FAILURE; - return; - } - memset (exhdr->buffer, 0, BLOCKSIZE); - for (counter = 0; counter < SPARSES_IN_SPARSE_HEADER; counter++) - { - if (counter + index_offset > upperbound) - break; - - SIZE_TO_OCT (sparsearray[counter + index_offset].numbytes, - exhdr->sparse_header.sp[counter].numbytes); - OFF_TO_OCT (sparsearray[counter + index_offset].offset, - exhdr->sparse_header.sp[counter].offset); - } - set_next_block_after (exhdr); -#if 0 - sum += counter; - if (sum < upperbound) - goto extend; -#endif - if (index_offset + counter <= upperbound) - { - index_offset += counter; - exhdr->sparse_header.isextended = 1; - goto extend; - } - - } - if (save_typeflag == GNUTYPE_SPARSE) - { - if (finish_sparse_file (f, &sizeleft, current_stat.st_size, p)) - goto padit; - } - else - while (sizeleft > 0) - { - if (multi_volume_option) - { - assign_string (&save_name, p); - save_sizeleft = sizeleft; - save_totsize = current_stat.st_size; - } - start = find_next_block (); - - bufsize = available_space_after (start); - - if (sizeleft < bufsize) - { - /* Last read -- zero out area beyond. */ - - bufsize = sizeleft; - count = bufsize % BLOCKSIZE; - if (count) - memset (start->buffer + sizeleft, 0, - (size_t) (BLOCKSIZE - count)); - } - if (f < 0) - count = bufsize; - else - count = full_read (f, start->buffer, bufsize); - if (count < 0) - { - char buf[UINTMAX_STRSIZE_BOUND]; - ERROR ((0, errno, _("\ -Read error at byte %s, reading %lu bytes, in file %s"), - STRINGIFY_BIGINT (current_stat.st_size - sizeleft, - buf), - (unsigned long) bufsize, p)); - goto padit; - } - sizeleft -= count; - - /* This is nonportable (the type of set_next_block_after's arg). */ - - set_next_block_after (start + (count - 1) / BLOCKSIZE); - - if (count == bufsize) - continue; - else - { - char buf[UINTMAX_STRSIZE_BOUND]; - ERROR ((0, 0, - _("File %s shrunk by %s bytes, padding with zeros"), - p, STRINGIFY_BIGINT (sizeleft, buf))); - goto padit; /* short read */ - } - } - - if (multi_volume_option) - assign_string (&save_name, NULL); - - if (f >= 0) - { - struct stat final_stat; - if (fstat (f, &final_stat) != 0) - ERROR ((0, errno, "%s: fstat", p)); - else if (final_stat.st_mtime != restore_times.modtime - || final_stat.st_size != restore_size) - ERROR ((0, errno, _("%s: file changed as we read it"), p)); - if (close (f) != 0) - ERROR ((0, errno, _("%s: close"), p)); - if (atime_preserve_option) - utime (p, &restore_times); - } - if (remove_files_option) - { - if (unlink (p) == -1) - ERROR ((0, errno, _("Cannot remove %s"), p)); - } - return; - - /* File shrunk or gave error, pad out tape to match the size we - specified in the header. */ - - padit: - while (sizeleft > 0) - { - save_sizeleft = sizeleft; - start = find_next_block (); - memset (start->buffer, 0, BLOCKSIZE); - set_next_block_after (start); - sizeleft -= BLOCKSIZE; - } - if (multi_volume_option) - assign_string (&save_name, NULL); - if (f >= 0) - { - close (f); - if (atime_preserve_option) - utime (p, &restore_times); - } - return; - } - -#ifdef S_ISLNK - else if (S_ISLNK (current_stat.st_mode)) - { - int size; - char *buffer = (char *) alloca (PATH_MAX + 1); - - size = readlink (p, buffer, PATH_MAX + 1); - if (size < 0) - { - WARN ((0, errno, _("Cannot add file %s"), p)); - if (!ignore_failed_read_option) - exit_status = TAREXIT_FAILURE; - return; - } - buffer[size] = '\0'; - if (size >= NAME_FIELD_SIZE) - write_long (buffer, GNUTYPE_LONGLINK); - assign_string (¤t_link_name, buffer); - - current_stat.st_size = 0; /* force 0 size on symlink */ - header = start_header (p, ¤t_stat); - if (header == NULL) - { - exit_status = TAREXIT_FAILURE; - return; - } - strncpy (header->header.linkname, buffer, NAME_FIELD_SIZE); - header->header.linkname[NAME_FIELD_SIZE - 1] = '\0'; - header->header.typeflag = SYMTYPE; - finish_header (header); /* nothing more to do to it */ - if (remove_files_option) - { - if (unlink (p) == -1) - ERROR ((0, errno, _("Cannot remove %s"), p)); - } - return; - } -#endif /* S_ISLNK */ - - else if (S_ISDIR (current_stat.st_mode)) + if (S_ISDIR (current_stat.st_mode)) { DIR *directory; struct dirent *entry; @@ -1296,18 +1019,15 @@ Read error at byte %s, reading %lu bytes, in file %s"), size_t len; dev_t our_device = current_stat.st_dev; - /* If this tar program is installed suid root, like for Amanda, the - access might look like denied, while it is not really. - - FIXME: I have the feeling this test is done too early. Couldn't it - just be bundled in later actions? I guess that the proper support - of --ignore-failed-read is the key of the current writing. */ + errno = 0; - if (access (p, R_OK) == -1 && geteuid () != 0) + directory = opendir (p); + if (! directory) { - WARN ((0, errno, _("Cannot add directory %s"), p)); - if (!ignore_failed_read_option) - exit_status = TAREXIT_FAILURE; + if (ignore_failed_read_option) + opendir_warn (p); + else + opendir_error (p); return; } @@ -1322,10 +1042,10 @@ Read error at byte %s, reading %lu bytes, in file %s"), namebuf[len++] = '/'; namebuf[len] = '\0'; - if (1) + if (! is_avoided_name (namebuf)) { - /* The "1" above used to be "archive_format != V7_FORMAT", GNU tar - was just not writing directory blocks at all. Daniel Trinkle + /* The condition above used to be "archive_format != V7_FORMAT". + GNU tar was not writing directory blocks at all. Daniel Trinkle writes: ``All old versions of tar I have ever seen have correctly archived an empty directory. The really old ones I checked included HP-UX 7 and Mt. Xinu More/BSD. There may be @@ -1346,11 +1066,6 @@ Read error at byte %s, reading %lu bytes, in file %s"), files, we'd better put the / on the name. */ header = start_header (namebuf, ¤t_stat); - if (header == NULL) - { - exit_status = TAREXIT_FAILURE; - return; /* eg name too long */ - } if (incremental_option) header->header.typeflag = GNUTYPE_DUMPDIR; @@ -1383,7 +1098,7 @@ Read error at byte %s, reading %lu bytes, in file %s"), p_buffer += tmp; } totsize++; - OFF_TO_OCT (totsize, header->header.size); + OFF_TO_CHARS (totsize, header->header.size); finish_header (header); p_buffer = buffer; sizeleft = totsize; @@ -1402,8 +1117,7 @@ Read error at byte %s, reading %lu bytes, in file %s"), bufsize = sizeleft; count = bufsize % BLOCKSIZE; if (count) - memset (start->buffer + sizeleft, 0, - (size_t) (BLOCKSIZE - count)); + memset (start->buffer + sizeleft, 0, BLOCKSIZE - count); } memcpy (start->buffer, p_buffer, bufsize); sizeleft -= bufsize; @@ -1411,7 +1125,7 @@ Read error at byte %s, reading %lu bytes, in file %s"), set_next_block_after (start + (bufsize - 1) / BLOCKSIZE); } if (multi_volume_option) - assign_string (&save_name, NULL); + assign_string (&save_name, 0); if (atime_preserve_option) utime (p, &restore_times); return; @@ -1430,31 +1144,18 @@ Read error at byte %s, reading %lu bytes, in file %s"), && parent_device != current_stat.st_dev) { if (verbose_option) - WARN ((0, 0, _("%s: On a different filesystem; not dumped"), p)); + WARN ((0, 0, _("%s is on a different filesystem; not dumped"), + quote (p))); return; } /* Now output all the files in the directory. */ - errno = 0; /* FIXME: errno should be read-only */ - - directory = opendir (p); - if (!directory) - { - ERROR ((0, errno, _("Cannot open directory %s"), p)); - return; - } - - /* Hack to remove "./" from the front of all the file names. */ - - if (len == 2 && namebuf[0] == '.' && namebuf[1] == '/') - len = 0; - /* FIXME: Should speed this up by cd-ing into the dir. */ - while (entry = readdir (directory), entry) + while (errno = 0, (entry = readdir (directory))) { - /* Skip `.' and `..'. */ + /* Skip `.', `..', and excluded file names. */ if (is_dot_or_dotdot (entry->d_name)) continue; @@ -1462,81 +1163,450 @@ Read error at byte %s, reading %lu bytes, in file %s"), if ((int) NAMLEN (entry) + len >= buflen) { buflen = len + NAMLEN (entry); - namebuf = (char *) xrealloc (namebuf, buflen + 1); -#if 0 - namebuf[len] = '\0'; - ERROR ((0, 0, _("File name %s%s too long"), - namebuf, entry->d_name)); - continue; -#endif + namebuf = xrealloc (namebuf, buflen + 1); } strcpy (namebuf + len, entry->d_name); - if (exclude_option && check_exclude (namebuf)) - continue; - dump_file (namebuf, our_device, 0); + if (!excluded_name (namebuf)) + dump_file (namebuf, 0, our_device); + } + + if (errno) + { + if (ignore_failed_read_option) + readdir_warn (p); + else + readdir_error (p); } - closedir (directory); + if (closedir (directory) != 0) + closedir_error (p); free (namebuf); if (atime_preserve_option) utime (p, &restore_times); return; } + else if (is_avoided_name (p)) + return; + else + { + /* Check for multiple links. -#ifdef S_ISCHR - else if (S_ISCHR (current_stat.st_mode)) - type = CHRTYPE; -#endif + We maintain a list of all such files that we've written so far. Any + time we see another, we check the list and avoid dumping the data + again if we've done it once already. */ -#ifdef S_ISBLK - else if (S_ISBLK (current_stat.st_mode)) - type = BLKTYPE; -#endif + if (1 < current_stat.st_nlink) + { + struct link *lp; - /* Avoid screwy apollo lossage where S_IFIFO == S_IFSOCK. */ + /* FIXME: First quick and dirty. Hashing, etc later. */ -#if (_ISP__M68K == 0) && (_ISP__A88K == 0) && defined(S_ISFIFO) - else if (S_ISFIFO (current_stat.st_mode)) - type = FIFOTYPE; -#endif + for (lp = linklist; lp; lp = lp->next) + if (lp->ino == current_stat.st_ino + && lp->dev == current_stat.st_dev) + { + char *link_name = lp->name; -#ifdef S_ISSOCK - else if (S_ISSOCK (current_stat.st_mode)) - type = FIFOTYPE; -#endif + /* We found a link. */ - else - goto unknown; + while (!absolute_names_option && *link_name == '/') + { + static int warned_once; + if (!warned_once) + { + warned_once = 1; + WARN ((0, 0, + _("Removing leading `/' from link names"))); + } + link_name++; + } + if (strlen (link_name) >= NAME_FIELD_SIZE) + write_long (link_name, GNUTYPE_LONGLINK); + assign_string (¤t_link_name, link_name); + + current_stat.st_size = 0; + header = start_header (p, ¤t_stat); + strncpy (header->header.linkname, + link_name, NAME_FIELD_SIZE); + + /* Force