Description of problem: KDE fails to start after successful log in. The result is a black screen with the summary message. This message comes from startkde: "# At this point all environment variables are set, let's send it to the DBus session server to update the activation environment /usr/libexec/ksyncdbusenv if test $? -ne 0; then # Startup error echo 'startkde: Could not sync environment to dbus.' 1>&2 test -n "$ksplash_pid" && kill "$ksplash_pid" 2>/dev/null xmessage -geometry 500x100 "Could not sync environment to dbus." exit 1 fi" Version-Release number of selected component (if applicable): plasma-workspace-5.2.0-4.fc22.x86_64 and later (5.2.0-6). How reproducible: always Steps to Reproduce: 1. update plasma-workspace to version 5.2.0-4 or later 2. log out/log in 3. Actual results: black screen with error message Expected results: kde start normally Additional info:
... and downgrading plasma-workspace solves this issue. Martin Kho
Fixed this upstream already some time ago... I wonder how come we don't have this... https://projects.kde.org/projects/kde/workspace/plasma-workspace/repository/revisions/c0ace3a3994ab024ba5301b0c6be24a907d57eaf Can we get it included in our packages Rex?
That patch has been included since * Mon Jan 12 2015 Daniel Vrátil <dvratil> - 5.1.2-5 - Add upstream patch to make ksyncdbusenv work with dbus-1.8.14 Recent changes in packaging are minimal cosmetic: %changelog * Mon Feb 09 2015 Daniel Vrátil <dvratil> - 5.2.0-6 - Revert the previous change * Mon Feb 09 2015 Daniel Vrátil <dvratil> - 5.2.0-5 - Provides/Obsoletes: kdeclassic-cursor-theme * Sun Feb 08 2015 Daniel Vrátil <dvratil> - 5.2.0-4 - Requires: powerdevil, oxygen-sound-theme Reporter, which is that last plasma-workspace build that works ok? 5.2.0-3 ?
Yes, 5.2.0-3 works ok.
And for any failed session, can you post the contents of your ~/.xsession-errors ? that should contain: qDebug() << reply.error().name() << reply.error().message(); which may help determine why the dbus sync failed
Hi Rex, That's not what I get. I have in .xsession-errors (all): "gpg-agent: a gpg-agent is already running - not starting a new one xset: bad font path element (#2), possible causes are: Directory does not exist or has wrong permissions Directory missing fonts.dir Incorrect font server address or syntax startkde: Starting up... QDBusMarshaller::appendVariantInternal: Found unknown D-BUS type '' "org.freedesktop.DBus.Error.InvalidArgs" "Call to UpdateActivationEnvironment has wrong args (, expected a{ss}) " startkde: Could not sync environment to dbus. Warning: Missing charsets in String to FontSet conversion" Martin Kho
B.t.w: dbus-1.8.14-1.fc22.x86_64 and qt-4.8.6-21.fc22.x86_64 (qdbus)
So apparently the qDebug sent: QDBusMarshaller::appendVariantInternal: Found unknown D-BUS type '' "org.freedesktop.DBus.Error.InvalidArgs" "Call to UpdateActivationEnvironment has wrong args (, expected a{ss}) " the first being the error name, the second being the error message. Lukáš, does that help at all? (I suspect not, since that part of the code hasn't changed except for being compiled with gcc5 now).
Aha.. so it fails within the same function but the cause is different (ie. not the path but the type of the argument being passed to UpdateActivationEnvironment).
So the only difference between 5.2.0-3 (working) and 5.2.0-4 (failing) is that it's compiled with GCC 5? :o Martin, what's your Qt5 version? Did you also revert that as part of your experiment?
