static int
record_hardlink(struct archive_read *a,
struct cpio *cpio, struct archive_entry *entry)
{
struct links_entry *le;
dev_t dev;
int64_t ino;
if (archive_entry_nlink(entry) <= 1)
return (ARCHIVE_OK);
dev = archive_entry_dev(entry);
ino = archive_entry_ino64(entry);
/*
* First look in the list of multiply-linked files. If we've
* already dumped it, convert this entry to a hard link entry.
*/
for (le = cpio->links_head; le; le = le->next) {
if (le->dev == dev && le->ino == ino) {
archive_entry_copy_hardlink(entry, le->name);
if (--le->links <= 0) {
if (le->previous != NULL)
le->previous->next = le->next;
if (le->next != NULL)
le->next->previous = le->previous;
if (cpio->links_head == le)
cpio->links_head = le->next;
free(le->name);
free(le);
}
return (ARCHIVE_OK);
}
}
le = (struct links_entry *)malloc(sizeof(struct links_entry));
if (le == NULL) {
archive_set_error(&a->archive,
ENOMEM, "Out of memory adding file to list");
return (ARCHIVE_FATAL);
}
if (cpio->links_head != NULL)
cpio->links_head->previous = le;
le->next = cpio->links_head;
le->previous = NULL;
cpio->links_head = le;
le->dev = dev;
le->ino = ino;
le->links = archive_entry_nlink(entry) - 1;
le->name = strdup(archive_entry_pathname(entry));
if (le->name == NULL) {
archive_set_error(&a->archive,
ENOMEM, "Out of memory adding file to list");
return (ARCHIVE_FATAL);
}
return (ARCHIVE_OK);
}
static int ar_entry_hardlink(lua_State *L) {
struct archive_entry* self = *ar_entry_check(L, 1);
int is_set;
if ( NULL == self ) return 0;
is_set = ( lua_gettop(L) == 2 );
lua_pushstring(L, archive_entry_hardlink(self));
if ( is_set ) {
archive_entry_copy_hardlink(self, lua_tostring(L, 2));
}
return 1;
}
/*-
* The logic here for -C <dir> attempts to avoid
* chdir() as long as possible. For example:
* "-C /foo -C /bar file" needs chdir("/bar") but not chdir("/foo")
* "-C /foo -C bar file" needs chdir("/foo/bar")
* "-C /foo -C bar /file1" does not need chdir()
* "-C /foo -C bar /file1 file2" needs chdir("/foo/bar") before file2
*
* The only correct way to handle this is to record a "pending" chdir
* request and combine multiple requests intelligently until we
* need to process a non-absolute file. set_chdir() adds the new dir
* to the pending list; do_chdir() actually executes any pending chdir.
*
* This way, programs that build tar command lines don't have to worry
* about -C with non-existent directories; such requests will only
* fail if the directory must be accessed.
*
*/
void
set_chdir(struct bsdtar *bsdtar, const char *newdir)
{
#if defined(_WIN32) && !defined(__CYGWIN__)
if (newdir[0] == '/' || newdir[0] == '\\' ||
/* Detect this type, for example, "C:\" or "C:/" */
(((newdir[0] >= 'a' && newdir[0] <= 'z') ||
(newdir[0] >= 'A' && newdir[0] <= 'Z')) &&
newdir[1] == ':' && (newdir[2] == '/' || newdir[2] == '\\'))) {
#else
if (newdir[0] == '/') {
#endif
/* The -C /foo -C /bar case; dump first one. */
free(bsdtar->pending_chdir);
bsdtar->pending_chdir = NULL;
}
if (bsdtar->pending_chdir == NULL)
/* Easy case: no previously-saved dir. */
bsdtar->pending_chdir = strdup(newdir);
else {
/* The -C /foo -C bar case; concatenate */
char *old_pending = bsdtar->pending_chdir;
size_t old_len = strlen(old_pending);
bsdtar->pending_chdir = malloc(old_len + strlen(newdir) + 2);
if (old_pending[old_len - 1] == '/')
old_pending[old_len - 1] = '\0';
if (bsdtar->pending_chdir != NULL)