null truncated. */ + + header->header.linkname[NAME_FIELD_SIZE - 1] = 0; + + header->header.typeflag = LNKTYPE; + finish_header (header); + + /* FIXME: Maybe remove from list after all links found? */ + + if (remove_files_option) + if (unlink (p) == -1) + unlink_error (p); + + /* We dumped it. */ + return; + } + + /* Not found. Add it to the list of possible links. */ + + lp = xmalloc (sizeof (struct link) + strlen (p)); + lp->ino = current_stat.st_ino; + lp->dev = current_stat.st_dev; + strcpy (lp->name, p); + lp->next = linklist; + linklist = lp; + } + + /* This is not a link to a previously dumped file, so dump it. */ + + if (S_ISREG (current_stat.st_mode) + || S_ISCTG (current_stat.st_mode)) + { + int f; /* file descriptor */ + size_t bufsize; + ssize_t count; + off_t sizeleft; + union block *start; + int header_moved; + char isextended = 0; + int upperbound; + + header_moved = 0; + + if (sparse_option) + { + /* Check the size of the file against the number of blocks + allocated for it, counting both data and indirect blocks. + If there is a smaller number of blocks that would be + necessary to accommodate a file of this size, this is safe + to say that we have a sparse file: at least one of those + blocks in the file is just a useless hole. For sparse + files not having more hole blocks than indirect blocks, the + sparseness will go undetected. */ + + /* Bruno Haible sent me these statistics for Linux. It seems + that some filesystems count indirect blocks in st_blocks, + while others do not seem to: + + minix-fs tar: size=7205, st_blocks=18 and ST_NBLOCKS=18 + extfs tar: size=7205, st_blocks=18 and ST_NBLOCKS=18 + ext2fs tar: size=7205, st_blocks=16 and ST_NBLOCKS=16 + msdos-fs tar: size=7205, st_blocks=16 and ST_NBLOCKS=16 + + Dick Streefland reports the previous numbers as misleading, + because ext2fs use 12 direct blocks, while minix-fs uses only + 6 direct blocks. Dick gets: + + ext2 size=20480 ls listed blocks=21 + minix size=20480 ls listed blocks=21 + msdos size=20480 ls listed blocks=20 + + It seems that indirect blocks *are* included in st_blocks. + The minix filesystem does not account for phantom blocks in + st_blocks, so `du' and `ls -s' give wrong results. So, the + --sparse option would not work on a minix filesystem. */ + + if (ST_NBLOCKS (current_stat) + < (current_stat.st_size / ST_NBLOCKSIZE + + (current_stat.st_size % ST_NBLOCKSIZE != 0))) + { + off_t filesize = current_stat.st_size; + int counter; + + header = start_header (p, ¤t_stat); + header->header.typeflag = GNUTYPE_SPARSE; + header_moved = 1; + + /* Call the routine that figures out the layout of the + sparse file in question. UPPERBOUND is the index of the + last element of the "sparsearray," i.e., the number of + elements it needed to describe the file. */ + + upperbound = deal_with_sparse (p, header); + + /* See if we'll need an extended header later. */ + + if (upperbound > SPARSES_IN_OLDGNU_HEADER - 1) + header->oldgnu_header.isextended = 1; + + /* We store the "real" file size so we can show that in + case someone wants to list the archive, i.e., tar tvf + . It might be kind of disconcerting if the + shrunken file size was the one that showed up. */ + + OFF_TO_CHARS (current_stat.st_size, + header->oldgnu_header.realsize); + + /* This will be the new "size" of the file, i.e., the size + of the file