Registered Member
|
Hi everyone !
First time poster, I'm not here to complain, but to add my voice on some issues, and maybe (maybe) find a few solutions... I'm on Kubuntu 15.04, with KDE 5.3.1 backported, but with Openbox as a WM (for speed issues, as my GPU is somewhat low quality/old (Geforce 8400 GS)). I'm not really a noob, I have looked for those issues, but found nothing helpful. So here we go, the list of the issues that are plaguing my installation, in no particular order :
Anyway, that's all for me now. There are a lot more of tiny bugs, but I can live with them. Those listed above are just those that are turning my hair white. If you have any solution for those, I would be eternally thanksful. If not, don't worry, I just wanted to list them so if someone look for them, he doesn't feel alone |
Registered Member
|
Sorry to bump this topic up (I swear I'll only do it once), but I hoped for a bit more than 0 response, even if I didn't expect a solution to all my issue.
Also, I'll add an issue : is there a way for the taskbar to remember the order of the tasks/apps when I choose to reorder them myself and plasmashell crashes ? |
|
"My list of problems" will simply scare people.
> 1) plasmashell always take ~30% of the CPU. Pretty much every plasmoid with constant animations/repaints has potential to cause this. You can either run plasmashell in "valgrind --tool=callgrind" or remove plasmoids (notably non-stock ones) one by one and see when the CPU load stops (I'd start with your own one; it's usually always ones own fault > 2) Every few hours, every dolphin instances together maxs out the CPU. Sure it's dolphin and not baloo indexing stuff? You could gdb into dolphin to check what its doing when doing too much:
"pkill -SIGSTOP dolphin" will stop every dolphin process (in case you've trouble with gdb taking too long while the system is loaded) > 3) The scroll seems broken in KDE 5. IIrc, I had that issue in kate/kwrite in early versions (presently 15.04 -KDE applications, not ubuntu- on KF 5.11.0 and Qt 5.4.1) but not seen it since a long time. Unless this is a bug in the Qt event handler, these will however be different issues (event interpretation is a client code job and kwrite doesn't have much in common with the QtQuick stuff plasmashell uses) > 5) The taskbar doesn't autohide. Ever. Some (other) plasmoid will keep the panel shown. Wild guess would be the systray or the notes plasmoid. Try an empty panel and then add plasmoids until you found the troublemaker. > 6) not one tooltip, notification, or even "Add graphic components" panel show at the same place over time See eg. https://bugs.kde.org/show_bug.cgi?id=347439 or https://bugs.kde.org/show_bug.cgi?id=337790 Bottom line is that geometry setting via QML seems buggy on multiscreens, but I don't know what's wrong there either. > 7) weather widget and yahoo weather widget. I'd report that to their authors - sounds like they query the weather service/s a bit too "intensively" or the service cuts the tcp connection after a while. |
Registered Member
|
I agree that the list of problems could scare people, but I had the choice between that, and spamming the forum. I choose the less evil of the two.
This time, it's not my fault ;) https://bugs.kde.org/show_bug.cgi?id=345696 https://bugs.kde.org/show_bug.cgi?id=346134 That the only issue I had the cause, but not the solution. Yes removing the plasmoids is a solution, but not an usable one. And no, drawing a plot only every second shouldn't take 25% CPU.
Yes. I see each instance of Dolphin doing this in top. I'll try gdb, it's a good advice.
I'm not sure what the cause is. But I see similar randomness in the way I scroll on taskbar, on the desktop, in kwrite, in kate, in konsole... Yes, they don't use QtQuick, but maybe they are using the same org.kde.plasma.components, the scrollable areas, etc. Or it is a Qt Event handler. In any case, it's buggy...
That could have been a good idea. But, 1) I don't have weird plasmoids on the panel, and 2) I tried with an empty panel, and that didn't work either.
No, because when I kill and restart immediately plasmashell, they work totally correctly. A server block that is solved by a restart of plasmashell doesn't seem probable. |
|
> https://bugs.kde.org/show_bug.cgi?id=345696
> https://bugs.kde.org/show_bug.cgi?id=346134 I thought that wouldn't "solve anything"? Anyway, from the second bug I take the widget calls "update" in a row and while "update" indeed compresses repaints, it still maintains a refresh rate of afair 50fps, so you're not updating this once a second, but 50 times a second with the same value in that case. > maybe they are using the same org.kde.plasma.components Neither Konsole nor Kate link plasma or qtquick. It's a bug in the Qt eventhandling or disjunct problems. > I tried with an empty panel, and that didn't work either. Well, either there's a bug in plasmashell or it falls into the potential Qt eventhandling category (notably both are mouse event caused) You could just file a respective bug and the developers of the plasmashell panel will either know "ah, yes - was bug foobar" or might have an idea why it could fail. It's unlikely gonna be resolved here, given nobody yet knew. > A server block that is solved by a restart of plasmashell doesn't seem probable. When you restart plasmashell, the plasmoids will trivially be enforced to create a new TCP connection. It's not like accuweather or yahoo would be down, but that the tcp connection might be closed from their side (and the plasmoids are not creating a new one) for some quota or timeout reason. PS: multiple posts are actually the less evil solution since it allows people to focus on one topic. The forum won't run out of topic IDs - don't worry |
Registered Member
|
This is typically the wrong behaviour by people in the "old-style" forums. I did it this way too all the time. In fact you should open for every problem one thread itself. This not only gives the forum and the threads more structure but also helps other people finding a possible solution more easily with the forum search or google.
|
Registered Member
|
A looonnng time ago I got insulted because I did it the other way around. Each forum has its rules, it's hard to know in which we should post in one post, and in which we should post a post per issue. Anyway, should I repost those issues one at a time now ?
Yes it doesn't solve anything, because I want/need those plasmoids. If the solution is to remove a functionnality, it's not a solution. Because killing plasmashell and not restarting it would also be a solution in this case. As for the 60Hz refresh rate, yes, the screen is painted 60 times per seconds. The whole screen. In fact both screens. But it eats at most 1% of CPU, whatever is on screen. Add one plasmoid, 30% of CPU. Something wrong here. And the data is pulled only once every second.
Yes, but they both use the same underlying Qt code, directly through the C++ lib, or through QtQuick/Plasma/Whatever.
That's why I posted here : I wanted to know if anybody knew anything about my issues, before I spam the bugtracker with those. As it seems most of them aren't known, maybe I should file a few bugs
Hum... Ok, I will report those to their respective authors |
|
> Yes it doesn't solve anything, because I want/need those plasmoids.
Of course, but asking about a bug that you already know is pretty pointless - so I took your statement as "I know this bug, but it's not what I experience" > As for the 60Hz refresh rate, yes, the screen is painted 60 times per seconds. You're interelaving unrelated things here. The screen refresh rate has nothing to do with how often a repaint is really triggered. Former only reposts the scanout buffer of your GPU to the output(s) but the latter causes CPU overhead (notably if rasterizing the data is expensive). If those elements repaint 50 times a second (for no actual reason) it's no wonder they stress the CPU. > Yes, but they both use the same underlying Qt code Yes - the Qt event dispatcher. I just wanted to ensure you understood that the event interpretation code in those items is so much different, that it's either two entirely different bugs or a bug in Qt and that they are notably *not* "using the same org.kde.plasma.components" - kate & konsole do not use QML at all. > before I spam the bugtracker with those. Bugzilla has btw. a pretty powerful search feature (though, frankly, google is smarter on searching bugs; because of it's heuristic near match results) |
Registered users: abc72656, Bing [Bot], daret, Google [Bot], Sogou [Bot], Yahoo [Bot]