sprintf(bsdtar->pending_chdir, "%s/%s",
old_pending, newdir);
free(old_pending);
}
if (bsdtar->pending_chdir == NULL)
lafe_errc(1, errno, "No memory");
}
void
do_chdir(struct bsdtar *bsdtar)
{
if (bsdtar->pending_chdir == NULL)
return;
if (chdir(bsdtar->pending_chdir) != 0) {
lafe_errc(1, 0, "could not chdir to '%s'\n",
bsdtar->pending_chdir);
}
free(bsdtar->pending_chdir);
bsdtar->pending_chdir = NULL;
}
static const char *
strip_components(const char *p, int elements)
{
/* Skip as many elements as necessary. */
while (elements > 0) {
switch (*p++) {
case '/':
#if defined(_WIN32) && !defined(__CYGWIN__)
case '\\': /* Support \ path sep on Windows ONLY. */
#endif
elements--;
break;
case '\0':
/* Path is too short, skip it. */
return (NULL);
}
}
/* Skip any / characters. This handles short paths that have
* additional / termination. This also handles the case where
* the logic above stops in the middle of a duplicate //
* sequence (which would otherwise get converted to an
* absolute path). */
for (;;) {
switch (*p) {
case '/':
#if defined(_WIN32) && !defined(__CYGWIN__)
case '\\': /* Support \ path sep on Windows ONLY. */
#endif
++p;
break;
case '\0':
return (NULL);
default:
return (p);
}
}
}
static void
warn_strip_leading_char(struct bsdtar *bsdtar, const char *c)
{
if (!bsdtar->warned_lead_slash) {
lafe_warnc(0,
"Removing leading '%c' from member names",
c[0]);
bsdtar->warned_lead_slash = 1;
}
}
static void
warn_strip_drive_letter(struct bsdtar *bsdtar)
{
if (!bsdtar->warned_lead_slash) {
lafe_warnc(0,
"Removing leading drive letter from "
"member names");
bsdtar->warned_lead_slash = 1;
}
}
/*
* Convert absolute path to non-absolute path by skipping leading
* absolute path prefixes.
*/
static const char*
strip_absolute_path(struct bsdtar *bsdtar, const char *p)
{
const char *rp;
/* Remove leading "//./" or "//?/" or "//?/UNC/"
* (absolute path prefixes used by Windows API) */
if ((p[0] == '/' || p[0] == '\\') &&
(p[1] == '/' || p[1] == '\\') &&
(p[2] == '.' || p[2] == '?') &&
(p[3] == '/' || p[3] == '\\'))
{
if (p[2] == '?' &&
(p[4] == 'U' || p[4] == 'u') &&
(p[5] == 'N' || p[5] == 'n') &&
(p[6] == 'C' || p[6] == 'c') &&
(p[7] == '/' || p[7] == '\\'))
p += 8;
else
p += 4;
warn_strip_drive_letter(bsdtar);
}
/* Remove multiple leading slashes and Windows drive letters. */
do {
rp = p;
if (((p[0] >= 'a' && p[0] <= 'z') ||
(p[0] >= 'A' && p[0] <= 'Z')) &&
p[1] == ':') {
p += 2;
warn_strip_drive_letter(bsdtar);
}
/* Remove leading "/../", "/./", "//", etc. */
while (p[0] == '/' || p[0] == '\\') {
if (p[1] == '.' &&
p[2] == '.' &&
(p[3] == '/' || p[3] == '\\')) {
p += 3; /* Remove "/..", leave "/" for next pass. */
} else if (p[1] == '.' &&
(p[2] == '/' || p[2] == '\\')) {
p += 2; /* Remove "/.", leave "/" for next pass. */
} else
p += 1; /* Remove "/". */
warn_strip_leading_char(bsdtar, rp);
}
} while (rp != p);
return (p);
}
/*
* Handle --strip-components and any future path-rewriting options.
* Returns non-zero if the pathname should not be extracted.
*
* Note: The rewrites are applied uniformly to pathnames and hardlink
* names but not to symlink bodies. This is deliberate: Symlink
* bodies are not necessarily filenames. Even when they are, they
* need to be interpreted relative to the directory containing them,
* so simple rewrites like this are rarely appropriate.
*
* TODO: Support pax-style regex path rewrites.