minus the blocks of holes that we're + skipping over. */ + + find_new_file_size (&filesize, upperbound); + current_stat.st_size = filesize; + OFF_TO_CHARS (filesize, header->header.size); + + for (counter = 0; counter < SPARSES_IN_OLDGNU_HEADER; counter++) + { + if (!sparsearray[counter].numbytes) + break; + + OFF_TO_CHARS (sparsearray[counter].offset, + header->oldgnu_header.sp[counter].offset); + SIZE_TO_CHARS (sparsearray[counter].numbytes, + header->oldgnu_header.sp[counter].numbytes); + } + + } + } + else + upperbound = SPARSES_IN_OLDGNU_HEADER - 1; + + sizeleft = current_stat.st_size; + + /* Don't bother opening empty, world readable files. Also do not open + files when archive is meant for /dev/null. */ + + if (dev_null_output + || (sizeleft == 0 + && MODE_R == (MODE_R & current_stat.st_mode))) + f = -1; + else + { + f = open (p, O_RDONLY | O_BINARY); + if (f < 0) + { + if (! top_level && errno == ENOENT) + WARN ((0, 0, _("File %s removed before we read it"), + quote (p))); + else if (ignore_failed_read_option) + open_warn (p); + else + open_error (p); + return; + } + } + + /* If the file is sparse, we've already taken care of this. */ + + if (!header_moved) + header = start_header (p, ¤t_stat); + + /* Mark contiguous files, if we support them. */ + + if (archive_format != V7_FORMAT && S_ISCTG (current_stat.st_mode)) + header->header.typeflag = CONTTYPE; + + isextended = header->oldgnu_header.isextended; + save_typeflag = header->header.typeflag; + finish_header (header); + if (isextended) + { + int counter; + /* static */ int index_offset = SPARSES_IN_OLDGNU_HEADER; + + extend: + exhdr = find_next_block (); + memset (exhdr->buffer, 0, BLOCKSIZE); + for (counter = 0; counter < SPARSES_IN_SPARSE_HEADER; counter++) + { + if (counter + index_offset > upperbound) + break; + + SIZE_TO_CHARS (sparsearray[counter + index_offset].numbytes, + exhdr->sparse_header.sp[counter].numbytes); + OFF_TO_CHARS (sparsearray[counter + index_offset].offset, + exhdr->sparse_header.sp[counter].offset); + } + set_next_block_after (exhdr); + if (index_offset + counter <= upperbound) + { + index_offset += counter; + exhdr->sparse_header.isextended = 1; + goto extend; + } + + } + if (save_typeflag == GNUTYPE_SPARSE) + { + if (f < 0 + || finish_sparse_file (f, &sizeleft, + current_stat.st_size, p)) + goto padit; + } + else + while (sizeleft > 0) + { + if (multi_volume_option) + { + assign_string (&save_name, p); + save_sizeleft = sizeleft; + save_totsize = current_stat.st_size; + } + start = find_next_block (); + + bufsize = available_space_after (start); + + if (sizeleft < bufsize) + { + /* Last read -- zero out area beyond. */ + + bufsize = sizeleft; + count = bufsize % BLOCKSIZE; + if (count) + memset (start->buffer + sizeleft, 0, BLOCKSIZE - count); + } + if (f < 0) + count = bufsize; + else + count = safe_read (f, start->buffer, bufsize); + if (count < 0) + { + char buf[UINTMAX_STRSIZE_BOUND]; + int e = errno; + ERROR ((0, e, + _("Read error at byte %s, reading %lu bytes, in file %s"), + STRINGIFY_BIGINT (current_stat.st_size - sizeleft, + buf), + (unsigned long) bufsize, quote (p))); + goto padit; + } + sizeleft -= count; + + /* This is nonportable (the type of set_next_block_after's arg). */ + + set_next_block_after (start + (count - 1) / BLOCKSIZE); + + if (count == bufsize) + continue; + else + { + char