Black Screen on Login
Apparently there's a bug, supposedly with kscreen (unconfirmed), that means as login as a new user you will be greeted with a black screen with plasmashell not drawing any windows. I've heard multiple people comment on various channels, but it seems no-one has had the sense to file a proper bug report so I'm commenting here. If someone can still reproduce this issue can you all please attach _all_ possible information so we can try to find a pattern. Pastebins on IRC really are not enough as they tend to get lost. This bug is super critical so it should be everyone's number 1 priority for the moment. David ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
I have a new theory as to what might be wrong. With KScreen's backend being out of process it could be signalling that there's a new screen before plasmashell has processed the X events and be trying to look for the relevant QScreen object before plasmashell has made it. Could someone put in some debug on shellcorona.cpp:851 before it returns a nullptr in ShellCorona::outputToScreen and see if that gets triggered. You may need to use printf() as plasmashell has this somewhat odd --shutup mode that suppresses all qDebug in an effort to make fixing this as difficult as possible. ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson wrote: > I have a new theory as to what might be wrong. > > With KScreen's backend being out of process it could be signalling that > there's a new screen before plasmashell has processed the X events and be > trying to look for the relevant QScreen object before plasmashell has made > it. > > Could someone put in some debug on shellcorona.cpp:851 before it returns a > nullptr in ShellCorona::outputToScreen and see if that gets triggered. > > You may need to use printf() as plasmashell has this somewhat odd --shutup > mode that suppresses all qDebug in an effort to make fixing this as > difficult as possible. I'll give it a go, as I've reproduced this problem once or twice. Aleix ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson wrote: > I have a new theory as to what might be wrong. > > With KScreen's backend being out of process it could be signalling that > there's a new screen before plasmashell has processed the X events and be > trying to look for the relevant QScreen object before plasmashell has made > it. > > Could someone put in some debug on shellcorona.cpp:851 before it returns a > nullptr in ShellCorona::outputToScreen and see if that gets triggered. > > You may need to use printf() as plasmashell has this somewhat odd --shutup > mode that suppresses all qDebug in an effort to make fixing this as > difficult as possible. My theory, is that ::load is not being executed though. Otherwise we'd be seeing things when this happens, instead of just a black screen. ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 4:30 PM, Aleix Pol wrote: > On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson > wrote: > > I have a new theory as to what might be wrong. > > > > With KScreen's backend being out of process it could be signalling that > > there's a new screen before plasmashell has processed the X events and be > > trying to look for the relevant QScreen object before plasmashell has > made > > it. > > > > Could someone put in some debug on shellcorona.cpp:851 before it returns > a > > nullptr in ShellCorona::outputToScreen and see if that gets triggered. > > > > You may need to use printf() as plasmashell has this somewhat odd > --shutup > > mode that suppresses all qDebug in an effort to make fixing this as > > difficult as possible. > > My theory, is that ::load is not being executed though. Otherwise we'd > be seeing things when this happens, instead of just a black screen. > I'm back to thinking you're right. load would call addOutput. addOutput would assert if outputToScreen failed. With the exception of being completely 100% wrong, it was a good theory. ___ > Plasma-devel mailing list > Plasma-devel@kde.org > https://mail.kde.org/mailman/listinfo/plasma-devel > ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 4:41 PM, David Edmundson wrote: > > > On Fri, Jan 2, 2015 at 4:30 PM, Aleix Pol wrote: >> >> On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson >> wrote: >> > I have a new theory as to what might be wrong. >> > >> > With KScreen's backend being out of process it could be signalling that >> > there's a new screen before plasmashell has processed the X events and >> > be >> > trying to look for the relevant QScreen object before plasmashell has >> > made >> > it. >> > >> > Could someone put in some debug on shellcorona.cpp:851 before it returns >> > a >> > nullptr in ShellCorona::outputToScreen and see if that gets triggered. >> > >> > You may need to use printf() as plasmashell has this somewhat odd >> > --shutup >> > mode that suppresses all qDebug in an effort to make fixing this as >> > difficult as possible. >> >> My theory, is that ::load is not being executed though. Otherwise we'd >> be seeing things when this happens, instead of just a black screen. > > > I'm back to thinking you're right. > load would call addOutput. addOutput would assert if outputToScreen failed. > > With the exception of being completely 100% wrong, it was a good theory. I appreciate a good theory. <3 I just added a warning that will output something if the GetConfigOperation fails. All those of you who can reproduce it, should rebuild and run again. Maybe we'll see the answer there. Aleix ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
