[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

Roman Gilg  changed:

   What|Removed |Added

 Status|UNCONFIRMED |RESOLVED
 Resolution|--- |FIXED

--- Comment #18 from Roman Gilg  ---
With

https://cgit.kde.org/kwin.git/commit/?id=6072b0cd6c370749f28418f22a1284de5aba3462

the OSDs have been removed. Therefore no freeze would be possible any longer.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #17 from Roman Gilg  ---
(based on https://wayland.freedesktop.org/ubuntu16.04.html)

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #16 from Roman Gilg  ---
You can use my script for compiling (change the rootdir and INST vars to your
liking, an install in /usr/local though makes KWin directly using the compiled
Xwayland on next launch):

#!/bin/sh

# setup environment for local install:
rootdir=$HOME/Entwicklung/linux/xorg
BLD=$rootdir/build-xwayland_only
#export INST=$rootdir/install
#export INST=/opt/xorg
export INST=/usr/local
export LD_LIBRARY_PATH=$INST/lib
export PKG_CONFIG_PATH=$INST/lib/pkgconfig/:$INST/share/pkgconfig/
export PATH=$INST/bin:$PATH
export ACLOCAL_PATH=$INST/share/aclocal
export ACLOCAL="aclocal -I $ACLOCAL_PATH"
mkdir -p src
mkdir -p $BLD
mkdir -p $INST
mkdir -p $INST/var/log
mkdir -p $ACLOCAL_PATH
export MAKEFLAGS="j5" # run 5 threads, or use your own flags

cd $rootdir/src

# * These three are not necessary with the --disable-documentation flag, and
#   pull 1,400 MB of packages:
# configure: error: Documentation build requested but doxygen not found.
#   Install doxygen or disable the documentation using --disable-documentation
# configure: error: Documentation build requested but xmlto not found.
#   Install xmlto or disable the documentation using --disable-documentation
# configure: error: Documentation build requested but graphviz's dot not found.
#   Install graphviz or disable the documentation using --disable-documentation

git clone git://anongit.freedesktop.org/wayland/wayland
cd wayland
git pull
cd $BLD
mkdir wayland
cd wayland
./../../src/wayland/autogen.sh --prefix=$INST # --disable-documentation
make check
make && make install
cd $rootdir/src

# wayland-protocols:
git clone git://anongit.freedesktop.org/wayland/wayland-protocols
cd wayland-protocols
git pull
cd $BLD
mkdir wayland-protocols
cd wayland-protocols
./../../src/wayland-protocols/autogen.sh --prefix=$INST
make check
make && make install
cd $rootdir/src

#git clone git://anongit.freedesktop.org/wayland/libinput
#cd libinput
#git pull
#cd $BLD
#mkdir libinput
#cd $rootdir/src/libinput
#meson --prefix=$INST $BLD/libinput
#ninja -C $BLD/libinput
#sudo ninja -C $BLD/libinput install
#sudo udevadm hwdb --update
#cd $rootdir/src

# No package 'xfont2' found
git clone git://anongit.freedesktop.org/xorg/lib/libXfont
cd libXfont
git pull
cd $BLD
mkdir libXfont
cd libXfont
./../../src/libXfont/autogen.sh --prefix=$INST
make check
make && make install
cd $rootdir/src

git clone git://anongit.freedesktop.org/xorg/xserver
cd xserver
#git pull
cd $BLD
mkdir xserver
cd xserver
./../../src/xserver/autogen.sh --prefix=$INST --disable-docs
--disable-devel-docs \
  --enable-xwayland --disable-xorg --disable-xvfb --disable-xnest \
  --disable-xquartz --disable-xwin
make check
make && make install
cd $rootdir/src

# Links needed so XWayland works:
mkdir -p $INST/share/X11/xkb/rules
ln -s /usr/share/X11/xkb/rules/evdev $INST/share/X11/xkb/rules/
ln -s /usr/bin/xkbcomp $INST/bin/

# git clone git://anongit.freedesktop.org/wayland/weston
# cd weston
# git pull
# cd $BLD
# mkdir weston
# cd weston
# ./../../src/weston/autogen.sh --prefix=$INST --disable-setuid-install
--with-xserver-path=$INST/bin/Xwayland
# make check # runs Xwayland
# make && make install
# cd $rootdir/src

# Weston configuration:
mkdir -p ~/.config
#cp weston/weston.ini.in ~/.config/weston.ini
#editor ~/.config/weston.ini # edit to set background and turn on xwayland.so
module

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

Martin Flöser  changed:

   What|Removed |Added

   Severity|major   |critical

--- Comment #15 from Martin Flöser  ---
I guess I need to compile the latest XWayland... the output looks exactly like
what I had in my test case.