*/
int
edit_pathname(struct bsdtar *bsdtar, struct archive_entry *entry)
{
const char *name = archive_entry_pathname(entry);
const char *original_name = name;
const char *hardlinkname = archive_entry_hardlink(entry);
const char *original_hardlinkname = hardlinkname;
#if defined(HAVE_REGEX_H) || defined(HAVE_PCREPOSIX_H)
char *subst_name;
int r;
/* Apply user-specified substitution to pathname. */
r = apply_substitution(bsdtar, name, &subst_name, 0, 0);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_pathname(entry, subst_name);
if (*subst_name == '\0') {
free(subst_name);
return -1;
} else
free(subst_name);
name = archive_entry_pathname(entry);
original_name = name;
}
/* Apply user-specified substitution to hardlink target. */
if (hardlinkname != NULL) {
r = apply_substitution(bsdtar, hardlinkname, &subst_name, 0, 1);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_hardlink(entry, subst_name);
free(subst_name);
}
hardlinkname = archive_entry_hardlink(entry);
original_hardlinkname = hardlinkname;
}
/* Apply user-specified substitution to symlink body. */
if (archive_entry_symlink(entry) != NULL) {
r = apply_substitution(bsdtar, archive_entry_symlink(entry), &subst_name, 1, 0);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_symlink(entry, subst_name);
free(subst_name);
}
}
#endif
/* Strip leading dir names as per --strip-components option. */
if (bsdtar->strip_components > 0) {
name = strip_components(name, bsdtar->strip_components);
if (name == NULL)
return (1);
if (hardlinkname != NULL) {
hardlinkname = strip_components(hardlinkname,
bsdtar->strip_components);
if (hardlinkname == NULL)
return (1);
}
}
if (!bsdtar->option_absolute_paths) {
/* By default, don't write or restore absolute pathnames. */
name = strip_absolute_path(bsdtar, name);
if (*name == '\0')
name = ".";
if (hardlinkname != NULL) {
hardlinkname = strip_absolute_path(bsdtar, hardlinkname);
if (*hardlinkname == '\0')
return (1);
}
} else {
/* Strip redundant leading '/' characters. */
while (name[0] == '/' && name[1] == '/')
name++;
}
/* Replace name in archive_entry. */
if (name != original_name) {
archive_entry_copy_pathname(entry, name);
}
if (hardlinkname != original_hardlinkname) {
archive_entry_copy_hardlink(entry, hardlinkname);
}
return (0);
}
/*
* It would be nice to just use printf() for formatting large numbers,
* but the compatibility problems are quite a headache. Hence the
* following simple utility function.
*/
const char *
tar_i64toa(int64_t n0)
{
static char buff[24];
uint64_t n = n0 < 0 ? -n0 : n0;
char *p = buff + sizeof(buff);
*--p = '\0';
do {
*--p = '0' + (int)(n % 10);
} while (n /= 10);
if (n0 < 0)
*--p = '-';
return p;
}
/*
* Like strcmp(), but try to be a little more aware of the fact that
* we're comparing two paths. Right now, it just handles leading
* "./" and trailing '/' specially, so that "a/b/" == "./a/b"
*
* TODO: Make this better, so that "./a//b/./c/" == "a/b/c"
* TODO: After this works, push it down into libarchive.
* TODO: Publish the path normalization routines in libarchive so
* that bsdtar can normalize paths and use fast strcmp() instead
* of this.
*
* Note: This is currently only used within write.c, so should
* not handle \ path separators.
*/
int
pathcmp(const char *a, const char *b)
{
/* Skip leading './' */
if (a[0] == '.' && a[1] == '/' && a[2] != '\0')
a += 2;
if (b[0] == '.' && b[1] == '/' && b[2] != '\0')
b += 2;
/* Find the first difference, or return (0) if none. */
while (*a == *b) {
if (*a == '\0')
return (0);
a++;
b++;
}
/*
* If one ends in '/' and the other one doesn't,
* they're the same.
*/
if (a[0] == '/' && a[1] == '\0' && b[0] == '\0')
return (0);
if (a[0] == '\0' && b[0] == '/' && b[1] == '\0')
return (0);
/* They're really different, return the correct sign. */
return (*(const unsigned char *)a - *(const unsigned char *)b);
}
#define PPBUFF_SIZE 1024
const char *
passphrase_callback(struct archive *a, void *_client_data)
{
struct bsdtar *bsdtar = (struct bsdtar *)_client_data;
(void)a; /* UNUSED */
if (bsdtar->ppbuff == NULL) {
bsdtar->ppbuff = malloc(PPBUFF_SIZE);
if (bsdtar->ppbuff == NULL)
lafe_errc(1, errno, "Out of memory");
}
return lafe_readpassphrase("Enter passphrase:",
bsdtar->ppbuff, PPBUFF_SIZE);
}
void
passphrase_free(char *ppbuff)
{
if (ppbuff != NULL) {
memset(ppbuff, 0, PPBUFF_SIZE);
free(ppbuff);
}
}
/*
* Display information about the current file.
*
* The format here roughly duplicates the output of 'ls -l'.
* This is based on SUSv2, where 'tar tv' is documented as
* listing additional information in an "unspecified format,"
* and 'pax -l' is documented as using the same format as 'ls -l'.
*/
void
list_item_verbose(struct bsdtar *bsdtar, FILE *out, struct archive_entry *entry)
{
char tmp[100];
size_t w;
const char *p;
const char *fmt;
time_t tim;
static time_t now;
/*
* We avoid collecting the entire list in memory at once by
* listing things as we see them. However, that also means we can't
* just pre-compute the field widths. Instead, we start with guesses
* and just widen them as necessary. These numbers are completely
* arbitrary.
