stat/lstat confusion in posix lua module

[keszybz]

posix.stat implements lstat that doesn't follow symlinks. Normal stat that follows symlinks is not available. The first issue is that people would expect stat to provide the semantics of stat, and get unexpected results. The second issue is that stat is not available.

$  strace -y rpmlua -e 'print(posix.stat("/usr/local/sbin").type)' |& rg /usr/local/sbin
newfstatat(AT_FDCWD</home/zbyszek/src/rpm>, "/usr/local/sbin", {st_mode=S_IFLNK|0777, st_size=3, ...}, AT_SYMLINK_NOFOLLOW) = 0

AT_SYMLINK_NOFOLLOW means that this implements lstat semantics.

[pmatilai]

Seems we negletted to comment here.

It's an interesting case. posix.stat() has been unconditionally using lstat() underneath ever since its addition in 2004. Kinda amazing that it took 21 years for anybody to notice, or at least, complain. Sometimes you need lstat() and sometimes you need stat(), and obviously we should provide access to both. It's just that at this point we cannot touch the existing posix.stat() because who knows what disasters may occur.

We could add a flag argument to change the meaning but it'd be kinda reverse from what you expect. Alternatively we'd need to come up with some other name for the thing.

[keszybz]

I agree that touching the existing function is not a good idea.

Ideas for naming:

• add posix.posix_lstat and posix.posix_stat
• add posix.stat_follow and posix.stat_nofollow
• add just posix.stat_follow