buf[UINTMAX_STRSIZE_BOUND]; + ERROR ((0, 0, + _("File %s shrunk by %s bytes, padding with zeros"), + quote (p), STRINGIFY_BIGINT (sizeleft, buf))); + goto padit; /* short read */ + } + } + + if (multi_volume_option) + assign_string (&save_name, 0); + + if (f >= 0) + { + struct stat final_stat; + if (fstat (f, &final_stat) != 0) + stat_error (p); + else if (final_stat.st_ctime != original_ctime) + ERROR ((0, 0, _("File %s changed as we read it"), quote (p))); + if (close (f) != 0) + close_error (p); + if (atime_preserve_option) + utime (p, &restore_times); + } + if (remove_files_option) + { + if (unlink (p) == -1) + unlink_error (p); + } + return; + + /* File shrunk or gave error, pad out tape to match the size we + specified in the header. */ + + padit: + while (sizeleft > 0) + { + save_sizeleft = sizeleft; + start = find_next_block (); + memset (start->buffer, 0, BLOCKSIZE); + set_next_block_after (start); + sizeleft -= BLOCKSIZE; + } + if (multi_volume_option) + assign_string (&save_name, 0); + if (f >= 0) + { + close (f); + if (atime_preserve_option) + utime (p, &restore_times); + } + return; + } +#ifdef HAVE_READLINK + else if (S_ISLNK (current_stat.st_mode)) + { + int size; + char *buffer = (char *) alloca (PATH_MAX + 1); + + size = readlink (p, buffer, PATH_MAX + 1); + if (size < 0) + { + if (ignore_failed_read_option) + readlink_warn (p); + else + readlink_error (p); + return; + } + buffer[size] = '\0'; + if (size >= NAME_FIELD_SIZE) + write_long (buffer, GNUTYPE_LONGLINK); + assign_string (¤t_link_name, buffer); + + current_stat.st_size = 0; /* force 0 size on symlink */ + header = start_header (p, ¤t_stat); + strncpy (header->header.linkname, buffer, NAME_FIELD_SIZE); + header->header.linkname[NAME_FIELD_SIZE - 1] = '\0'; + header->header.typeflag = SYMTYPE; + finish_header (header); /* nothing more to do to it */ + if (remove_files_option) + { + if (unlink (p) == -1) + unlink_error (p); + } + return; + } +#endif + else if (S_ISCHR (current_stat.st_mode)) + type = CHRTYPE; + else if (S_ISBLK (current_stat.st_mode)) + type = BLKTYPE; + else if (S_ISFIFO (current_stat.st_mode)) + type = FIFOTYPE; + else if (S_ISSOCK (current_stat.st_mode)) + { + WARN ((0, 0, _("%s: socket ignored"), quotearg_colon (p))); + return; + } + else if (S_ISDOOR (current_stat.st_mode)) + { + WARN ((0, 0, _("%s: door ignored"), quotearg_colon (p))); + return; + } + else + goto unknown; + } if (archive_format == V7_FORMAT) goto unknown; current_stat.st_size = 0; /* force 0 size */ header = start_header (p, ¤t_stat); - if (header == NULL) - { - exit_status = TAREXIT_FAILURE; - return; /* eg name too long */ - } - header->header.typeflag = type; -#if defined(S_IFBLK) || defined(S_IFCHR) if (type != FIFOTYPE) { - MAJOR_TO_OCT (major (current_stat.st_rdev), header->header.devmajor); - MINOR_TO_OCT (minor (current_stat.st_rdev), header->header.devminor); + MAJOR_TO_CHARS (major (current_stat.st_rdev), header->header.devmajor); + MINOR_TO_CHARS (minor (current_stat.st_rdev), header->header.devminor); } -#endif finish_header (header); if (remove_files_option) { if (unlink (p) == -1) - ERROR ((0, errno, _("Cannot remove %s"), p)); + unlink_error (p); } return; unknown: - ERROR ((0, 0, _("%s: Unknown file type; file ignored"), p)); + ERROR ((0, 0, _("%s: Unknown file type; file ignored"), + quotearg_colon (p))); }