Dne 2.1.2015 v 16:47 Aleix Pol napsal(a): On Fri, Jan 2, 2015 at 4:41 PM, David Edmundson wrote: On Fri, Jan 2, 2015 at 4:30 PM, Aleix Pol wrote: On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson wrote: I have a new theory as to what might be wrong. With KScreen's backend being out of process it could be signalling that there's a new screen before plasmashell has processed the X events and be trying to look for the relevant QScreen object before plasmashell has made it. Could someone put in some debug on shellcorona.cpp:851 before it returns a nullptr in ShellCorona::outputToScreen and see if that gets triggered. You may need to use printf() as plasmashell has this somewhat odd --shutup mode that suppresses all qDebug in an effort to make fixing this as difficult as possible. My theory, is that ::load is not being executed though. Otherwise we'd be seeing things when this happens, instead of just a black screen. I'm back to thinking you're right. load would call addOutput. addOutput would assert if outputToScreen failed. With the exception of being completely 100% wrong, it was a good theory. I appreciate a good theory. <3 I just added a warning that will output something if the GetConfigOperation fails. All those of you who can reproduce it, should rebuild and run again. Maybe we'll see the answer there. Can you update to the most recent powerdevil and try to reproduce the black screen problem again? My theory is that there lies the problem; plasmashell is starting up and requesting the config while kded is (also) waiting for plasmashell to show up (because of the notifications it wants to display). Eventually the dbus calls time out, plasmashell start but as there is no config, you just get a black screen. After updating powerdevil, I can no longer reproduce this; kscreen is broken in a different way (doesn't apply the config) but that's another story -- Lukáš Tinkl ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 6:14 PM, Lukáš Tinkl wrote: > Can you update to the most recent powerdevil and try to reproduce the > black screen problem again? My theory is that there lies the problem; > plasmashell is starting up and requesting the config while kded is (also) > waiting for plasmashell to show up (because of the notifications it wants > to display). Eventually the dbus calls time out, plasmashell start but as > there is no config, you just get a black screen. > Notifications should (and do here) fall back to KPassivePopup if the service is not on the bus, it's not waiting for it to appear. Or did you mean other notifications? Cheers -- Martin Klapetek | KDE Developer ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 6:14 PM, Lukáš Tinkl wrote: > Dne 2.1.2015 v 16:47 Aleix Pol napsal(a): > > On Fri, Jan 2, 2015 at 4:41 PM, David Edmundson >> wrote: >> >>> >>> >>> On Fri, Jan 2, 2015 at 4:30 PM, Aleix Pol wrote: >>> On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson wrote: > I have a new theory as to what might be wrong. > > With KScreen's backend being out of process it could be signalling that > there's a new screen before plasmashell has processed the X events and > be > trying to look for the relevant QScreen object before plasmashell has > made > it. > > Could someone put in some debug on shellcorona.cpp:851 before it > returns > a > nullptr in ShellCorona::outputToScreen and see if that gets triggered. > > You may need to use printf() as plasmashell has this somewhat odd > --shutup > mode that suppresses all qDebug in an effort to make fixing this as > difficult as possible. > My theory, is that ::load is not being executed though. Otherwise we'd be seeing things when this happens, instead of just a black screen. >>> >>> >>> I'm back to thinking you're right. >>> load would call addOutput. addOutput would assert if outputToScreen >>> failed. >>> >>> With the exception of being completely 100% wrong, it was a good theory. >>> >> >> I appreciate a good theory. <3 >> >> I just added a warning that will output something if the >> GetConfigOperation fails. >> >> All those of you who can reproduce it, should rebuild and run again. >> Maybe we'll see the answer there. >> > > Can you update to the most recent powerdevil and try to reproduce the > black screen problem again? My theory is that there lies the problem; > plasmashell is starting up and requesting the config while kded is (also) > waiting for plasmashell to show up (because of the notifications it wants > to display). Eventually the dbus calls time out, plasmashell start but as > there is no config, you just get a black screen. > After updating powerdevil, I can no longer reproduce this; kscreen is > broken in a different way (doesn't apply the config) but that's another > story That theory doesn't make sense - plasmashell doesn't query kded for the config. plasmashell links libkscreen, libkscreen starts it's own unique backend process for the current config then we query that. This might delay kscreen setting the screen config, but it shouldn't prevent reading the current state unless there's something else going on... > > -- > Lukáš Tinkl > > ___ > Plasma-devel mailing list > Plasma-devel@kde.org > https://mail.kde.org/mailman/listinfo/plasma-devel > ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Fri, Jan 2, 2015 at 6:14 PM, Lukáš Tinkl wrote: > Dne 2.1.2015 v 16:47 Aleix Pol napsal(a): > >> On Fri, Jan 2, 2015 at 4:41 PM, David Edmundson >> wrote: >>> >>> >>> >>> On Fri, Jan 