*/
if (!bsdtar->u_width) {
bsdtar->u_width = 6;
bsdtar->gs_width = 13;
}
if (!now)
time(&now);
fprintf(out, "%s %d ",
archive_entry_strmode(entry),
archive_entry_nlink(entry));
/* Use uname if it's present, else uid. */
p = archive_entry_uname(entry);
if ((p == NULL) || (*p == '\0')) {
sprintf(tmp, "%lu ",
(unsigned long)archive_entry_uid(entry));
p = tmp;
}
w = strlen(p);
if (w > bsdtar->u_width)
bsdtar->u_width = w;
fprintf(out, "%-*s ", (int)bsdtar->u_width, p);
/* Use gname if it's present, else gid. */
p = archive_entry_gname(entry);
if (p != NULL && p[0] != '\0') {
fprintf(out, "%s", p);
w = strlen(p);
} else {
sprintf(tmp, "%lu",
(unsigned long)archive_entry_gid(entry));
w = strlen(tmp);
fprintf(out, "%s", tmp);
}
/*
* Print device number or file size, right-aligned so as to make
* total width of group and devnum/filesize fields be gs_width.
* If gs_width is too small, grow it.
*/
if (archive_entry_filetype(entry) == AE_IFCHR
|| archive_entry_filetype(entry) == AE_IFBLK) {
sprintf(tmp, "%lu,%lu",
(unsigned long)archive_entry_rdevmajor(entry),
(unsigned long)archive_entry_rdevminor(entry));
} else {
strcpy(tmp, tar_i64toa(archive_entry_size(entry)));
}
if (w + strlen(tmp) >= bsdtar->gs_width)
bsdtar->gs_width = w+strlen(tmp)+1;
fprintf(out, "%*s", (int)(bsdtar->gs_width - w), tmp);
/* Format the time using 'ls -l' conventions. */
tim = archive_entry_mtime(entry);
#define HALF_YEAR (time_t)365 * 86400 / 2
#if defined(_WIN32) && !defined(__CYGWIN__)
#define DAY_FMT "%d" /* Windows' strftime function does not support %e format. */
#else
#define DAY_FMT "%e" /* Day number without leading zeros */
#endif
if (tim < now - HALF_YEAR || tim > now + HALF_YEAR)
fmt = bsdtar->day_first ? DAY_FMT " %b %Y" : "%b " DAY_FMT " %Y";
else
fmt = bsdtar->day_first ? DAY_FMT " %b %H:%M" : "%b " DAY_FMT " %H:%M";
strftime(tmp, sizeof(tmp), fmt, localtime(&tim));
fprintf(out, " %s ", tmp);
safe_fprintf(out, "%s", archive_entry_pathname(entry));
/* Extra information for links. */
if (archive_entry_hardlink(entry)) /* Hard link */
safe_fprintf(out, " link to %s",
archive_entry_hardlink(entry));
else if (archive_entry_symlink(entry)) /* Symbolic link */
safe_fprintf(out, " -> %s", archive_entry_symlink(entry));
}
void
archive_entry_linkify(struct archive_entry_linkresolver *res,
struct archive_entry **e, struct archive_entry **f)
{
struct links_entry *le;
struct archive_entry *t;
*f = NULL; /* Default: Don't return a second entry. */
if (*e == NULL) {
le = next_entry(res, NEXT_ENTRY_DEFERRED);
if (le != NULL) {
*e = le->entry;
le->entry = NULL;
}
return;
}
/* If it has only one link, then we're done. */
if (archive_entry_nlink(*e) == 1)
return;
/* Directories, devices never have hardlinks. */
if (archive_entry_filetype(*e) == AE_IFDIR
|| archive_entry_filetype(*e) == AE_IFBLK
|| archive_entry_filetype(*e) == AE_IFCHR)
return;
switch (res->strategy) {
case ARCHIVE_ENTRY_LINKIFY_LIKE_TAR:
le = find_entry(res, *e);
if (le != NULL) {
archive_entry_unset_size(*e);
archive_entry_copy_hardlink(*e,
archive_entry_pathname(le->canonical));
} else
insert_entry(res, *e);
return;
case ARCHIVE_ENTRY_LINKIFY_LIKE_MTREE:
le = find_entry(res, *e);
if (le != NULL) {
archive_entry_copy_hardlink(*e,
archive_entry_pathname(le->canonical));
} else
insert_entry(res, *e);
return;
case ARCHIVE_ENTRY_LINKIFY_LIKE_OLD_CPIO:
/* This one is trivial. */
return;
case ARCHIVE_ENTRY_LINKIFY_LIKE_NEW_CPIO:
le = find_entry(res, *e);
if (le != NULL) {
/*
* Put the new entry in le, return the
* old entry from le.
