[11:56] <alan_g> alf__: IIRC the mesa support did/does some unusual stuff with loading symbols?
[11:58] <anpok> i think the interesting part is that the mesa egl library does not link against mirclient
[11:59] <anpok> (while it links against wayland c/s and libX*)
[12:00] <anpok> alan_g: mesa does lib = dlopen(null) dlsym(.. "mir_egl_mesa_display_is_valid") ..
[12:01] <alan_g> anpok: Ah. That's probably what I'm seeing.
[12:01] <alf__> alan_g: anpok: yes, it assumes that libmirclient/server is linked by the executable, and searches for that function at runtime
[12:02] <alan_g> I tried removing the ensure_loaded_with_rtld_global() hack and mesa "blew up"
[12:02] <anpok> then it knows that the surface is of a specific type and uses function pointers there to advance buffers..
[12:02] <anpok> ah
[14:44] <alan_g> alf__: @1455109 - so we change status to confirmed? And push to the kernel team?
[14:47] <alf__> alan_g: that's one way to go, although I still haven't verified that the commit really fixes the problem.
[14:47] <alan_g> alf__: OK, then just update the status for now?
[14:48] <alf__> alan_g: yes
[14:51]  * alan_g is glad that all the cleanup of linkage to static libraries allows us to remove the ensure_loaded_with_rtld_global() hack from libmirclient and libmirplatform
[14:52] <alf__> alan_g: I will mark confirmed in Mir for now, although the kernel crash is of course the kernel's fault.
[14:52] <alf__> alan_g: Not sure what's so special about "sudo mir_demo_server --vt 2 --test-timeout 1 --test-client mir_demo_client_basic"
[14:52] <alf__> alan_g: the command also fails on radeon too (but doesn't crash the kernel)
[14:53] <alan_g> Indeed, especially as I couldn't get the error with the non-installed build (even with use_debflags)
[15:27] <alan_g> davmor2: not nagging, but what timeframe are we expecting for mir-0.13.0?
[15:28] <davmor2> alan_g: I have been blocked for 2 days and the fubar images in the channels so it should be by end of day today hopefully
[15:28] <davmor2> s/and/with
[15:29] <alan_g> Thanks. :)
[15:29] <davmor2> now that we finally have vivid images all round
[16:07] <alan_g> alf__: FYI it is now "vivid all round" ^^