2, 2015 at 4:30 PM, Aleix Pol wrote: On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson wrote: > > I have a new theory as to what might be wrong. > > With KScreen's backend being out of process it could be signalling that > there's a new screen before plasmashell has processed the X events and > be > trying to look for the relevant QScreen object before plasmashell has > made > it. > > Could someone put in some debug on shellcorona.cpp:851 before it > returns > a > nullptr in ShellCorona::outputToScreen and see if that gets triggered. > > You may need to use printf() as plasmashell has this somewhat odd > --shutup > mode that suppresses all qDebug in an effort to make fixing this as > difficult as possible. My theory, is that ::load is not being executed though. Otherwise we'd be seeing things when this happens, instead of just a black screen. >>> >>> >>> >>> I'm back to thinking you're right. >>> load would call addOutput. addOutput would assert if outputToScreen >>> failed. >>> >>> With the exception of being completely 100% wrong, it was a good theory. >> >> >> I appreciate a good theory. <3 >> >> I just added a warning that will output something if the >> GetConfigOperation fails. >> >> All those of you who can reproduce it, should rebuild and run again. >> Maybe we'll see the answer there. > > > Can you update to the most recent powerdevil and try to reproduce the black > screen problem again? My theory is that there lies the problem; plasmashell > is starting up and requesting the config while kded is (also) waiting for > plasmashell to show up (because of the notifications it wants to display). > Eventually the dbus calls time out, plasmashell start but as there is no > config, you just get a black screen. > > After updating powerdevil, I can no longer reproduce this; kscreen is broken > in a different way (doesn't apply the config) but that's another story > I can't reproduce myself, that's why I just added debug information... Aleix ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
Okay people, I reported this bug on bugzilla https://bugs.kde.org/show_bug.cgi?id=342431 On Sat, Jan 3, 2015 at 3:47 AM, Aleix Pol wrote: > On Fri, Jan 2, 2015 at 6:14 PM, Lukáš Tinkl wrote: >> Dne 2.1.2015 v 16:47 Aleix Pol napsal(a): >> >>> On Fri, Jan 2, 2015 at 4:41 PM, David Edmundson >>> wrote: On Fri, Jan 2, 2015 at 4:30 PM, Aleix Pol wrote: > > > On Fri, Jan 2, 2015 at 4:23 PM, David Edmundson > wrote: >> >> I have a new theory as to what might be wrong. >> >> With KScreen's backend being out of process it could be signalling that >> there's a new screen before plasmashell has processed the X events and >> be >> trying to look for the relevant QScreen object before plasmashell has >> made >> it. >> >> Could someone put in some debug on shellcorona.cpp:851 before it >> returns >> a >> nullptr in ShellCorona::outputToScreen and see if that gets triggered. >> >> You may need to use printf() as plasmashell has this somewhat odd >> --shutup >> mode that suppresses all qDebug in an effort to make fixing this as >> difficult as possible. > > > My theory, is that ::load is not being executed though. Otherwise we'd > be seeing things when this happens, instead of just a black screen. I'm back to thinking you're right. load would call addOutput. addOutput would assert if outputToScreen failed. With the exception of being completely 100% wrong, it was a good theory. >>> >>> >>> I appreciate a good theory. <3 >>> >>> I just added a warning that will output something if the >>> GetConfigOperation fails. >>> >>> All those of you who can reproduce it, should rebuild and run again. >>> Maybe we'll see the answer there. >> >> >> Can you update to the most recent powerdevil and try to reproduce the black >> screen problem again? My theory is that there lies the problem; plasmashell >> is starting up and requesting the config while kded is (also) waiting for >> plasmashell to show up (because of the notifications it wants to display). >> Eventually the dbus calls time out, plasmashell start but as there is no >> config, you just get a black screen. >> >> After updating powerdevil, I can no longer reproduce this; kscreen is broken >> in a different way (doesn't apply the config) but that's another story >> > I can't reproduce myself, that's why I just added debug information... > > Aleix > ___ > Plasma-devel mailing list > Plasma-devel@kde.org > https://mail.kde.org/mailman/listinfo/plasma-devel -- Bhushan Shah http://bhush9.github.io IRC Nick : bshah on Freenode ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
I have a branch in kscreen, david/fixedit in which I fix it for sure, maybe. Can people test that please. David ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
After about 2 gazillion attempts, I'm pleased to report that the bug is fixed. It wasn't in kscreen, or powerdevil...but a quirk in KActivities which became a bug with the kscreen change Big thanks to einar and bshah for relentlessly testing al the patches. ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
Re: Black Screen on Login
On Sun, Jan 4, 2015 at 5:17 PM, David Edmundson wrote: > After about 2 gazillion attempts, I'm pleased to report that the bug is > fixed. > > It wasn't in kscreen, or powerdevil...but a quirk in KActivities which > became a bug with the kscreen change > > Big thanks to einar and bshah for relentlessly testing al the patches. And big thanks to you! For worrying :D Aleix ___ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel