Created attachment 139805 [details] screenshots SUMMARY Most default widgets (default bottom bar + menu) render with dirty, empty-boxed "shadows", giving Plasma a sketchy look. STEPS TO REPRODUCE 1. Create a new user and open a Plasma Wayland session from it. OBSERVED RESULT See the attached files. The first is from my current session. The second is from a new user's session. EXPECTED RESULT Clean shadows or no shadow, not the blurry, sketchy middle-ground shown here. SOFTWARE/OS VERSIONS Operating System: openSUSE Tumbleweed 20210629 KDE Plasma Version: 5.22.2 KDE Frameworks Version: 5.83.0 Qt Version: 5.15.2 Kernel Version: 5.12.13-1-default (64-bit) Graphics Platform: Wayland Processors: 8 × Intel® Core™ i7-6700HQ CPU @ 2.60GHz Memory: 15.5 Gio of RAM Graphics Processor: Mesa DRI Intel® HD Graphics 530
Can reproduce this one. Changing Plasma theme seems to fix it temporarily but the artifacts come back. Not sure what exactly triggers it.
X or wayland?
Wayland for me. KDE Plasma Version: 5.22.2 KDE Frameworks Version: 5.83.0 Qt Version: 5.15.2
For me too on wayland
The shadows seem to be mirrored. I sometimes observe the same behavior as well. It does not happen all the time, but it seems to be consistent within one session.
Can you confirm if this is "resolved" by removing ~/.cache/*.kcache and restarting plasmashell
possibly duplicate of bug 429057
> possibly duplicate of bug 429057 Might be related, but the glitch there looks pretty different to me. Bug 438696 is probably a duplicate of the present issue instead.
> Can you confirm if this is "resolved" by removing ~/.cache/*.kcache and restarting plasmashell In fact, just restarting plasmashell seemed to be enough.
Good catch David Edmundson, removing ~/.cache/*.kcache and restarting plasmashell fixed it for me. Le mar. 6 juil. 2021 à 18:09, Alois Wohlschlager <bugzilla_noreply@kde.org> a écrit : > > https://bugs.kde.org/show_bug.cgi?id=439408 > > --- Comment #9 from Alois Wohlschlager <alois1@gmx-topmail.de> --- > > Can you confirm if this is "resolved" by removing ~/.cache/*.kcache and restarting plasmashell > > In fact, just restarting plasmashell seemed to be enough. > > -- > You are receiving this mail because: > You reported the bug.
(In reply to Ad from comment #10) > Good catch David Edmundson, removing ~/.cache/*.kcache and restarting > plasmashell fixed it for me. > > Le mar. 6 juil. 2021 à 18:09, Alois Wohlschlager > <bugzilla_noreply@kde.org> a écrit : > > > > https://bugs.kde.org/show_bug.cgi?id=439408 > > > > --- Comment #9 from Alois Wohlschlager <alois1@gmx-topmail.de> --- > > > Can you confirm if this is "resolved" by removing ~/.cache/*.kcache and restarting plasmashell > > > > In fact, just restarting plasmashell seemed to be enough. > > > > -- > > You are receiving this mail because: > > You reported the bug. In fact the problem happens again on every new fresh sessions even after removing the cache. I would not even consider it a temporary fix.
I think I have a reliable reproducer now (I use a left panel with Kickoff at the top and Digital Clock at the bottom, to put the locations of the mirrored shadows into context). * When the first thing I do with a fresh plasmashell (i.e. after login or plasmashell restart) is opening Kickoff, subsequent popups will have mirrored shadows at the bottom and right sides. Kickoff itself only gets the mirrored shadow after another popup has appeared in the meantime. * If the first thing I do is opening the calendar, subsequent popus will have mirrored shadows at the top and right sides. Similar to above, the calendar itself only displays mirrored shadows if something else was popped up in the meantime. * If any popup other than the two mentioned before opens first (including the tooltip of Kickoff or Digital Clock), all shadows display in the correct orientation. * Panel and window shadows are unaffected in any case. The behavior is completely reproducible in my setup. Deleting the cache files does not seem to have any impact in any way.
*** Bug 439974 has been marked as a duplicate of this bug. ***
Raising to VHI since this gives the impression that the desktop is completely unstable and broken.
Possibly relevant: whenever I open an applet popup on Wayland, I see the following printed to the console: kf.plasma.quick: Couldn't create KWindowShadow for PlasmaQuick::Dialog_QML_230(0x32bdf60, name="popupWindow")
I don't think that's too relevant. If someone can reproduce, can they send me their ~/.cache/plasma*
Sent offline since the file was too big to attach to the bug report.
*** This bug has been marked as a duplicate of bug 438644 ***
Not the same as Bug 438644; that bug report is about the white/gray artifacts in the corners caused by the background contrast effect not being able to round its mask properly (IIRC); this is about a cache invalidation bug whereby the wrong shadow SVGs are sometimes used.
