The patch leverages the existing -C code. In order to do that, every
entry in the wd[] table gets another companion entry in the table that
represents the --one-top-level directory. There is one additional field
in each entry that allows skipping the companion entries if they are not
desired.
The actual directory is created lazily by chdir_do() if needed, as you
requested, to avoid empty "a/foo" after --one-top-level=foo -C a -C b.
The patch "by the way" fixes also extraction of hardlinks with
--one-top-level which currently is broken in the typical case (the
transform is not applied to the target, so the hardlink is wrong).
The patch does not yet handle the --show-transformed case with
--one-top-level that you discussed in another subthread. As a result, two
tests now fail (onetop02.at and onetop04.at). I suppose that this would
be quite easy to fix.
Another issue that I am aware of is that I am not sure whether to call
repair_delayed_set_stat and/or delay_set_stat on the newly created
directories like extract_dir() does (the whole delay_set code is abit
mysterious to me).
Use of --create together with --one-top-level should probably be
forbidden, as unlink.c uses wd[] in a way that will likely break in the
presence of companion entries (the chdir_do call in
flush_deferred_unlinks).
Show the one-top-level arg as a transformation
Since commit b8d8a61b, tar started honoring size field in headers
of the directory archive members when listing and extracting. That
doesn't seem right: although GNU tar always stores 0 for such members,
there are other tar implementation that don't. As a result, GNU tar
skips the actual directory contents when listing or extracting archives
created by such implementations.
This commit fixes that by assuming that size is 0 for directory archive
members.
* src/list.c (member_is_dir): Restore function.
(skim_member): Don't apply skim_file to directory members.
* tests/skipdir.at: Fix expectations.
Problem reported by Antonio Teixeira.
* src/list.c (read_header): When POSIX says a size field must
be zero or does not represent a data count, treat it as zero.
* tests/extrac32.at: Update to match new behavior.
We now treat hard link sizes as zero even when the size fields are
nonzero, and this means the “injected” file is treated as valid
regardless of whether we list or extract.
* tests/extrac34.at: New test.
* tests/Makefile.am (TESTSUITE_AT), tests/testsuite.at: Add it.
* src/tar.c (one_top_level_option): Remove. All uses removed.
All external uses changed to equivalent of !!one_top_level_dir.
(decode_options): Set one_top_level_dir to be consistent with
the old one_top_level_option.
* src/list.c (enforce_one_top_level): Do not strip prefixes like
"../" (which should not be allowed unless -P) and ".../" (which
are ordinary file names). The "../" not being allowed should be
addressed in a different way.
Problem reported by Guillermo de Angel in:
https://lists.gnu.org/r/bug-tar/2026-03/msg00007.html
* THANKS: Add him, and sort.
* src/extract.c (extract_dir, extract_file):
* src/incremen.c (purge_directory):
Do not call skip_member, as the caller now does that, and does it
more reliably.
* src/extract.c (extract_file):
Mark file as skipped when we’ve read it.
(extract_archive): Always call skip_member after extracting,
as it suppresses the skip as needed.
* src/incremen.c (try_purge_directory): Remove; no longer
needed. Move internals to purge_directory.
* src/list.c (read_header): Do not treat LNKTYPE header as having
size zero, as it can be nonzero (e.g., ‘pax -o linkdata’).
Set info->skipped field according to how the header was read.
(member_is_dir): Remove; no longer needed.
(skim_member): Skip directory data too, unless it’s already been
skipped (i.e., read).
* tests/extrac32.at: New file.
* tests/Makefile.am (TESTSUITE_AT):
* tests/testsuite.at:
Add it.
* tests/skipdir.at (skip directory members):
Fix test to match the correct behavior.
This fixes a bug introduced in commit
b009124ffd
dated 2025-05-12 17:17:21 +0300.