*/
t = *e;
*e = le->entry;
le->entry = t;
/* Make the old entry into a hardlink. */
archive_entry_unset_size(*e);
archive_entry_copy_hardlink(*e,
archive_entry_pathname(le->canonical));
/* If we ran out of links, return the
* final entry as well. */
if (le->links == 0) {
*f = le->entry;
le->entry = NULL;
}
} else {
/*
* If we haven't seen it, tuck it away
* for future use.
*/
le = insert_entry(res, *e);
if (le == NULL)
/* XXX We should return an error code XXX */
return;
le->entry = *e;
*e = NULL;
}
return;
default:
break;
}
return;
}
/*
* Handle --strip-components and any future path-rewriting options.
* Returns non-zero if the pathname should not be extracted.
*
* TODO: Support pax-style regex path rewrites.
*/
int
edit_pathname(struct bsdtar *bsdtar, struct archive_entry *entry)
{
const char *name = archive_entry_pathname(entry);
#if HAVE_REGEX_H
char *subst_name;
int r;
#endif
#if HAVE_REGEX_H
r = apply_substitution(bsdtar, name, &subst_name, 0);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_pathname(entry, subst_name);
if (*subst_name == '\0') {
free(subst_name);
return -1;
} else
free(subst_name);
name = archive_entry_pathname(entry);
}
if (archive_entry_hardlink(entry)) {
r = apply_substitution(bsdtar, archive_entry_hardlink(entry), &subst_name, 1);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_hardlink(entry, subst_name);
free(subst_name);
}
}
if (archive_entry_symlink(entry) != NULL) {
r = apply_substitution(bsdtar, archive_entry_symlink(entry), &subst_name, 1);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_symlink(entry, subst_name);
free(subst_name);
}
}
#endif
/* Strip leading dir names as per --strip-components option. */
if (bsdtar->strip_components > 0) {
const char *linkname = archive_entry_hardlink(entry);
name = strip_components(name, bsdtar->strip_components);
if (name == NULL)
return (1);
if (linkname != NULL) {
linkname = strip_components(linkname,
bsdtar->strip_components);
if (linkname == NULL)
return (1);
archive_entry_copy_hardlink(entry, linkname);
}
}
/* By default, don't write or restore absolute pathnames. */
if (!bsdtar->option_absolute_paths) {
const char *rp, *p = name;
int slashonly = 1;
/* Remove leading "//./" or "//?/" or "//?/UNC/"
* (absolute path prefixes used by Windows API) */
if ((p[0] == '/' || p[0] == '\\') &&
(p[1] == '/' || p[1] == '\\') &&
(p[2] == '.' || p[2] == '?') &&
(p[3] == '/' || p[3] == '\\'))
{
if (p[2] == '?' &&
(p[4] == 'U' || p[4] == 'u') &&
(p[5] == 'N' || p[5] == 'n') &&
(p[6] == 'C' || p[6] == 'c') &&
(p[7] == '/' || p[7] == '\\'))
p += 8;
else
p += 4;
slashonly = 0;
}
do {
rp = p;
/* Remove leading drive letter from archives created
* on Windows. */
if (((p[0] >= 'a' && p[0] <= 'z') ||
(p[0] >= 'A' && p[0] <= 'Z')) &&
p[1] == ':') {
p += 2;
slashonly = 0;
}
/* Remove leading "/../", "//", etc. */
while (p[0] == '/' || p[0] == '\\') {
if (p[1] == '.' && p[2] == '.' &&
(p[3] == '/' || p[3] == '\\')) {
p += 3; /* Remove "/..", leave "/"
* for next pass. */
slashonly = 0;
} else
p += 1; /* Remove "/". */
}
} while (rp != p);
if (p != name && !bsdtar->warned_lead_slash) {
/* Generate a warning the first time this happens. */
if (slashonly)
lafe_warnc(0,
"Removing leading '%c' from member names",
name[0]);
else
lafe_warnc(0,
"Removing leading drive letter from "
"member names");
bsdtar->warned_lead_slash = 1;
}
/* Special case: Stripping everything yields ".". */
if (*p == '\0')
name = ".";
else
name = p;
} else {
/* Strip redundant leading '/' characters. */
while (name[0] == '/' && name[1] == '/')
name++;
}
/* Safely replace name in archive_entry. */
if (name != archive_entry_pathname(entry)) {
char *q = strdup(name);
archive_entry_copy_pathname(entry, q);
free(q);
}
return (0);
}
/*-
* The logic here for -C <dir> attempts to avoid
* chdir() as long as possible. For example:
* "-C /foo -C /bar file" needs chdir("/bar") but not chdir("/foo")
* "-C /foo -C bar file" needs chdir("/foo/bar")
* "-C /foo -C bar /file1" does not need chdir()
* "-C /foo -C bar /file1 file2" needs chdir("/foo/bar") before file2
*
* The only correct way to handle this is to record a "pending" chdir
* request and combine multiple requests intelligently until we
* need to process a non-absolute file. set_chdir() adds the new dir
* to the pending list; do_chdir() actually executes any pending chdir.
