Re: master fails to build on FreeBSD when ACL support is on

[Paul Eggert]

Eli Zaretskii wrote:
> Why doesn't copy-file (or qcopy_acl?) use acl_errno_valid, as we do in
> set-file-acl?  Is there a reason for this inconsistency?

set-file-acl is documented to return t on success, nil on ordinary failure, and 
(this part is undocumented) it signals an error on a serious failure. 
acl_errno_valid is used to decide whether the error is serious.

copy-file simply signals an error on failure; its return value is undocumented. 
So it doesn't have the three possibilities that set-file-acl has.

Perhaps copy-file's API should be changed to have those three possibilities, but 
that'd be a fairly big change to its API.