* src/buffer.c (check_compressed_archive):
* src/list.c (read_header, decode_header):
Use memcmp, not strcmp, when looking for magic strings in
headers, since input headers are not guaranteed to be
strings and strcmp has undefined behavior otherwise.
Problem with extract_file reported by Kirill Furman in:
https://lists.gnu.org/r/bug-tar/2025-07/msg00003.html
Since the UBSan thing seems to be a recurring issue,
I fixed other instances of the problem that I found.
Also, I noticed that the same line of code had another failure to
conform to C23’s rules for pointers (an alignment issue not caught
by UBSan), so I fixed that too. None of these issues matter on
practical production hosts.
* src/common.h (charptr): New function.
* src/buffer.c (available_space_after, short_read, flush_archive)
(backspace_output, try_new_volume, simple_flush_read)
(_gnu_flush_read, _gnu_flush_write):
* src/compare.c (read_and_process):
* src/create.c (write_eot, write_gnu_long_link)
(dump_regular_file, dump_dir0):
* src/extract.c (extract_file):
* src/incremen.c (get_gnu_dumpdir):
* src/list.c (read_header):
* src/sparse.c (sparse_dump_region, sparse_extract_region):
* src/system.c (sys_write_archive_buffer)
(sys_child_open_for_compress, sys_child_open_for_uncompress):
* src/update.c (append_file, update_archive):
Use it.
* src/buffer.c (set_next_block_after): Arg is now void *,
not union block *, since it need not be a valid union block * pointer
and this can matter on unusual or debugging implementations.
Turn a loop into an if so that the code is O(1) not O(N).
* src/list.c (skim_member): Recognize directory members using
the same rules as during extraction.
* tests/skipdir.at: New testcase.
* tests/testsuite.at: Add new test.
* tests/Makefile.am: Likewise.
* src/buffer.c (_flush_write, short_read, seek_archive)
(_gnu_flush_write):
* src/create.c (write_gnu_long_link, dump_regular_file)
(dump_dir0):
* src/delete.c (write_recent_bytes, flush_file)
(delete_archive_members):
* src/list.c (read_header):
* src/sparse.c (sparse_dump_region, sparse_extract_region)
(pax_dump_header_1):
* src/tar.c (parse_opt):
* src/update.c (append_file):
Prefer shifting and masking to dividing and remaindering by
BLOCKSIZE. This reclaims some compiler optimizations lost
by our recent preference for signed integers.
* src/tar.h (LG_BLOCKSIZE): New constant, for shifting.
* src/list.c (decode_xform): Last arg is now int, not a void *
pointer to that int. All uses changed.
(enforce_one_top_level): Don’t assume string length fits in int.
(transform_stat_info): Prefer char to int for typeflag.
All uses changed.
(decode_header): Prefer bool for booleans. All uses changed.
(ugswidth): Now idx_t, not int, since in theory it could
exceed INT_MAX. All uses changed.
(simple_print_header, print_for_mkdir): Don’t assume printf length
fits in int, and similarly for length of user or group name.
* src/transform.c (transform_name_fp): Last arg is now int, not void *.
All uses changed.
* src/create.c (max_octal_val, to_octal, tar_copy_str)
(tar_name_copy_str, to_base256, to_chars_subst, to_chars)
(gid_to_chars, major_to_chars, minor_to_chars, mode_to_chars)
(off_to_chars, time_to_chars, uid_to_chars, string_to_chars)
(split_long_name, write_ustar_long_name, simple_finish_header):
* src/list.c (from_header, gid_from_header, major_from_header)
(minor_from_header, mode_from_header, off_from_header)
(time_from_header, uid_from_header):
Prefer int to idx_t where either will do because the buffer sizes
are known to be small, as this can be a performance win on 32-bit
platforms. Also, in a few cases the values were negative, whereas
idx_t is supposed to be nonnegative.
In common.h, replace macros with constants or functions when that
is easy. This makes code a bit more reliable (functions evaluate
their args exactly once) and easier to debug (many debugging
environments cannot access macros).