*
* This way, programs that build tar command lines don't have to worry
* about -C with non-existent directories; such requests will only
* fail if the directory must be accessed.
*
*/
void
set_chdir(struct bsdtar *bsdtar, const char *newdir)
{
#if defined(_WIN32) && !defined(__CYGWIN__)
if (newdir[0] == '/' || newdir[0] == '\\' ||
/* Detect this type, for example, "C:\" or "C:/" */
(((newdir[0] >= 'a' && newdir[0] <= 'z') ||
(newdir[0] >= 'A' && newdir[0] <= 'Z')) &&
newdir[1] == ':' && (newdir[2] == '/' || newdir[2] == '\\'))) {
#else
if (newdir[0] == '/') {
#endif
/* The -C /foo -C /bar case; dump first one. */
free(bsdtar->pending_chdir);
bsdtar->pending_chdir = NULL;
}
if (bsdtar->pending_chdir == NULL)
/* Easy case: no previously-saved dir. */
bsdtar->pending_chdir = strdup(newdir);
else {
/* The -C /foo -C bar case; concatenate */
char *old_pending = bsdtar->pending_chdir;
size_t old_len = strlen(old_pending);
bsdtar->pending_chdir = malloc(old_len + strlen(newdir) + 2);
if (old_pending[old_len - 1] == '/')
old_pending[old_len - 1] = '\0';
if (bsdtar->pending_chdir != NULL)
sprintf(bsdtar->pending_chdir, "%s/%s",
old_pending, newdir);
free(old_pending);
}
if (bsdtar->pending_chdir == NULL)
lafe_errc(1, errno, "No memory");
}
void
do_chdir(struct bsdtar *bsdtar)
{
if (bsdtar->pending_chdir == NULL)
return;
if (chdir(bsdtar->pending_chdir) != 0) {
lafe_errc(1, 0, "could not chdir to '%s'\n",
bsdtar->pending_chdir);
}
free(bsdtar->pending_chdir);
bsdtar->pending_chdir = NULL;
}
static const char *
strip_components(const char *p, int elements)
{
/* Skip as many elements as necessary. */
while (elements > 0) {
switch (*p++) {
case '/':
#if defined(_WIN32) && !defined(__CYGWIN__)
case '\\': /* Support \ path sep on Windows ONLY. */
#endif
elements--;
break;
case '\0':
/* Path is too short, skip it. */
return (NULL);
}
}
/* Skip any / characters. This handles short paths that have
* additional / termination. This also handles the case where
* the logic above stops in the middle of a duplicate //
* sequence (which would otherwise get converted to an
* absolute path). */
for (;;) {
switch (*p) {
case '/':
#if defined(_WIN32) && !defined(__CYGWIN__)
case '\\': /* Support \ path sep on Windows ONLY. */
#endif
++p;
break;
case '\0':
return (NULL);
default:
return (p);
}
}
}
/*
* Handle --strip-components and any future path-rewriting options.
* Returns non-zero if the pathname should not be extracted.
*
* TODO: Support pax-style regex path rewrites.