Hi Lukáš, I think it's 5.4.0. rpm -qa | grep qt5 qt5-qtquickcontrols-5.4.0-1.fc22.x86_64 libkscreen-qt5-5.2.0-2.fc22.x86_64 qt5-qtgraphicaleffects-5.4.0-1.fc22.x86_64 qt5-qtbase-gui-5.4.0-11.fc22.x86_64 phonon-qt5-backend-gstreamer-4.8.2-1.fc22.x86_64 qt5-qtscript-5.4.0-1.fc22.x86_64 phonon-qt5-4.8.3-1.fc22.x86_64 qt5-qtsvg-5.4.0-1.fc22.x86_64 poppler-qt5-0.30.0-2.fc22.x86_64 qt5-qtwebkit-5.4.0-2.fc22.x86_64 qt5-qtx11extras-5.4.0-1.fc22.x86_64 polkit-qt5-1-0.112.0-3.fc22.x86_64 qt5-qtxmlpatterns-5.4.0-1.fc22.x86_64 qt5-qtsensors-5.4.0-2.fc22.x86_64 qt5-style-oxygen-5.2.0-2.fc22.x86_64 qt5-qttools-libs-designer-5.4.0-1.fc22.x86_64 qt5-qttools-common-5.4.0-1.fc22.noarch qt5-qtbase-5.4.0-11.fc22.x86_64 dbusmenu-qt5-0.9.3-0.3.20140604bzr.fc22.x86_64 qt5-qtdeclarative-5.4.0-1.fc22.x86_64 qt5-qtlocation-5.4.0-2.fc22.x86_64 qt5-qttools-5.4.0-1.fc22.x86_64 qt5-qtmultimedia-5.4.0-1.fc22.x86_64
Created attachment 990482 [details] Tentative patch for ksyncdbusenv
%changelog * Wed Feb 11 2015 Rex Dieter <rdieter> 5.2.0-7 - "Could not sync environment to dbus." (startkde) (#1191171)
This didn't fix the issue. The last version that consistently works for me is 5.2.0-3. I have a vague recollection that 1) The problem happened before the new gcc 2) At one time it seems like I just waited until there was a new selinux update. I haven't tried running with selinux off to see if the problem disappears.
Re: comment #14 The only difference between -3 and -4 is: added: Requires: powerdevil oxygen-sound-theme and being built with gcc5 To be clear, you're saying the just-built plasma-workspace-5.2.0-7 fails the same way? if so, can you post the tail snippet from ~/.xsession-errors similar to comment #6 ? hopefully the qDebug output may give a different message this time about the failure.
With 5.2.0-7 the message is startkde: Starting up... QDBusMarshaller::appendVariantInternal: Found unknown D-BUS type '' "org.freedesktop.DBus.Error.InvalidArgs" "Call to UpdateActivationEnvironment has wrong args (, expected a{ss}) " startkde: Could not sync environment to dbus. Warning: Missing charsets in String to FontSet conversion XIO: fatal IO error 11 (Resource temporarily unavailable) on X server ":0" after 212 requests (212 known processed) with 0 events remaining.
Proposed as a Blocker for 22-alpha by Fedora user rdieter using the blocker tracking app because: This bug prevents Plasma5 session start, makes kde spin dead.
showstopper for KDE? +1 blocker then.
Hi, FYI, fwiw, my Fedora 20 + dvratil's plasma-5 copr repo has no problem with plasma-workspace-5.2.0-6.fc20.x86_64, as expected (gcc-4.8.3-7) Martin Kho
Yeah, all indications the problem is specific to f22+ and/or packages built with gcc5.
A theory brought up on IRC was that this might actually be due to the dbus package. If that's the case, then it could be the same bug as F21 bug #1192826.
Discussed at 2015-02-16 blocker review meeting: http://meetbot.fedoraproject.org/fedora-blocker-review/2015-02-16/f22-blocker-review.2015-02-16-17.30.log.txt . As described, accepted as an Alpha blocker as a violation of criterion "A system installed with a release-blocking desktop must boot to a log in screen where it is possible to log in to a working desktop using a user account created during installation or a 'first boot' utility."
Can you test if upgrading to qt5-qtbase-5.4.0-13.fc22 (or newer) helps?
Works with qt5-qtbase-5.4.0-13.fc22
Hi Rex, Congratulations! It works fo me too. Nice work. Martin Kho
Yay, marking closed/fixed in qt5-qtbase-5.4.0-13 %changelog * Mon Feb 16 2015 Rex Dieter <rdieter> 5.4.0-13 - -no-use-gold-linker (f22+, #1193044) essentially a dup of bug #1193044 then, but won't , since this one is considered a blocker (and that one is still open)