* src/common.h (CHKBLANKS): Remove. All uses removed.
(NAME_FIELD_SIZE, PREFIX_FIELD_SIZE, UNAME_FIELD_SIZE)
(GNAME_FIELD_SIZE, TAREXIT_SUCCESS, TAREXIT_DIFFERS)
(TAREXIT_FAILURE, LG_8, LG_256, DEFAULT_CHECKPOINT)
(MAX_OLD_FILES, TF_READ, TF_WRITE, TF_DELETED, XFORM_REGFILE)
(XFORM_LINK, XFORM_SYMLINK, XFORM_ALL, WARN_ALONE_ZERO_BLOCK)
(WARN_BAD_DUMPDIR, WARN_CACHEDIR, WARN_CONTIGUOUS_CAST)
(WARN_FILE_CHANGED, WARN_FILE_IGNORED, WARN_FILE_REMOVED)
(WARN_FILE_SHRANK, WARN_FILE_UNCHANGED, WARN_FILENAME_WITH_NULS)
(WARN_IGNORE_ARCHIVE, WARN_IGNORE_NEWER, WARN_NEW_DIRECTORY)
(WARN_RENAME_DIRECTORY, WARN_SYMLINK_CAST, WARN_TIMESTAMP)
(WARN_UNKNOWN_CAST, WARN_UNKNOWN_KEYWORD, WARN_XDEV)
(WARN_DECOMPRESS_PROGRAM, WARN_EXISTING_FILE, WARN_XATTR_WRITE)
(WARN_RECORD_SIZE, WARN_FAILED_READ, WARN_MISSING_ZERO_BLOCKS)
(WARN_VERBOSE_WARNINGS, WARN_ALL, EXCL_DEFAULT, EXCL_RECURSIVE)
(EXCL_NON_RECURSIVE): Now enum constants rather than macros.
(time_option_initialized, isfound, wasfound, warning_enabled):
Now functions rather than macros TIME_OPTION_INITIALIZED, ISFOUND,
WASFOUND, WARNING_ENABLED. All uses changed.
(OLDER_STAT_TIME, OLDER_TAR_STAT_TIME, EXTRACT_OVER_PIPE)
(TAR_ARGS_INITIALIZER): Remove. All uses replaced with their
definiens or equivalent.
Simplify the code by assuming C99 initializers.
* src/list.c (base_64_digits): Remove.
(base64_map): Now a constant. Now has its (old value + 1) % 65,
as that’s the only easy portable way to do it with a static
initializer (even on platforms where CHAR_BIT != 8); all uses changed.
(base64_init): Remove; only use removed.
(from_header): Adjust to new values in base64_map.
* src/list.c (base_64_digits): Remove; no longer needed.
(base64_map): Now const, initialized statically, and with
invalid entries being 0 not 64, and with valid entries
being 1 greater than before.
This is part of the general guideline that signed integer types
are safer.
* src/names.c (stripped_prefix_len): Return ptrdiff_t,
not size_t. All callers changed.
It’s now safe to assume support for C99 formats like %jd, so remove
some of the longwinded formatting code put in only to be portable to
pre-C99 platforms.
* gnulib.modules: Add intprops.
* src/buffer.c (format_total_stats, try_new_volume)
(write_volume_label):
* src/checkpoint.c (format_checkpoint_string):
* src/compare.c (verify_volume):
* src/create.c (to_chars_subst, dump_regular_file):
* src/incremen.c (read_num):
* src/list.c (read_and, from_header, simple_print_header)
(print_for_mkdir):
* src/sparse.c (sparse_dump_region):
* src/system.c (dec_to_env, sys_exec_info_script)
(sys_exec_checkpoint_script):
* src/xheader.c (out_of_range_header):
Prefer C99 formats like %jd and %ju to STRINGIFY_BIGINT.
