Re: How to enable BSD defined names / functions
Date: Sun, 21 Nov 2021 21:18:10 UTC
On Sun, Nov 21, 2021 at 1:58 PM Konstantin Belousov <kostikbel@gmail.com> wrote: > On Sun, Nov 21, 2021 at 01:51:00PM -0700, Warner Losh wrote: > > There's rather a lot of software that defines _POSIX_C_SOURCE to some > value > > (usually 200809L for POSIX-1.2008), but also wants to use other things. > > > > One can generally work around this issue by defining __BSD_VISIBLE=1, but > > most other systems have something more specific. NetBSD has > _NETBSD_SOURCE. > > OpenBSD has _OPENBSD_SOURCE as well. There's also some expectation that > > _BSD_SOURCE can be defined, but none of the currently active BSDs has > that. > > > > It appears from casual inspection that _NETBSD_SOURCE=1 means > approximately > > the same as __BSD_VISIBLE=1 in FreeBSD. > > > > Would it make sense to add a _FREEBSD_SOURCE=1 case and have it include > > __BSD_VISIBLE=1 regardless of what other macros (especially > > _POSIX_C_SOURCE) are defined to be a more regimented and defined way to > > expand the namespace when multiple namespaces are defined? > > Please note that _BSD_VISIBLE works by presence, and not by value. > Same as things like _GNU_SOURCE. > This stuff is complicated, so I may have overlooked something. _GNU_SOURCE is indeed by presence. But I couldn't find any instances of _BSD_VISIBLE in the tree, just __BSD_VISIBLE where it's tested as simply '#if __BSD_VISIBLE' since it seems to be defined always. Have I overlooked something? > Yes, it makes sense to have _FREEBSD_SOURCE symbol that would imply total > visibility and override any _POSIX_C_SOURCE. > Great. I'll work up a patch. Warner