| Summary: | Desktop Effects were restarted due to a graphics reset... | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | PGillespie <vermontpoet> |
| Component: | general | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | VERIFIED DUPLICATE | ||
| Severity: | grave | CC: | bhush94, plasma-bugs |
| Priority: | NOR | ||
| Version: | 5.5.5 | ||
| Target Milestone: | --- | ||
| Platform: | Other | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
it's not a bug at all - kwin noticed that your gpu too a nap and restarts the compositor. it's certainly not related to the other bug but might have a common source: far too many gl contexts (thanks to qtquick) and the driver not suited for that load. try to enable dri3 and use sna instead of glamor acceleration, https://www.reddit.com/r/archlinux/comments/3pey73/graphical_corruption_on_gnome_with_intel_graphics/ //'s not a bug at all - kwin noticed that your gpu too a nap and restarts the compositor.// Except when my gpu doesn't take a nap... As I wrote above, it can occur "seemingly at random" and not always when recovering from suspend. Should I file another bug report? This is a misunderstanding - I didn't mean to imply an STR condition by "took a nap". glGetGraphicsResetStatusARB (this is part of the robust gl contexts) returns some error, ie. it *was* reset. The reset can be caused (anytime) by kwin or not (you didn't provide the exact warning), but neither the error message itself nor the compositor restart is a bug and is NOT responsible for causing failures in other GL contexts. (it *might* induce them, but that's a bug in the driver or the other gl context - and it's more likely the reset is the direct cause) Well, there's extreme instability after every suspend. Almost without exception the compositor restarts without warning (or error message) and sometimes not at all. Just this morning, waking up my laptop, I tried to demonstrate accessing TTY to my kids (because of the disappearing mouse bug) and arrived at a blank screen (no prompt), when I returned to plasma, compositing was off (due to a compositing error--as stated in settings) and I could discover no workaround to recover any functionality. Restarting plasmashell didn't work and ALT+Shift+F12 failed to toggle the compositor. Windows were opening without borders or not opening at all, and the panel was non-functional. Had to reboot. Unless I hear otherwise, I'm opening another bug report. If you think the current bug should be renamed, that's your call. Okay, after some research, it appears that this bug might be a duplicate of the following: https://bugs.kde.org/show_bug.cgi?id=357551 Which is itself a duplicate of: https://bugs.kde.org/show_bug.cgi?id=356394 I notice that this has been resolved upstream in QT 5.6. Any idea when Kubuntu Backports will be available to Xenial, or KDE NEON. *** This bug has been marked as a duplicate of bug 356394 *** That's the mighty bug #341497 but won't cause or relate to a graphics reset/compositor restart. No idea about Ubuntu policy, but there're pending bugs even in Qt 5.6 itr (at least partially, see https://bugreports.qt.io/browse/QTBUG-42803, resolved in 5.6.1) crashing every QtQuick client (incl. plasmashell & KWin) I'm inclined disagree with you based on Bug 357551, which perfectly describes my symptoms and was itself marked as a duplicate of Bug 356394 by David Edmundson (take it up with *him* maybe?): "After waking from suspend Plasma is always broken, most commonly the pannel and window borders are gone. In the best case they restart, and the bug tracker pops up, but sometimes I am left with a broken session with no pannel, borders, or way to switch windows. Once in a while I get a blank screen and can't even switch to a tty. I also can't switch to TTY. I see the login prompt but can't interact with the session. Should I file a separate bug on this? Quatsch... my legs are getting stiff and I'm tired of trying to sort this out. If I don't hear anything definite concerning the bug with TTY, I guess I'll file a separate bug report including everything above. Not a threat, it's just what it is. It's a bug. (In reply to PGillespie from comment #7) > "After waking from suspend Plasma is always broken, most commonly the pannel > and window borders are gone. In the best case they restart, and the bug > tracker pops up That's not even remotely the same. The process (kwin_x11 or plasmashell) segfaulted and - ideally - caught the segfault to restart itself. That's fundamentally different from "kwin got a gpu reset signal from the driver and "restarted" (ie. deleted and reconstructed) the compositor GL context. The latter one will not affect the process itself at all - eg. restarting (suspend/resume) the compositor is even a regular feature of kwin. Here's what I can tell you: 1.) The missing text issues have seemingly gone away -- solved by updates? 