* src/common.h: Sort includes.
Include intprops.h, verify.h. All other includes of verify.h
removed.
(intmax, uintmax): New functions and macros.
(STRINGIFY_BIGINT): Remove; no longer used.
(TIMESPEC_STRSIZE_BOUND): Make it 1 byte bigger, for negatives.
* src/create.c (MAX_VAL_WITH_DIGITS, to_base256):
Use *_WIDTH macros rather than assuming no padding bits.
Prefer UINTMAX_MAX to (uintmax_t) -1.
* src/list.c (tartime): Use strftime result rather
than running strlen later.
* src/misc.c (timetostr): New function. Prefer it when
printing time_t values.
* src/list.c (read_header): Use ckd_add instead of
doing overflow checking by hand. Although the old code
was correct on all practical hosts, the new code is simpler
and works even on weird hosts where SIZE_MAX <= INT_MAX.
* gnulib.modules: Add assert-h, for static_assert.
* src/common.h, src/list.c, src/misc.c:
Prefer static_assert to #if + #error. This doesn’t fix any bugs; it’s
just that in general it’s better to avoid the preprocessor.
update submodules to latest
* gnulib.modules: Add c-ctype.
* lib/wordsplit.c, src/buffer.c, src/exclist.c, src/incremen.c:
* src/list.c, src/misc.c, src/names.c, src/sparse.c, src/tar.c:
* src/xheader.c:
Include c-ctype.h, and use its API rather than ctype.h’s.
This is more likely to work when oddball locales are used.
* src/transform.c: Include ctype.h, since this module still uses
tolower and toupper (this is probably wrong - should be multi-byte).
* gnulib.modules: Remove alloca.
* src/create.c (dump_file0): Return address of any allocated
storage. Caller changed to free it. Use xmalloc instead
of alloca, to obtain this storage.
* src/list.c (from_header): Use quote_mem instead of quote,
removing the need to use alloca.
This also ports to C23 [[maybe_unused]].
* configure.ac (WARN_CFLAGS): Do not add -Wno-unused-parameter.
Add MAYBE_UNUSED where needed in source code.
Also, put it at the front where C23 requires it.
(In response to savannah bug #63574)
* doc/intern.texi: Document actual tar behaviour in regard to
missing end-of-file marker.
* doc/tar.texi: Rewrite the "warnings" section. Document
--warning=missing-zero-blocks
* src/common.h (WARN_MISSING_ZERO_BLOCKS): New constant.
(WARN_ALL): Include all warning bits.
* src/list.c (read_and): If EOF is reached without seeing end-of-file
blocks and the "missing-zero-blocks" warning is requested, warn about
the fact.
* src/warning.c: New warnings: "missing-zero-blocks", "verbose".
(warning_option): Change definition to reflect changes in common.h
This bug was introduced by the recent lseek-related changes.
* src/delete.c (delete_archive_members):
* src/update.c (update_archive):
Copy the member if acting as a filter, rather than lseeking over
it, which is possible if stdin is a regular file.
* src/list.c (skim_file, skim_member):
* src/sparse.c (sparse_skim_file):
New functions, for copying when a filter.
* src/list.c (skip_file): Remove; replaced with skim_file.
All callers changed.
(skip_member): Reimplement in terms of skim_member.
* src/sparse.c (sparse_skip_file):
Remove; replaced with sparse_skim_file. All callers changed.
* src/update.c (acting_as_filter): New static var.
(update_archive): Set it; this is like delete.c.
* tests/delete01.at (deleting a member after a big one):
* tests/delete02.at (deleting a member from stdin archive):
Also test filter case.
Bug reported in https://savannah.gnu.org/bugs/?59897
* src/list.c (read_header): Don't return directly from the loop.
Instead set the status and break. Return the status. Free
next_long_name and next_long_link before returning.
Pavel CahynaandPaul Eggert
Sergey Poznyakoff