Some apps upon being ran, do not update their content. I noticed such behavior in google-chrome and Emacs. In Emacs reproducible always. I thought it is a problem in OpenGL rendering KWin uses, and for testing purposes wanted to enable another rendered, but docs saying that there's no way to choose it now. When the problem happens, ~/.xsession-errors do not produce any output. Xorg.0.log here: https://pastebin.mozilla.org/8884980 (sorry, doesn't see a button to attach). Reproducible: Always Steps to Reproduce: 1. Run Emacs, and try any movement, the content would stand still. 2. Switch to another desktop, then back: the content of Emacs would change, but it still wouldn't respond to new actions until a new switch to another desktop and back.
> enable another rendered, but docs saying that there's no way to choose it now. Hummwhat? Please provide the output of "qdbus org.kde.KWin /KWin supportInformation" and ensure that emacs doesn't expose this behavior with suspended compositor (SHIFT+Alt+F12 toggles) Afaik you cannot attach files on reporting but only afterwards. No idea why.
Created attachment 100176 [details] qdbus org.kde.KWin /KWin supportInformation Attached the qdbus output. Wow, indeed, Alt+Shift+F12 fixes the problem. But when I enable compositioning back, the problem appears again. Why is this? > Hummwhat? s/rendered/renderer, sorry, typo. I mean, I wanted to test with another renderer (I remember there was a choice between OpenGL versions, and XRender), but didn't find the setting, and docs https://userbase.kde.org/Desktop_Effects_Performance#OpenGL_mode says it was removed. Though it is probably wrong part of documentation, I ought to look at the "Compositing type" paragraph. Anyway, I didn't find the setting either way.
run "kcmshell5 kwincompositing" it wasn't removed but the settings split so users won't find them easily (unfortunately that's not a joke)
> run "kcmshell5 kwincompositing" Okay, I ran, and tried XRender, OpenGL-2.0, OpenGL-3.0 (both EGL and GLX). Neither fixed the problem. I don't know if that help, but the problem is neither present on Enlightement, nor on Gnome, not on Awesome+Compton. (yeah, I know, it's too many DEs, it's just, as everyone goes to Wayland, I'm considering to use a full-fledged DE with tiling. kwin has more or less usable tiling btw; "more or less" mostly because of an old kwin bug; but still, non-updating content in windows is a problem)
if it does even happen with xrender chances are good the client simply stops causing damage events (for whatever reason) you probably tried all backends on the same instance? can you cause it on xrender after a logout and fresh emacs process?
Yep, I set Xrender, logged out/in, ran Emacs, and the problem still persist.
I'm not sure if that relevant, but Qt apps on my PC are crashing very often. I remember, I've read blog posts of Martin Graesslin about some faulty drivers which causing different problems, so I'm wondering: could that be it?
Random crashes in "foo" are not necessarily related to "bar" - this is about emacs (apparently under kwin, maybe only even plasma) and (likely) damage events. I don't think it's a driver issue for *all* backends are affected. - Does anything change when you resize the window (does it at least update then)? - Does it work when running compton on KDE? (it's not a WM, so suspending the KWin compositor and starting compton should just work "fine" - minus some integration between WM and compositor, of course)
Wow! I found what were causing it! I'm using tiling add-on to kwin https://github.com/faho/kwin-tiling/ so, to fulfill the resize test I ought to disable it. And… magically, it started working! In fact, the problem appears only if Emacs was launched for tiling extension was enabled, and disabling it fixes the problem. And if I enable it back later (i.e. when Emacs already launched), Emacs still works fine.
The tiling probably disobeys emacs size hints and emacs has errrr.... a "record" of being very errr.... "special"? "picky"? on this (it starts a war on its geometry) Quick tiling emacs might cause this as well juas as disobeying the geometry hints via "kcmshell5 kwinrules" - this very behavior is new to me, though (*sigh*)
Turns out, the bug is a duplicate of https://bugs.kde.org/show_bug.cgi?id=341195 Can we mark it as such?
As per comment #11 marked as dup *** This bug has been marked as a duplicate of bug 341195 ***