As it's a freeze, I'm raising to critical.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #14 from Roman Gilg  ---
I just tried and the problem is still there. KWin log with
WAYLAND_DEBUG=server:

[2240489,397] wl_surface@20.attach(wl_buffer@11, 0, 0)
[2240489,410] wl_surface@20.damage(0, 0, 3840, 2160)
[2240489,425] wl_surface@20.frame(new id wl_callback@19)
[2240489,432] wl_surface@20.commit()
[2240505,902]  -> wl_buffer@11.release()
[2240506,079]  -> wl_callb...@19.done(1020125252)
[2240506,089]  -> wl_display@1.delete_id(19)
[2240506,120] wl_surface@20.attach(wl_buffer@11, 0, 0)
[2240506,133] wl_surface@20.damage(0, 0, 3840, 2160)
[2240506,141] wl_surface@20.frame(new id wl_callback@19)
[2240506,150] wl_surface@20.commit()
[2240515,906]  -> wl_pointer@15.button(64, 19836925, 272, 1)
[2240516,972] zwp_pointer_constraints_v1@13.confine_pointer(new id
zwp_confined_pointer_v1@21, wl_surface@20, wl_pointer@15, nil, 2)
[2240517,053]  -> zwp_confined_pointer_v1@21.confined()
[2240531,221] wl_shm@6.create_pool(new id wl_shm_pool@22, fd 98, 4)
[2240531,246] wl_shm_pool@22.create_buffer(new id wl_buffer@23, 0, 1, 1, 4, 0)
[2240531,260] wl_shm_pool@22.destroy()
[2240531,264]  -> wl_display@1.delete_id(22)
[2240531,267] zwp_confined_pointer_v1@21.destroy()
[2240531,296]  -> wl_display@1.delete_id(21)
[2240531,300] zwp_pointer_constraints_v1@13.lock_pointer(new id
zwp_locked_pointer_v1@24, wl_surface@20, wl_pointer@15, nil, 2)
[2240531,345]  -> zwp_locked_pointer_v1@24.locked()

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #13 from Martin Flöser  ---
Could you please check whether the freeze matches #384441 ?

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #12 from Martin Flöser  ---
I'm not able to reproduce the problem, could you please add an output of
WAYLAND_DEBUG, so that I can get the actual sequence of commands?