2.) Bug 357551 describes the remaining symptoms: "sometimes I am left with a broken session with no pannel, borders, or way to switch windows. Once in a while I get a blank screen and can't even switch to a tty." 3.) If you want to parse out the "Desktop Effects were restarted due to a graphics reset," then take it up with my laptop. Of note: I have not received "Desktop Effects" error messages since reporting this bug yet the problem continues. The compositor simply crashes alongside the symptoms described in Bug 357551. And TTY doesn't work. Yet you insist these aren't related!?! Then it's a new bug. I'm reporting it. Done. Two new bugs reported: Bug 362588 & Bug 362589 And that's that. If you or anyone wants to go deeper, I'm available. All I can do is to report. Thanks and good luck. wtf is so hard to understand about "the graphics reset messages do not describe the process crash"? if the process crashes (and restarts) you should get a crash dialog with a backtrace indicating the reason of the crash. whether or not that is the darn qt Bug that crashes every application whenever you touch the randr config (what can happen after an STR) can be said by looking at that bug. but you can expect a hell lot of crashes in random applications w/ qt 5.5, yes. "i noticed that the gpu reset and therefore restart the compositor" otoh is not a bug in kwin at all but desired behavior since the gpu is in an uncertain state. the two things are just really not related and only the former is a bug (in qt) //wtf is so hard to understand about "the graphics reset messages do not describe the process crash"?// And as to wtf's so hard about it? I'm not a developer/programmer or hacker. I'm a builder and writer who used to work as a tech support technician in the early 1990's when Windows vs. DOS was still open to debate. So yeah, it's hard to understand. I've spent several hours this morning contributing my time reporting bugs/problems I'm experiencing with KDE/Kubuntu 16.04, having very little knowledge beyond the GUI. Nonetheless, I'm trying to educate myself in the use of gdb and bug reporting so that I waste less of my time and yours. FYI: I enabled dri3 and used sna. Didn't help. And Intel microcode is installed. "it's not a bug at all - kwin noticed that your gpu too a nap and restarts the compositor." Or doesn't. How this isn't a bug is beyond me. So it's perfectly normal for KDE's compositor to crash? Okay. okay, I think I might have Googled the source of all these problems: http://www.spinics.net/lists/intel-gfx/msg82135.html I reverted to Kernel 4.2.0-25-generic x86_64 (64 bit) and, so far, TTY is stable, there are no missing glyphs/letters after switching to TTY or suspend/resume, and no graphic resets/plasmashell crashes. Stay tuned. |
This error is a regular occurrence after upgrading to 16.04. I've been using this system since 14.10 and have never before run into this bug. Kernel: 4.4.0-18-generic x86_64 (64 bit) Desktop: KDE Plasma 5.5.5 Distro: Ubuntu 16.04 xenial Machine: System: LENOVO (portable) product: 20428 v: Lenovo Yoga 2 11 Mobo: LENOVO model: VIUU4 v: 31900058WIN Bios: LENOVO v: AACN21WW date: 01/30/2015 CPU: Dual core Intel Core i3-4012Y (-HT-MCP-) cache: 3072 KB clock speeds: max: 1500 MHz 1: 1500 MHz 2: 1500 MHz 3: 1155 MHz 4: 1498 MHz Graphics: Card: Intel Device 0a1e Display Server: X.Org 1.18.3 drivers: intel (unloaded: fbdev,vesa) Resolution: 1366x768@60.00hz GLX Renderer: Mesa DRI Intel Haswell GLX Version: 3.0 Mesa 11.2.0 Reproducible: Sometimes Steps to Reproduce: 1.) It can seemingly occur at random---though possibly related to page-rendering in browsers. 2.) It also frequently occurs (though not always) when recovering from suspend mode. 3.) If the system has been unplugged during suspend, restarting can not only cause the above error but also lead to missing text as described in Bug: https://bugs.kde.org/show_bug.cgi?id=361211 But the two Bugs may not be directly related? That is, restarting Desktop Effects only leads to Missing Text 1 out of every 10 restarts. I've marked the Bug as Grave because of its unpredictability and because, at its worst (with missing text) the system is "basically unusable now" and needs to be rebooted.