Bug 754501 – Trying to run a Qt app natively on Wayland session crashes it

After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.

Bug 754501 - Trying to run a Qt app natively on Wayland session crashes it


Summary:	Trying to run a Qt app natively on Wayland session crashes it


Status:	RESOLVED DUPLICATE of bug 757623

Product:	mutter
Classification:	Core
Component:	wayland
Version:	3.17.x
Hardware:	Other Linux

Importance:	Normal normal
Target Milestone:	---
Assigned To:	mutter-maint
QA Contact:	mutter-maint

URL:
Whiteboard:

Depends on:
Blocks:

Reported:	2015-09-03 08:29 UTC by Dmitry Shachnev
Modified:	2015-12-22 00:23 UTC

See Also:
GNOME target:	---
GNOME version:	---

Attachments
Output of a test Qt cleint (6.69 KB, text/plain) 2015-09-04 09:46 UTC, Dmitry Shachnev	Details

Description Dmitry Shachnev 2015-09-03 08:29:57 UTC

When trying to run any Qt 5 application natively (i.e. with "-platform wayland" added to command line) on GNOME Wayland session (3.17.90), the whole shell crashes (and I get the GDM login screen).

I don't have the stacktrace because I didn't figure out how to obtain it. If someone tells me how I can debug this, I will do it.

This is a semi-regression from 3.16: in that version, the Qt app itself receives a "Cannot commit a NULL buffer to an xdg_surface" message from Mutter and exits.

I tested on Debian unstable with some packages from experimental. Software versions used:

mutter 3.17.90
gnome-shell 3.17.90
gnome-session 3.16.0
qtbase 5.5.0
qtwayland 5.5.0

Running any Qt application (i.e. Qt Assistant or Qt Designer) with -platform wayland is enough to reproduce the crash (without -platform wayland they would be run under XWayland, where they don't crash).

The same applications run just fine in Weston session.

Comment 1 Dmitry Shachnev 2015-09-03 15:16:48 UTC

I forgot to mention a version of libwayland-client that Qt is using: it's 1.8.1.

Comment 2 Dmitry Shachnev 2015-09-04 09:20:57 UTC

I have finally managed to get a stacktrace:

Program received signal SIGSEGV, Segmentation fault.
meta_window_set_title (window=0x0, title=0x369847c "") at core/window.c:7292

+ Trace 235415

#0 meta_window_set_title
at core/window.c line 7292
#1 ffi_call_unix64
at ../src/x86/unix64.S line 76
#2 ffi_call
at ../src/x86/ffi64.c line 525
#3 wl_closure_invoke
at ../src/connection.c line 942
#4 wl_client_connection_data
at ../src/wayland-server.c line 335
#5 wl_event_loop_dispatch
at ../src/event-loop.c line 419
#6 wayland_event_source_dispatch
at wayland/meta-wayland.c line 85
#7 g_main_context_dispatch
at /build/glib2.0-eaKVfK/glib2.0-2.45.6/./glib/gmain.c line 3154
#8 g_main_context_dispatch
at /build/glib2.0-eaKVfK/glib2.0-2.45.6/./glib/gmain.c line 3769
#9 g_main_context_iterate
at /build/glib2.0-eaKVfK/glib2.0-2.45.6/./glib/gmain.c line 3840
#10 g_main_loop_run
at /build/glib2.0-eaKVfK/glib2.0-2.45.6/./glib/gmain.c line 4034
#11 meta_run
at core/main.c line 437
#12 main
at main.c line 463

Comment 3 Jonas Ådahl 2015-09-04 09:37:55 UTC

(In reply to Dmitry Shachnev from comment #2)
> I have finally managed to get a stacktrace:
> 
> Program received signal SIGSEGV, Segmentation fault.
> meta_window_set_title (window=0x0, title=0x369847c "") at core/window.c:7292

Could you run the client with the WAYLAND_DEBUG=1 and attach the output here?

Comment 4 Dmitry Shachnev 2015-09-04 09:46:04 UTC

Created attachment 310655 [details]
Output of a test Qt cleint

(In reply to Jonas Ådahl from comment #3)
> Could you run the client with the WAYLAND_DEBUG=1 and attach the output here?

Here it is.

Comment 5 Dima Ryazanov 2015-12-03 10:10:12 UTC

*** Bug 757623 has been marked as a duplicate of this bug. ***

Comment 6 Jonas Ådahl 2015-12-22 00:23:18 UTC

Closing this one as dup of 757623 because I happened to attach the patches to that bug instead of this one.

*** This bug has been marked as a duplicate of bug 757623 ***