coreutils-9.5

[Bruce Dubbs]

New minor version.

[Xi Ruoyao]

• Noteworthy changes in release 9.5 (2024-03-28) [stable]

Bug fixes

• chmod -R now avoids a race where an attacker may replace a traversed file with a symlink, causing chmod to operate on an unintended file. [This bug was present in "the beginning".]

• cp, mv, and install no longer issue spurious diagnostics like "failed to preserve ownership" when copying to GNU/Linux CIFS file systems. They do this by working around some Linux CIFS bugs.

• cp --no-preserve=mode will correctly maintain set-group-ID bits for created directories. Previously on systems that didn't support ACLs, cp would have reset the set-group-ID bit on created directories. [bug introduced in coreutils-8.20]

• join and uniq now support multi-byte characters better. For example, 'join -tX' now works even if X is a multi-byte character, and both programs now treat multi-byte characters like U+3000 IDEOGRAPHIC SPACE as blanks if the current locale treats them so.

• numfmt options like --suffix no longer have an arbitrary 127-byte limit. [bug introduced with numfmt in coreutils-8.21]

• mktemp with --suffix now better diagnoses templates with too few X's. Previously it conflated the insignificant --suffix in the error. [bug introduced in coreutils-8.1]

• sort again handles thousands grouping characters in single-byte locales where the grouping character is greater than CHAR_MAX. For e.g. signed character platforms with a 0xA0 (aka &nbsp) grouping character. [bug introduced in coreutils-9.1]

• split --line-bytes with a mixture of very long and short lines no longer overwrites the heap (CVE-2024-0684). [bug introduced in coreutils-9.2]

• tail no longer mishandles input from files in /proc and /sys file systems, on systems with a page size larger than the stdio BUFSIZ. [This bug was present in "the beginning".]

• timeout avoids a narrow race condition, where it might kill arbitrary processes after a failed process fork. [bug introduced with timeout in coreutils-7.0]

• timeout avoids a narrow race condition, where it might fail to kill monitored processes immediately after forking them. [bug introduced with timeout in coreutils-7.0]

• wc no longer fails to count unprintable characters as parts of words. [bug introduced in textutils-2.1]

Changes in behavior

• base32 and base64 no longer require padding when decoding. Previously an error was given for non padded encoded data.

• base32 and base64 have improved detection of corrupted encodings. Previously encodings with non zero padding bits were accepted.

• basenc --base16 -d now supports lower case hexadecimal characters. Previously an error was given for lower case hex digits.

• cp --no-clobber, and mv -n no longer exit with failure status if existing files are encountered in the destination. Instead they revert to the behavior from before v9.2, silently skipping existing files.

• ls --dired now implies long format output without hyperlinks enabled, and will take precedence over previously specified formats or hyperlink mode.

• numfmt will accept lowercase 'k' to indicate Kilo or Kibi units on input, and uses lowercase 'k' when outputting such units in '--to=si' mode.

• pinky no longer tries to canonicalize the user's login location by default, rather requiring the new --lookup option to enable this often slow feature.

• wc no longer ignores encoding errors when counting words. Instead, it treats them as non white space.

New features

• chgrp now accepts the --from=OWNER:GROUP option to restrict changes to files with matching current OWNER and/or GROUP, as already supported by chown(1).

• chmod adds support for -h, -H,-L,-P, and --dereference options, providing more control over symlink handling. This supports more secure handling of CLI arguments, and is more consistent with chown, and chmod on other systems.

• cp now accepts the --keep-directory-symlink option (like tar), to preserve and follow existing symlinks to directories in the destination.

• cp and mv now accept the --update=none-fail option, which is similar to the --no-clobber option, except that existing files are diagnosed, and the command exits with failure status if existing files. The -n,--no-clobber option is best avoided due to platform differences.

• env now accepts the -a,--argv0 option to override the zeroth argument of the command being executed.

• mv now accepts an --exchange option, which causes the source and destination to be exchanged. It should be combined with --no-target-directory (-T) if the destination is a directory. The exchange is atomic if source and destination are on a single file system that supports atomic exchange; --exchange is not yet supported in other situations.

• od now supports printing IEEE half precision floating point with -t fH, or brain 16 bit floating point with -t fB, where supported by the compiler.

• tail now supports following multiple processes, with repeated --pid options.

Improvements

• cp,mv,install,cat,split now read and write a minimum of 256KiB at a time. This was previously 128KiB and increasing to 256KiB was seen to increase throughput by 10-20% when reading cached files on modern systems.

• env,kill,timeout now support unnamed signals. kill(1) for example now supports sending such signals, and env(1) will list them appropriately.

• SELinux operations in file copy operations are now more efficient, avoiding unneeded MCS/MLS label translation.

• sort no longer dynamically links to libcrypto unless -R is used. This decreases startup overhead in the typical case.

• wc is now much faster in single-byte locales and somewhat faster in multi-byte locales.

[Xi Ruoyao]

Replying to Xi Ruoyao:

• join and uniq now support multi-byte characters better. For example, 'join -tX' now works even if X is a multi-byte character, and both programs now treat multi-byte characters like U+3000 IDEOGRAPHIC SPACE as blanks if the current locale treats them so.

So we need to remove the hunks for join and uniq in i18n patch as the i18n support is already upstream. The other hunks seem just applying fine.

[Xi Ruoyao]

split --line-bytes with a mixture of very long and short lines no longer overwrites the heap (CVE-2024-0684). [bug introduced in coreutils-9.2]

Note that we already have a sed to fix this, so don't issue a SA again.

[Xi Ruoyao]

Replying to Xi Ruoyao:

Replying to Xi Ruoyao:

• join and uniq now support multi-byte characters better. For example, 'join -tX' now works even if X is a multi-byte character, and both programs now treat multi-byte characters like U+3000 IDEOGRAPHIC SPACE as blanks if the current locale treats them so.

So we need to remove the hunks for join and uniq in i18n patch as the i18n support is already upstream. The other hunks seem just applying fine.

Also need to add the "mbchar" gnulib module back.

[Xi Ruoyao]

i18n patch rebased and adjusted: ​https://www.linuxfromscratch.org/patches/downloads/coreutils/coreutils-9.5-i18n-1.patch

[Bruce Dubbs]

Fixed at commit 9b01d1b4ba.