Unrelated found a crash in another test case:
Thread 1 "testPointerCons" received signal SIGSEGV, Segmentation fault.
0x7f33637ca3be in KWin::QPA::Window::unmap (this=0x560d99c0) at
/home/martin/src/kf5/kde/workspace/kwin/plugins/qpa/window.cpp:113
113
m_shellClient->setInternalFramebufferObject(QSharedPointer());
(gdb) bt
#0  0x7f33637ca3be in KWin::QPA::Window::unmap() (this=0x560d99c0) at
/home/martin/src/kf5/kde/workspace/kwin/plugins/qpa/window.cpp:113
#1  0x7f33637c9fd0 in KWin::QPA::Window::setVisible(bool)
(this=0x560d99c0, visible=false) at
/home/martin/src/kf5/kde/workspace/kwin/plugins/qpa/window.cpp:76
#2  0x7f33744f7dac in QWindow::setVisible(bool) () at
/usr/lib/x86_64-linux-gnu/libQt5Gui.so.5
#3  0x7f3344867b66 in PlasmaQuick::Dialog::setVisible(bool)
(this=0x5603a160, visible=false) at
/home/martin/src/kf5/frameworks/plasma-framework/src/plasmaquick/dialog.cpp:1379
#4  0x7f3344868476 in PlasmaQuick::Dialog::qt_static_metacall(QObject*,
QMetaObject::Call, int, void**) (_o=0x5603a160,
_c=QMetaObject::WriteProperty, _id=9, _a=0x7fffa7f0)
at
/opt/build/kf5/frameworks/plasma-framework/src/plasmaquick/KF5PlasmaQuick_autogen/include/moc_dialog.cpp:346
#5  0x7f334486861d in PlasmaQuick::Dialog::qt_metacall(QMetaObject::Call,
int, void**) (this=0x5603a160, _c=QMetaObject::WriteProperty, _id=9,
_a=0x7fffa7f0)
at
/opt/build/kf5/frameworks/plasma-framework/src/plasmaquick/KF5PlasmaQuick_autogen/include/moc_dialog.cpp:399
#6  0x7f337aaaef52 in QQmlPropertyPrivate::write(QObject*, QQmlPropertyData
const&, QVariant const&, QQmlContextData*,
QFlags) (object=object@entry=0x5603a160,
property=..., value=..., context=context@entry=0x561d1fa0, flags=...) at
qml/qqmlproperty.cpp:1259
#7  0x7f337aab0693 in QQmlPropertyPrivate::writeValueProperty(QObject*,
QQmlPropertyData const&, QVariant const&, QQmlContextData*,
QFlags) (object=0x5603a160, core=...,
value=..., context=0x561d1fa0, flags=...) at qml/qqmlproperty.cpp:1183
#8  0x7f337ab1cef5 in QQmlBinding::write(QQmlPropertyData const&,
QV4::Value const&, bool, QFlags)
(this=this@entry=0x561174c0, core=..., result=..., isUndefined=, flags=...)
at qml/qqmlbinding.cpp:333
#9  0x7f337ab1df67 in
QQmlBinding::update(QFlags)
(this=0x561174c0, flags=...) at qml/qqmlbinding.cpp:197
#10 0x7f337ab1e9ce in QQmlBinding::refresh() (this=) at
qml/qqmlbinding.cpp:409
#11 0x7f337aaf3934 in QQmlNotifier::emitNotify(QQmlNotifierEndpoint*,
void**) (endpoint=, a=a@entry=0x0) at qml/qqmlnotifier.cpp:100
#12 0x7f337aa9e61c in QQmlData::signalEmitted(QAbstractDeclarativeData*,
QObject*, int, void**) (object=0x55ef7de0, index=3, a=0x0) at
qml/qqmlengine.cpp:779
#13 0x7f3373cb2524 in QMetaObject::activate(QObject*, int, int, void**)
(sender=0x55ef7de0, signalOffset=,
local_signal_index=, argv=0x0) at kernel/qobject.cpp:3618
#14 0x7f337c580e2b in KWin::OnScreenNotification::visibleChanged()
(this=0x55ef7de0) at
/opt/build/kf5/kde/workspace/kwin/kwin_autogen/EWIEGA46WW/moc_onscreennotification.cpp:220
#15 0x7f337c4bbd8b in KWin::OnScreenNotification::setVisible(bool)
(this=0x55ef7de0, visible=false) at
/home/martin/src/kf5/kde/workspace/kwin/onscreennotification.cpp:115
#16 0x7f337c4bd205 in std::__invoke_impl(std::__invoke_memfun_deref, void (KWin::OnScreenNotification::*&)(bool),
KWin::OnScreenNotification*&, bool&) (__f=@0x55ef69e0: (void
(KWin::OnScreenNotification::*)(KWin::OnScreenNotification * const, bool))
0x7f337c4bbd56 ,
__t=@0x55ef69f8: 0x55ef7de0, __args#0=@0x55ef69f0: false)
at /usr/include/c++/7/bits/invoke.h:73
#17 0x7f337c4bd13f in std::__invoke(void (KWin::OnScreenNotification::*&)(bool),
KWin::OnScreenNotification*&, bool&) (__fn=
@0x55ef69e0: (void
(KWin::OnScreenNotification::*)(KWin::OnScreenNotification * const, bool))
0x7f337c4bbd56 ,
__args#0=@0x55ef69f8: 0x55ef7de0, __args#1=@0x55ef69f0: false)
at /usr/include/c++/7/bits/invoke.h:95
#18 0x7f337c4bd04b in std::_Bind::__call(std::tuple&&, std::_Index_tuple<0ul, 1ul>)
(this=0x55ef69e0, __args=)
at /usr/include/c++/7/functional:467
#19 0x7f337c4bcfcb in std::_Bind::operator()(QTimer::QPrivateSignal&) (this=0x55ef69e0, __args#0=...)
at /usr/include/c++/7/functional:551
#20 0x7f337c4bcec2 in QtPrivate::FunctorCall,
QtPrivate::List, void, 

