Small test program and CLUTTER_DEBUG=all output attached. The color of the stage should periodically (1hz) change between blue and yellow. This works as expected when using the x11 backend running in an X session. When changing to "eglnative" and starting from a VT it freezes on blue. The stacktrace is: #0 0x00007ff12a5de8a0 in __read_nocancel () from /nix/store/a5g*-glibc-2.25/lib/libpthread.so.0 #1 0x00007ff12dbccac7 in drmHandleEvent () from /nix/store/kbx*-libdrm-2.4.75/lib/libdrm.so.2 #2 0x00007ff12e462c50 in handle_drm_event.isra.0.part () from /nix/store/j6f*-cogl-1.22.2/lib/libcogl.so.20 #3 0x00007ff12e452823 in cogl_poll_renderer_dispatch () from /nix/store/j6f*-cogl-1.22.2/lib/libcogl.so.20 #4 0x00007ff12e45b737 in cogl_glib_source_dispatch () from /nix/store/j6f*-cogl-1.22.2/lib/libcogl.so.20 #5 0x00007ff12be0ed47 in g_main_context_dispatch () from /nix/store/x7h*-glib-2.50.3/lib/libglib-2.0.so.0 #6 0x00007ff12be0efa0 in g_main_context_iterate.isra () from /nix/store/x7h*-glib-2.50.3/lib/libglib-2.0.so.0 #7 0x00007ff12be0f2c2 in g_main_loop_run () from /nix/store/x7h*-glib-2.50.3/lib/libglib-2.0.so.0 Running 'continue' and sampling a few more stacktraces indicates it's blocking in "__read_nocancel ()" The most suspicious thing I can see in the debug log is the "bail" messages. Clutter-Message: [ +250]:[CLIPPING]:clutter-actor.c:17418: Bail from get_paint_volume (<unnamed>[<ClutterGroup>:0xd8ae10]): Actor failed to report a volume We use clutter v1.26 built with "--enable-egl-native" The gpu tested on is Intel HD Graphics 3000 . We're trying to write a wayland compositor based on clutter and ran into this problem. Narrowing it down the problem occurs without any wayland libraries/interaction though. It might just be that we initiate things incorrectly - if so - feedback is greatly appreciated :) Best regard Ole Jørgen
Attachment:
clutter_debug_output.txt
Description: Text document
Attachment:
minimal-freeze.c
Description: Text Data
Attachment:
build.sh
Description: Bourne shell script