See review D20099 for the patch. This bug is for better visibility. Please, don't reply with empty promises to review.
Hi, any updates on this? I'd like to remove the workaround we have in place for FreeBSD's outdated libwayland.
Can someone from x11@ give Jan approval to commit this? I think this has been open for way too long.
I will have a look at this, but there are other things (mesa and xorg infrastructure) taht's above this on my todo.
(In reply to Niclas Zeising from comment #3) > I will have a look at this, but there are other things (mesa and xorg > infrastructure) taht's above this on my todo. What QA would you do and is there anything I can help with?
I've added the update + a patch to use SHM_ANON instead of posix_fallocate(), since posix_fallocate() doesn't work with ZFS.
Ping? Still not sure what QA I should do if any. This update has not caused any problems for me on 13.0 for the last 2 months with Sway etc.
comment 5 pushed https://github.com/FreeBSDDesktop/freebsd-ports/commit/689501412ec9 on the same bandwagon. x11/waybar seems to crash more often after that change but not like Waybar was very stable before. Other Wayland clients appear to work fine. Does anyone see something similar? $ sway $ waybar ** GLib-GIO:ERROR:giostream.c:255:g_io_stream_get_output_stream: assertion failed: (klass->get_output_stream != NULL) Process 89916 stopped * thread #3, name = 'waybar', stop reason = signal SIGABRT frame #0: 0x000000080122e5ca libc.so.7`__sys_thr_kill at thr_kill.S:4 (lldb) bt * thread #3, name = 'waybar', stop reason = signal SIGABRT * frame #0: 0x000000080122e5ca libc.so.7`__sys_thr_kill at thr_kill.S:4 frame #1: 0x000000080122df44 libc.so.7`__raise(s=6) at raise.c:52:10 frame #2: 0x000000080118f6f9 libc.so.7`abort at abort.c:67:8 frame #3: 0x000000080235cb56 libglib-2.0.so.0`g_assertion_message(domain="GLib-GIO", file="giostream.c", line=255, func="g_io_stream_get_output_stream", message="assertion failed: (klass->get_output_stream != NULL)") at gtestutils.c:2532:5 frame #4: 0x000000080235cbd4 libglib-2.0.so.0`g_assertion_message_expr(domain="GLib-GIO", file="giostream.c", line=255, func="g_io_stream_get_output_stream", expr="klass->get_output_stream != NULL") at gtestutils.c:2555:3 frame #5: 0x000000080211cf64 libgio-2.0.so.0`g_io_stream_get_output_stream(stream=0x000000080397aa20) at giostream.c:255:3 frame #6: 0x000000080219838c libgio-2.0.so.0`_g_dbus_auth_run_client(auth=0x00000008038aa0a0, observer=0x0000000000000000, offered_capabilities=G_DBUS_CAPABILITY_FLAGS_UNIX_FD_PASSING, out_negotiated_capabilities=0x00000008038b7238, cancellable=0x00000008034a1c70, error=0x00000008038b71c8) at gdbusauth.c:616:9 frame #7: 0x00000008021aca2c libgio-2.0.so.0`initable_init(initable=0x00000008038b7160, cancellable=0x00000008034a1c70, error=0x00007fffdfdfcf08) at gdbusconnection.c:2542:26 frame #8: 0x0000000802116a35 libgio-2.0.so.0`g_initable_init(initable=0x00000008038b7160, cancellable=0x00000008034a1c70, error=0x00007fffdfdfcf08) at ginitable.c:128:10 frame #9: 0x00000008020d5484 libgio-2.0.so.0`async_init_thread(task=0x0000000803851280, source_object=0x00000008038b7160, task_data=0x0000000000000000, cancellable=0x00000008034a1c70) at gasyncinitable.c:260:7 frame #10: 0x000000080215cabe libgio-2.0.so.0`g_task_thread_pool_thread(thread_data=0x0000000803851280, pool_data=0x0000000000000000) at gtask.c:1331:3 frame #11: 0x0000000802361aa7 libglib-2.0.so.0`g_thread_pool_thread_proxy(data=0x00000008038e5b00) at gthreadpool.c:307:15 frame #12: 0x000000080236057d libglib-2.0.so.0`g_thread_proxy(data=0x0000000803903c50) at gthread.c:784:20 frame #13: 0x000000080052a776 libthr.so.3`thread_start(curthread=0x000000080382a900) at thr_create.c:291:16
Another victim: net/waypipe ca no longer forward EGL clients with SHM_ANON fix. I still need someone to confirm as my environment is too weird.
(In reply to Jan Beich from comment #8) Is this related to https://github.com/FreeBSDDesktop/kms-drm/issues/156 ?
No. kms-drm#156 was filed a few days before coment 5.
Hmm, I can't reproduce comment 7 and comment 8 anymore. No one else confirmed, so those could either be due to a -CURRENT bump, fixed by a dependency update, intermittent (e.g., related to libepoll-shim), specific to old kms-drm versions or a pilot error (optimizations, partial upgrade, etc).
Hi, Any update on this? As a side note, 1.18 will be release soon: https://lists.freedesktop.org/archives/wayland-devel/2020-January/041128.html Thanks
Niclas, how do we proceed from here? There has been a long silence from x11@. Per comment #11, Jan no longer has the problem from comment #7, so I do not think anything is really blocking this? (In any case the SHM_ANON patch is trivially revertable too if anyone runs into problems with it.) Should we ask for an exp-run and then ship it?
(In reply to Tobias Kortkamp from comment #13) There is no need for an exp-run honestly. I've said to Jan on saturday that I will commit after testing and I haven't had time to test for now. If it's not commited tomorow I'll take an hour and test/commit.
I'm working on it. I got side tracked yesterday.
A commit references this bug: Author: zeising Date: Tue Jan 28 11:53:09 UTC 2020 New revision: 524473 URL: https://svnweb.freebsd.org/changeset/ports/524473 Log: graphics/wayland: Update to 1.17 Update graphics/wayland to 1.17 [1] Add a patch to use shm_anon instead of posix_fallocate(), as the latter doesn't work on ZFS. [2] PR: 237650 [1], 224226 [2] Submitted by: jbeich [1], jbeich and Greg V [2] (original version) Reviewed by: Greg V, manu Differential Revision: https://reviews.freebsd.org/D20099 Changes: head/graphics/wayland/Makefile head/graphics/wayland/distinfo head/graphics/wayland/files/patch-Makefile.am head/graphics/wayland/files/patch-cursor_os-compatibility.c head/graphics/wayland/files/patch-tests_sanity-test.c head/graphics/wayland/files/patch-tests_test-runner.c
Leave this open for a bit, just in case.
Thank you.
No reports of fallout, closing this.