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #11 from Martin Flöser  ---
OK, I'll try to adjust the test case for this condition.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #10 from Roman Gilg  ---
Thanks for digging into this. You're right that the cursor can't be locked and
confined at the same time.

The Xwayland code works like this: First confine. If then a certain condition
is met (cursor hidden) release the confinement and directly afterwards lock
instead.

Then showing the lock notice instead of the confine notice somehow kills KWin.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #9 from Martin Flöser  ---
Created attachment 107391
  --> https://bugs.kde.org/attachment.cgi?id=107391=edit
Patch to autotest

For reference the patch added to the auto test

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #8 from Martin Flöser  ---
I just created a test case for the situation and this results in:
wl_surface@13: error 1: Surface already constrained

It looks to me like what XWayland does is not valid on Wayland. And I just
checked with the spec
(https://cgit.freedesktop.org/wayland/wayland-protocols/tree/unstable/pointer-constraints/pointer-constraints-unstable-v1.xml):
"  For any surface, only one lock or confinement may be active across all
  wl_pointer objects of the same seat. If a lock or confinement is
requested
  when another lock or confinement is active or requested on the same
surface
  and with any of the wl_pointer objects of the same seat, an
  'already_constrained' error will be raised."

The problem might now be that XWayland gets a protocol error that causes the
hang in KWin as XWayland goes down.

I think the change in XWayland should be reverted or fixed ASAP!

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #7 from Martin Flöser  ---
See http://en.cppreference.com/w/cpp/language/storage_duration

Section "Static local variables"

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #6 from Martin Flöser  ---
It's a static initialization: it's only evaluated the first time it's executed

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #5 from Roman Gilg  ---
You're sure it's created only once? While s_osd is a static
OnScreenNotification pointer, we still always call create() in osd() and in
create() a new OnScreenNotification instance is created with its address being
set as the new value of the static pointer s_osd.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Karl Ove Hufthammer]

https://bugs.kde.org/show_bug.cgi?id=383410

Karl Ove Hufthammer  changed:

   What|Removed |Added

 CC||k...@huftis.org

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #4 from Martin Flöser  ---
Looking at the code everything seems fine. The OSD is only created on very
first show (it's a static instance, so no leak). So the problem might be the
change of the Qml prior to actual showing.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #3 from Martin Flöser  ---
I'm not sure whether we actually support locking and confining at the same
time, but yeah that's not the problem.

It looks like the OSD just doesn't support being shown multiple times. Or in
general that our internal windows don't handle getting shown multiple times.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Roman Gilg]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #2 from Roman Gilg  ---
We get confined and locked probably at the same time, but we setConfined first
in https://cgit.kde.org/kwin.git/tree/pointer_input.cpp#n599, show the confined
OSD and return on
https://cgit.kde.org/kwin.git/tree/pointer_input.cpp#n626

Then on the next call to PointerInputRedirection::update(), which again calls
PointerInputRedirection::enablePointerConstraints(), we go into
https://cgit.kde.org/kwin.git/tree/pointer_input.cpp#n638 and show the other
OSD.

The thing is though when one of the two OSD::show(..) calls get removed, the
code works without problems, so there seems to be no general problem with
confining and locking at the same time.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

--- Comment #1 from Martin Flöser  ---
So the main point here is that we get both confined and locked?

-- 
You are receiving this mail because:
You are watching all bug changes.

[kwin] [Bug 383410] Creating both OSD for pointer confinement + lock in short succesion hangs Wayland session

[Martin Flöser]

https://bugs.kde.org/show_bug.cgi?id=383410

Martin Flöser  changed:

   What|Removed |Added

  Component|wayland-generic |input
  Flags||Wayland+, X11-

-- 
You are receiving this mail because:
You are watching all bug changes.