KDE Developer
|
Does compositor work for you with proprietary NVidia driver 358.16 or it is just my problem?
Can you have transparent konsole window? My setup: Distro: Fedora 23 64-bit plasmashell 5.5.4 Driver: NVidia 358.16 Card: 01:00.0 VGA compatible controller: NVIDIA Corporation GM107 [GeForce GTX 750 Ti] (rev a2) Linux kernel: 4.3.5-300.fc23.x86_64 Thanks
Daylight is coming...
Krita developer | http://lukast.mediablog.sk/log |
|
Yes, works.
What's the output of "glxinfo" (please post it in code tags!) and the contens of ~/.config/kwinrc ? (nb: that EGL will not work on the nvidia blob for now, only GLX does) |
KDE Developer
|
glxinfo
kwinrc
Daylight is coming...
Krita developer | http://lukast.mediablog.sk/log |
|
> Enabled=false
It's not enabled (on startup) but for the rest looks fine. SHIFT+Alt+F12 should toggle it on (for a functional test) and you can alter the initial state in "kcmshell5 kwincompositing". Afair, we disable this setting if kwin crashes too often™ in a row (but regardless of the actual reason) |
KDE Developer
|
I must say that there are crashes regularly with compositor enabled. That's why I disabled it at the start.
Can you have transparent konsole window? I cannot. When I drag window with compositor enabled, I see the translucency of window works: http://i.imgur.com/PMqGHPZ.png But the konsole window is black, even when I set in profile to be 21% transparent. https://imgur.com/c5INnZy
Daylight is coming...
Krita developer | http://lukast.mediablog.sk/log |
Manager
|
I get transparent Konsole, what happens of you try 99% transparency
historically, pre 358.16, where there issues? have you tried a diff set of drivers? ps - I am running 358.16 |
|
You got to restart konsole - it only checks for a compositor once on startup to decide whether it needs an ARGB drawable or not.
Reg. the kwin crashes: bug id? Backtrace? |
KDE Developer
|
Is it possible to force konsole to recheck for ARGB drawable?
It is still full black, but compositor is on. > Reg. the kwin crashes: bug id? Backtrace? I use Fedora 23 and it uses abrt tool which consumes all the backtraces and marks them as useless. I have to find solution so that I can provide the useful bug reports with NVidia proprietary drivers. Probably the crashes are in drivers..
Daylight is coming...
Krita developer | http://lukast.mediablog.sk/log |
KDE Developer
|
I tried nouveau, it was not supporting dual monitors I use.
Daylight is coming...
Krita developer | http://lukast.mediablog.sk/log |
|
re-checking isn't sufficient, the entire window would have to be re-created.
You might want to file a bug against konsole to support "--forcetransparency" or similar, but that's rather moot since you "only" need to ensure a compositor when launching the konsole process (and if there's none, the ARGB window makes little sense) "abrt tool" is a sigabrt? That means you ran into an assert (different state than expected) and the "useless" state comes from the lack of debug symbols => "sudo debuginfo-install kwin" should do (not 100% sure about the package name being "kwin") See https://techbase.kde.org/Development/Tu ... rts#Fedora on details. |
KDE Developer
|
Aha, is Konsole doing the check every KDE session start? I though it just checks once, and caches the result of the check somewhere in $KDEHOME dir.
This tool https://imgur.com/lGfk3JV https://github.com/abrt/abrt/wiki/ABRT-Project Here is example of the backtrace marked as "useless" by Abrt:
Daylight is coming...
Krita developer | http://lukast.mediablog.sk/log |
|
It checks for the compositor with every konsole process start (but if there's a running konsole process, calling konsole will just open a new window in that instance)
The backtrace is super-weird. The process is ksplashqml (not kwin, but that likely doesn't matter) Main is in thread #2 (??) and currently exitting, but some thread #1 aborts in apparently libpthread from some unknown entry and a malloc error. From the table below, that would either be the vdso lib (unlikely) or indeed one of the nvidia libraries. Those will never provide any debug symbols, but you can report bugs to nvidia about this (they can map the stack positions to actual code) Try to
(just a wild guess) |
Registered users: bartoloni, Bing [Bot], Evergrowing, Google [Bot], ourcraft