Created attachment 201567 [details] patch Base GCC doesn't support -std=gnu90. Remove it from meson.build to fix build. Tested on powerpc64 and amd64. Hardware sponsored by IntegriCloud.
I don't think it needs to be constrained to GCC architectures, "gnu90" doesn't exist. I'd wager it's a typo for gnu99.
Okay, after a quick Google, it does exist, I'm wrong. I've never seen it used before, and it doesn't make sense for it to be in there.
-std=gnu90 is an alias for -std=gnu89. Removing -std=gnu90 may change "inline" from GNU to C99 semantics as Clang defaults to C11 (ditto GCC >= 5). In this case, it's not important as -std=gnu90 was only used for header sanity check. I've submitted a better fix upstream. https://gitlab.freedesktop.org/libinput/libinput/merge_requests/160
A commit references this bug: Author: zeising Date: Wed Mar 6 21:57:43 UTC 2019 New revision: 494850 URL: https://svnweb.freebsd.org/changeset/ports/494850 Log: x11/libinput: Fix build on GCC-based architectures Fix build of x11/libinput on GCC-based architectures. gnu90 is an alias for gnu89, but is not understood by all versions of GCC, most notably, our base version. Switch to use gnu89 instead. While here, adjust patch context a bit. A similar fix has been merged upstream. PR: 235361 Submitted by: Jan Beich (fix) Reported by: Piotr Kubaj (PR) MFH: 2019Q1 X-MFH-note: If applicable, 2019Q1 might have old libinput version Changes: head/x11/libinput/Makefile head/x11/libinput/files/patch-meson.build
Fix committed, pending MFH. Instead of removing the option, I switched to gnu89, like what Jan submitted upstream.
(In reply to Niclas Zeising from comment #5) Since Q2 has been created, is there any reason to MFH this to Q1 now? And if not, can this PR just be closed?
Yes, it can be closed.