*/
int
edit_pathname(struct bsdtar *bsdtar, struct archive_entry *entry)
{
const char *name = archive_entry_pathname(entry);
#if defined(HAVE_REGEX_H) || defined(HAVE_PCREPOSIX_H)
char *subst_name;
int r;
r = apply_substitution(bsdtar, name, &subst_name, 0, 0);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_pathname(entry, subst_name);
if (*subst_name == '\0') {
free(subst_name);
return -1;
} else
free(subst_name);
name = archive_entry_pathname(entry);
}
if (archive_entry_hardlink(entry)) {
r = apply_substitution(bsdtar, archive_entry_hardlink(entry), &subst_name, 0, 1);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_hardlink(entry, subst_name);
free(subst_name);
}
}
if (archive_entry_symlink(entry) != NULL) {
r = apply_substitution(bsdtar, archive_entry_symlink(entry), &subst_name, 1, 0);
if (r == -1) {
lafe_warnc(0, "Invalid substitution, skipping entry");
return 1;
}
if (r == 1) {
archive_entry_copy_symlink(entry, subst_name);
free(subst_name);
}
}
#endif
/* Strip leading dir names as per --strip-components option. */
if (bsdtar->strip_components > 0) {
const char *linkname = archive_entry_hardlink(entry);
name = strip_components(name, bsdtar->strip_components);
if (name == NULL)
return (1);
if (linkname != NULL) {
linkname = strip_components(linkname,
bsdtar->strip_components);
if (linkname == NULL)
return (1);
archive_entry_copy_hardlink(entry, linkname);
}
}
/* By default, don't write or restore absolute pathnames. */
if (!bsdtar->option_absolute_paths) {
const char *rp, *p = name;
int slashonly = 1;
/* Remove leading "//./" or "//?/" or "//?/UNC/"
* (absolute path prefixes used by Windows API) */
if ((p[0] == '/' || p[0] == '\\') &&
(p[1] == '/' || p[1] == '\\') &&
(p[2] == '.' || p[2] == '?') &&
(p[3] == '/' || p[3] == '\\'))
{
if (p[2] == '?' &&
(p[4] == 'U' || p[4] == 'u') &&
(p[5] == 'N' || p[5] == 'n') &&
(p[6] == 'C' || p[6] == 'c') &&
(p[7] == '/' || p[7] == '\\'))
p += 8;
else
p += 4;
slashonly = 0;
}
do {
rp = p;
/* Remove leading drive letter from archives created
* on Windows. */
if (((p[0] >= 'a' && p[0] <= 'z') ||
(p[0] >= 'A' && p[0] <= 'Z')) &&
p[1] == ':') {
p += 2;
slashonly = 0;
}
/* Remove leading "/../", "//", etc. */
while (p[0] == '/' || p[0] == '\\') {
if (p[1] == '.' && p[2] == '.' &&
(p[3] == '/' || p[3] == '\\')) {
p += 3; /* Remove "/..", leave "/"
* for next pass. */
slashonly = 0;
} else
p += 1; /* Remove "/". */
}
} while (rp != p);
if (p != name && !bsdtar->warned_lead_slash) {
/* Generate a warning the first time this happens. */
if (slashonly)
lafe_warnc(0,
"Removing leading '%c' from member names",
name[0]);
else
lafe_warnc(0,
"Removing leading drive letter from "
"member names");
bsdtar->warned_lead_slash = 1;
}
/* Special case: Stripping everything yields ".". */
if (*p == '\0')
name = ".";
else
name = p;
} else {
/* Strip redundant leading '/' characters. */
while (name[0] == '/' && name[1] == '/')
name++;
}
/* Safely replace name in archive_entry. */
if (name != archive_entry_pathname(entry)) {
char *q = strdup(name);
archive_entry_copy_pathname(entry, q);
free(q);
}
return (0);
}
/*
* It would be nice to just use printf() for formatting large numbers,
* but the compatibility problems are quite a headache. Hence the
* following simple utility function.
*/
const char *
tar_i64toa(int64_t n0)
{
static char buff[24];
uint64_t n = n0 < 0 ? -n0 : n0;
char *p = buff + sizeof(buff);
*--p = '\0';
do {
*--p = '0' + (int)(n % 10);
} while (n /= 10);
if (n0 < 0)
*--p = '-';
return p;
}
/*
* Like strcmp(), but try to be a little more aware of the fact that
* we're comparing two paths. Right now, it just handles leading
* "./" and trailing '/' specially, so that "a/b/" == "./a/b"
*
* TODO: Make this better, so that "./a//b/./c/" == "a/b/c"
* TODO: After this works, push it down into libarchive.
* TODO: Publish the path normalization routines in libarchive so
* that bsdtar can normalize paths and use fast strcmp() instead
* of this.
*
* Note: This is currently only used within write.c, so should
* not handle \ path separators.
*/
int
pathcmp(const char *a, const char *b)
{
/* Skip leading './' */
if (a[0] == '.' && a[1] == '/' && a[2] != '\0')
a += 2;
if (b[0] == '.' && b[1] == '/' && b[2] != '\0')
b += 2;
/* Find the first difference, or return (0) if none. */
while (*a == *b) {
if (*a == '\0')
return (0);
a++;
b++;
}
/*
* If one ends in '/' and the other one doesn't,
* they're the same.
