Paul Eggert wrote: > On 4/9/25 04:09, Corinna Vinschen wrote: > > I would like to ask you to reconsider and to remove the > > Cygwin-special code and let Cygwin decide by itself, what a trivial vs. > > non-trivial ACL is by using the POSIXy functions Cygwin provides. > > Just use acl_extended_file(), please. > > Sorry, I've lost context. I assume you're not suggesting that we revert > the patch I recently installed
No, that's not what Corinna means. It's about whether Gnulib's file-has-acl module (and, by consequence, the coreutils 'ls' program) shows freshly created files and directories (e.g. "mkdir foo", "echo > bar") as having ACLs. My answer is "it should not", and my arguments are in <https://sourceware.org/pipermail/cygwin/2025-March/257765.html>. Her answer is "it should", and her arguments are in <https://sourceware.org/pipermail/cygwin/2025-April/257902.html>, referring to <https://sourceware.org/pipermail/cygwin/2025-March/257762.html>. These points of view are not immediately reconcilable. But what I could do is to change Gnulib's file-has-acl module so that if the environment variable CYGWINLY_PEDANTIC is set, it implements Corinna's point of view. (It's not a large change.) The effects for a user who has set this environment variable would be: - "ls -l" shows the presence of an ACL on nearly all files. - The gnulib unit tests of this module and of 'copy-acl' fail. The CYGWINLY_PEDANTIC name is meant to imitate POSIXLY_CORRECT, but here we are not talking about POSIX (since POSIX:2024 does not specify ACLs [1]), and what is "correct" is subjective. Then users can choose which behaviour of 'ls' they prefer... Bruno [1] https://en.wikipedia.org/wiki/Access-control_list#POSIX_ACL
