ARCHIVE_ENTRY_PERMS(3)

πŸ“¦
version 3.8.9 Β· 3 versions
, in arch rolling, centos-stream 10, fedora 43, fedora 44, freebsd 14.5, freebsd 15.1, nixos 26.05, nixpkgs unstable, opensuse-leap 16.0, ubuntu questing, ubuntu resolute, ubuntu stonking

πŸ’– ARCHIVE_ENTRY_PERMS β€” the ownership + perms glow-up page πŸ’–

okay bestie, this is archive_entry_perms(3) from the Library Functions Manual, and it's literally the whole group chat of functions that mess with who owns a file and what perms it has in an archive entry description. no cap.

✨ LIBRARY ✨

Streaming Archive Library (libarchive, -larchive) β€” she's streaming, she's archiving, she's iconic πŸŽ€

🌸 SYNOPSIS 🌸

#include <archive_entry.h>

gid_t
archive_entry_gid(struct archive_entry *a);

void
archive_entry_set_gid(struct archive_entry *a, gid_t gid);

uid_t
archive_entry_uid(struct archive_entry *a);

void
archive_entry_set_uid(struct archive_entry *a, uid_t uid);

mode_t
archive_entry_perm(struct archive_entry *a);

void
archive_entry_set_perm(struct archive_entry *a, mode_t mode);

const char *
archive_entry_strmode(struct archive_entry *a);

const char *
archive_entry_gname(struct archive_entry *a);

const wchar_t *
archive_entry_gname_w(struct archive_entry *a);

void
archive_entry_set_gname(struct archive_entry *a, const char *a);

void
archive_entry_copy_gname(struct archive_entry *a, const char *name);

void
archive_entry_copy_gname_w(struct archive_entry *a,
    const wchar_t *name);

int
archive_entry_update_gname_utf8(struct archive_entry *a,
    const char *name);

const char *
archive_entry_uname(struct archive_entry *a);

const wchar_t *
archive_entry_uname_w(struct archive_entry *a);

void
archive_entry_set_uname(struct archive_entry *a, const char *name);

void
archive_entry_copy_uname(struct archive_entry *a, const char *name);

void
archive_entry_copy_uname_w(struct archive_entry *a,
    const wchar_t *name);

int
archive_entry_update_uname_utf8(struct archive_entry *a,
    const char *name);

void
archive_entry_fflags(struct archive_entry *a, unsigned long *set_bits,
    unsigned long *clear_bits);

const char *
archive_entry_fflags_text(struct archive_entry *a);

void
archive_entry_set_fflags(struct archive_entry *a,
    unsigned long set_bits, unsigned long clear_bits);

const char *
archive_entry_copy_fflags_text(struct archive_entry *a,
    const char *text);

const wchar_t *
archive_entry_copy_fflags_text_w(struct archive_entry *a,
    const wchar_t *text);

πŸ’… DESCRIPTION πŸ’…

🫢 User id, group id and mode

so like, archive_entry_uid(), archive_entry_gid(), and archive_entry_perm() are the little grabby hands πŸ™Œ that pull the user id, group id and permission out of the given entry.

and then their matching besties archive_entry_set_uid(), archive_entry_set_gid(), and archive_entry_set_perm() put the given user id, group id and permission into the entry. slay.

also, tiny lil side quest: the permission is also set as a side effect of calling archive_entry_set_mode(). iykyk.

archive_entry_strmode() returns a string representation of the permission, exactly like the long mode of ls(1) does. it's giving -rw-r--r-- realness πŸ‘‘

πŸ¦‹ User and group name

user and group names can be provided in one of three different ways, pick your fighter:

  • char * β€” multibyte strings in the current locale. classic.
  • wchar_t * β€” wide character strings in the current locale. the accessor functions are named XXX_w(). (the _w is the whole personality ✨)
  • UTF-8 β€” Unicode strings encoded as UTF-8. these are convenience functions that update both the multibyte and wide character strings at the same time. multitasking queen πŸ’–

archive_entry_set_XXX() is an alias for archive_entry_copy_XXX(). the strings are copied, so they don't need to outlive the call. no lurking pointers, we love a clean copy 🧚

🌈 File Flags

file flags get transparently converted between a bitmap representation and a textual format, which is honestly so versatile of them. for example, if you set the bitmap and ask for text, the library will build a canonical text format. but β€” and hear me out β€” if you set a text format and request a text format, you get back the same text, even if it's ill-formed. no judgment, it's just what it is 😭

if you need to canonicalize a textual flags string, the ritual is:

  1. set the text form,
  2. request the bitmap form,
  3. then use that to set the bitmap form.

setting the bitmap format will clear the internal text representation and force it to be reconstructed when you next request the text form. periodt.

the bitmap format consists of two integers β€” one containing bits that should be set, the other specifying bits that should be cleared. bits not mentioned in either bitmap will be ignored. usually, the bitmap of bits to be cleared will be set to zero. in unusual circumstances (spooky πŸ‘€), you can force a fully-specified set of file flags by setting the bitmap of flags to clear to the complement of the bitmap of flags to set.

ok important note bestie: this differs from fflagstostr(3), which only includes names for set bits. don't mix them up πŸ’€

converting a bitmap to a textual string is a platform-specific operation; bits that are not meaningful on the current platform will be ignored.

the canonical text format is a comma-separated list of flag names. archive_entry_copy_fflags_text() and archive_entry_copy_fflags_text_w() parse the provided text and set the internal bitmap values. this is a platform-specific operation; names that are not meaningful on the current platform will be ignored. the function returns a pointer to the start of the first name that was not recognized, or NULL if every name was recognized.

🚨 and this is the part everyone gets wrong: every name β€” including names that follow an unrecognized name β€” will be evaluated, and the bitmaps will be set to reflect every name that is recognized. (in particular, this differs from strtofflags(3), which stops parsing at the first unrecognized name.) she does NOT quit early, she's a completionist πŸ’…

πŸ‘‘ SEE ALSO πŸ‘‘


πŸ’€ BUGS πŸ’€

the platform types uid_t and gid_t are often 16 or 32 bit wide. in this case it is possible that the ids can not be correctly restored from archives and get truncated. like. the numbers literally do not fit. truncation era, and not in a cute way 😭


okay that's the whole page, you ate it fr fr πŸ“ go forth and archive responsibly, bestie πŸ’œ

with love, Debian β€” dated February 2, 2012 πŸŽ€βœ¨

πŸ“¦ where this page comes from