*/
if (a[0] == '/' && a[1] == '\0' && b[0] == '\0')
return (0);
if (a[0] == '\0' && b[0] == '/' && b[1] == '\0')
return (0);
/* They're really different, return the correct sign. */
return (*(const unsigned char *)a - *(const unsigned char *)b);
}
void Entry::set_hardlink(const char *hardlink)
{
archive_entry_copy_hardlink(_entry, hardlink);
}
/*
* Set the locale and write a pathname containing invalid characters.
* This should work; the underlying implementation should automatically
* fall back to storing the pathname in binary.
*/
static void
test_pax_filename_encoding_2(void)
{
char filename[] = "abc\314\214mno\374xyz";
struct archive *a;
struct archive_entry *entry;
char buff[65536];
char longname[] = "abc\314\214mno\374xyz"
"/abc\314\214mno\374xyz/abcdefghijklmnopqrstuvwxyz"
"/abc\314\214mno\374xyz/abcdefghijklmnopqrstuvwxyz"
"/abc\314\214mno\374xyz/abcdefghijklmnopqrstuvwxyz"
"/abc\314\214mno\374xyz/abcdefghijklmnopqrstuvwxyz"
"/abc\314\214mno\374xyz/abcdefghijklmnopqrstuvwxyz"
"/abc\314\214mno\374xyz/abcdefghijklmnopqrstuvwxyz"
;
size_t used;
/*
* We need a starting locale which has invalid sequences.
* en_US.UTF-8 seems to be commonly supported.
*/
/* If it doesn't exist, just warn and return. */
if (NULL == setlocale(LC_ALL, "en_US.UTF-8")) {
skipping("invalid encoding tests require a suitable locale;"
" en_US.UTF-8 not available on this system");
return;
}
assert((a = archive_write_new()) != NULL);
assertEqualIntA(a, 0, archive_write_set_format_pax(a));
assertEqualIntA(a, 0, archive_write_add_filter_none(a));
assertEqualIntA(a, 0, archive_write_set_bytes_per_block(a, 0));
assertEqualInt(0,
archive_write_open_memory(a, buff, sizeof(buff), &used));
assert((entry = archive_entry_new()) != NULL);
/* Set pathname, gname, uname, hardlink to nonconvertible values. */
archive_entry_copy_pathname(entry, filename);
archive_entry_copy_gname(entry, filename);
archive_entry_copy_uname(entry, filename);
archive_entry_copy_hardlink(entry, filename);
archive_entry_set_filetype(entry, AE_IFREG);
failure("This should generate a warning for nonconvertible names.");
assertEqualInt(ARCHIVE_WARN, archive_write_header(a, entry));
archive_entry_free(entry);
assert((entry = archive_entry_new()) != NULL);
/* Set path, gname, uname, and symlink to nonconvertible values. */
archive_entry_copy_pathname(entry, filename);
archive_entry_copy_gname(entry, filename);
archive_entry_copy_uname(entry, filename);
archive_entry_copy_symlink(entry, filename);
archive_entry_set_filetype(entry, AE_IFLNK);
failure("This should generate a warning for nonconvertible names.");
assertEqualInt(ARCHIVE_WARN, archive_write_header(a, entry));
archive_entry_free(entry);
assert((entry = archive_entry_new()) != NULL);
/* Set pathname to a very long nonconvertible value. */
archive_entry_copy_pathname(entry, longname);
archive_entry_set_filetype(entry, AE_IFREG);
failure("This should generate a warning for nonconvertible names.");
assertEqualInt(ARCHIVE_WARN, archive_write_header(a, entry));
archive_entry_free(entry);
assertEqualIntA(a, ARCHIVE_OK, archive_write_close(a));
assertEqualInt(ARCHIVE_OK, archive_write_free(a));
/*
* Now read the entries back.
*/
assert((a = archive_read_new()) != NULL);
assertEqualInt(0, archive_read_support_format_tar(a));
assertEqualInt(0, archive_read_open_memory(a, buff, used));
assertEqualInt(0, archive_read_next_header(a, &entry));
assertEqualString(filename, archive_entry_pathname(entry));
assertEqualString(filename, archive_entry_gname(entry));
assertEqualString(filename, archive_entry_uname(entry));
assertEqualString(filename, archive_entry_hardlink(entry));
assertEqualInt(0, archive_read_next_header(a, &entry));
assertEqualString(filename, archive_entry_pathname(entry));
assertEqualString(filename, archive_entry_gname(entry));
assertEqualString(filename, archive_entry_uname(entry));
assertEqualString(filename, archive_entry_symlink(entry));
assertEqualInt(0, archive_read_next_header(a, &entry));
assertEqualString(longname, archive_entry_pathname(entry));
assertEqualIntA(a, ARCHIVE_OK, archive_read_close(a));
assertEqualInt(ARCHIVE_OK, archive_read_free(a));
}