*** Bug 438644 has been marked as a duplicate of this bug. ***
Was able to reproduce twice a day. It cleans up itself and happens again with no visible reason..
It happens randomly and disappears again. I tried to reproduce it by taking the cache while it happens and restoring it after logout, no luck.
I can reproduce it somewhat reliably like this: 1. Be in an X11 session 2. Reboot 3. Log into a Wayland session 4. Change scale factor from 200% to to 175% and then back to 200% (working around Bug 433460) 5. Reboot Then the shadows will be broken until I restart plasmashell
Nothing is particularly obvious, so I propose the following action plan: - we make a qmlscene that shows shadow elements at their natural size - we make a plasma theme that has blocky base colours (red, green, blue) for the different dialog-background-shadow-top/left/right elements - we find the numerical cache IDs for each element so we can retrieve them from a cache relatively easily
Documenting things we know: This affects Dialog shadows mostly so: - the bug is not in FrameSVG - the bug is not in SVGItem - Within p-f only SVG directly - But no other SVG is affected? So that implies it's not SVG, or there's something special about this - It affects both X and wayland This /implies/ it's not at a KWindowSystem level, but we can't yet completely rule it out - When the client goes wrong it remains wrong until a client restart This /implies/ it's not kwin as a new window vs a new client should be mostly the same
Ids "shadow-top" : "1449101648" "shadow-topright" : "1336515456" "shadow-right" : "3033709554" "shadow-bottomright" : "3158977252" "shadow-bottom" : "2093078684" "shadow-bottomleft" : "578256148" "shadow-left" : "2934766096" "shadow-topleft" : "2066772017"
Yeah, the net effect is that the bottom and right edges use the wrong shadow images.
(In reply to David Edmundson from comment #26) > - When the client goes wrong it remains wrong until a client restart I'm not sure about this. I recall once the issue cleaned up itself while I was in the same session. As I see it right now I'll look when it restores.
*** Bug 429057 has been marked as a duplicate of this bug. ***
*** Bug 431761 has been marked as a duplicate of this bug. ***
*** Bug 438696 has been marked as a duplicate of this bug. ***
*** Bug 438723 has been marked as a duplicate of this bug. ***
Created attachment 141233 [details] colored shadows theme I've created a theme with colored shadow's elements. Thus the bug should be clearly noticeable even on dark background - assuming there should be elements with the same color.
Created attachment 141234 [details] Colored shadows, normal behaviour I'll add new theme screenshot with the problem when I face it next time (will try also Nate's approach reproducing it).
For others installing it that attachment will extract to ./local/share/plasma/desktopthemes from within your CWD
(In reply to Alois Wohlschlager from comment #12) > I think I have a reliable reproducer now (I use a left panel with Kickoff at > the top and Digital Clock at the bottom, to put the locations of the > mirrored shadows into context). > > * When the first thing I do with a fresh plasmashell (i.e. after login or > plasmashell restart) is opening Kickoff, subsequent popups will have > mirrored shadows at the bottom and right sides. Kickoff itself only gets the > mirrored shadow after another popup has appeared in the meantime. Oh, thank you - was able to reproduce with it too!
Created attachment 141243 [details] Broken shadows with colored theme It was reproduced with https://bugs.kde.org/show_bug.cgi?id=439408#c12
(In reply to Andrey from comment #37) > (In reply to Alois Wohlschlager from comment #12) > > I think I have a reliable reproducer now (I use a left panel with Kickoff at > > the top and Digital Clock at the bottom, to put the locations of the > > mirrored shadows into context). > > > > * When the first thing I do with a fresh plasmashell (i.e. after login or > > plasmashell restart) is opening Kickoff, subsequent popups will have > > mirrored shadows at the bottom and right sides. Kickoff itself only gets the > > mirrored shadow after another popup has appeared in the meantime. > > Oh, thank you - was able to reproduce with it too! I can also reproduce with those steps and that setup. It didn't occur to me that my vertical panel might be a part of it.
A possibly relevant merge request was started @ https://invent.kde.org/plasma/kwayland-integration/-/merge_requests/12
Git commit 11769ddffe25ca808228f7733aea5697ea41264d by Andrey Butirsky. Committed on 08/09/2021 at 09:21. Pushed by davidedmundson into branch 'master'. fix a dirty shadows ShmPool tries to resize itself when creating new images if the existing allocated memory is too small. As shadows currently use a single shared pool, this potentially happens multiple times. When an SHMPool is resized, the existing memory allocation is invalid and all existing buffers have to be re-uploaded. This patch introduces one pool per tile, which alleviates that problem. M +4 -8 src/windowsystem/waylandintegration.cpp M +1 -2 src/windowsystem/waylandintegration.h M +6 -3 src/windowsystem/windowshadow.cpp M +4 -0 src/windowsystem/windowshadow.h https://invent.kde.org/plasma/kwayland-integration/commit/11769ddffe25ca808228f7733aea5697ea41264d
*** Bug 444533 has been marked as a duplicate of this bug. ***