https://bugs.kde.org/show_bug.cgi?id=438097
vmpn changed:
What|Removed |Added
CC||volc...@gmail.com
--
You are receiving this mail
https://bugs.kde.org/show_bug.cgi?id=438097
Nate Graham changed:
What|Removed |Added
Version Fixed In||5.23
--
You are receiving this mail because:
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #42 from Lewis Lakerink ---
*** Bug 438381 has been marked as a duplicate of this bug. ***
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=438097
Vlad Zahorodnii changed:
What|Removed |Added
Latest Commit|https://invent.kde.org/plas |https://invent.kde.org/plas
https://bugs.kde.org/show_bug.cgi?id=438097
Vlad Zahorodnii changed:
What|Removed |Added
Resolution|--- |FIXED
Latest Commit|
https://bugs.kde.org/show_bug.cgi?id=438097
Bug Janitor Service changed:
What|Removed |Added
Status|REPORTED|ASSIGNED
Ever confirmed|0
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #38 from Lewis Lakerink ---
+1, your patch prevents the uncontrolled growth of fds for me Vlad!
Thanks!
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #37 from Ash Blake ---
(In reply to Vlad Zahorodnii from comment #36)
> Created attachment 141876 [details]
> Potential solution (untested)
>
> If you apply the attached patch, does the issue go away?
Yes, I just tested it with 1000
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #36 from Vlad Zahorodnii ---
Created attachment 141876
--> https://bugs.kde.org/attachment.cgi?id=141876=edit
Potential solution (untested)
If you apply the attached patch, does the issue go away?
--
You are receiving this mail
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #35 from Vlad Zahorodnii ---
Aha, I know what it is... It's a bug in our handling of inert resources.
If a plasma window is deleted, our wrapper code won't call the handler and
therefore the file descriptor will be leaked.
--
You are
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #34 from Ash Blake ---
(In reply to Vlad Zahorodnii from comment #33)
> Thanks for the great analysis, Ash! This is definitely very strange. At
> quick glance, I don't see how file descriptors can be leaked in kwin.
Well, it doesn't seem
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #33 from Vlad Zahorodnii ---
Thanks for the great analysis, Ash! This is definitely very strange. At quick
glance, I don't see how file descriptors can be leaked in kwin.
--
You are receiving this mail because:
You are watching all bug
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #32 from Ash Blake ---
Created attachment 141875
--> https://bugs.kde.org/attachment.cgi?id=141875=edit
A Python script that parses strace output for FD information
I wrote a script to parse strace output and abbreviate it, displaying
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #31 from Ash Blake ---
(In reply to Ash Blake from comment #30)
> break PlasmaWindow::Private::iconChangedCallback
> commands
> silent
> printf "hello\n"
> end
Oops, forgot a continue there.
break
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #30 from Ash Blake ---
> The short burst length of 2 was motivated by observing the popup windows
> flicker without any content sometimes, before the actual popup appeared.
Post scriptum: in fact, when GDB is attached to plasmashell in
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #29 from Ash Blake ---
Created attachment 141874
--> https://bugs.kde.org/attachment.cgi?id=141874=edit
A C program that reproduces the bug
I figured out which behaviour of the Jetbrains IDEs causes the bug to occur and
I wrote a program
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #28 from Lewis Lakerink ---
Ok, so setting the hard/soft fd limit to a high value appears to be an
effective solution for me for now until this issue is resolved.
Using Ash's xdotool script modified for pycharm, I got the kwin fd count up
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #27 from Ash Blake ---
The file descriptor flood comes from PlasmaWindow::Private::iconChangedCallback
(kwayland/src/client/plasmawindowmanagement.cpp:655, master branch).
This can be easily verified by catching pipe2 in the plasmashell
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #26 from Lewis Lakerink ---
Firstly can I just say Ash, thanks for your in depth analysis! It's been super
educational for me to see your process and how you use GDB to debug the issue.
Nice find with the fds, I can confirm that that is the
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #25 from Ash Blake ---
This is pretty weird. I just tried running IDEA in KWin only, started from a
TTY.
The file descriptor amount does not skyrocket, and everything is stable.
It also works fine in nested KWin inside a Plasma session.
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #24 from Ash Blake ---
Created attachment 141808
--> https://bugs.kde.org/attachment.cgi?id=141808=edit
Timelapse of lsof count, after increasing the open fd limit to 8192
This shows the test running for much longer (5 minutes), after
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #23 from Ash Blake ---
Created attachment 141807
--> https://bugs.kde.org/attachment.cgi?id=141807=edit
Outputs of lsof -p $KWIN_PID taken every second, from the test start to KWin
crash
Additional logs for #22
--
You are receiving
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #22 from Ash Blake ---
Created attachment 141806
--> https://bugs.kde.org/attachment.cgi?id=141806=edit
Screencast demonstrating the rising amount of open files by KWin
Regarding the previous error about too many open files:
This
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #21 from Ash Blake ---
This time KWin didn't crash completely, but entered a weird state.
Plasmashell stopped working, but most of the applications were still running,
including IDEA. I stopped the test, because if it continued it would
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #20 from Ash Blake ---
This short script can be used to trigger a crash:
#!/bin/bash
win_id=$(sort <(xdotool search --name "Content window") <(xdotool search
--class "jetbrains-idea") | uniq -d)
sleep 10
while :
do
xdotool key --window
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #19 from Ash Blake ---
Created attachment 141801
--> https://bugs.kde.org/attachment.cgi?id=141801=edit
Code completion popup window IDs in the debug console
Update: It appears that the crashes are highly correlated with the amount of
https://bugs.kde.org/show_bug.cgi?id=438097
Ash Blake changed:
What|Removed |Added
Version|5.22.1 |5.22.90
--
You are receiving this mail because:
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #18 from Ash Blake ---
I didn't get a crash so far, but I revisited the previous coredump, as it seems
weird that the EffectWindow was at least a bit intact this time.
Disassembly of the current instruction and the one before it:
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #17 from Ash Blake ---
This time, however, it seems EffectWindow was not completely destroyed.
Unlike in the last crash, the data inside EffectWindow is not complete garbage.
For instance, the addresses in toplevel and
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #16 from Ash Blake ---
Okay, it crashed again. The backtrace looks the same as the last one.
The crash is again in kwin4_effect_fade:
(gdb) p this
$3 = (KWin::ScriptedEffect * const) 0x55f805fa1440
(gdb) set $data_start =
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #15 from Ash Blake ---
Additional information about that EffectWindow, as I forgot it's actually an
instance of EffectWindowImpl and didn't dump members of the latter.
Both the toplevel and sw pointers lead to inaccessible memory
https://bugs.kde.org/show_bug.cgi?id=438097
Ash Blake changed:
What|Removed |Added
CC||telepath...@tutanota.com
--- Comment #14 from Ash
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #13 from Lewis Lakerink ---
(In reply to Aleix Pol from comment #10)
> @Lewis, you mentioned that you were using 5.21.90, maybe you could try
> 5.22.90?
>
> Could you mention which effects you have enabled?
Exactly the same symptoms for
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #12 from Luis Cáceres ---
Created attachment 141776
--> https://bugs.kde.org/attachment.cgi?id=141776=edit
Backtrace on 5.22.4
I'm also attaching a new backtrace I got just today. Hopefully it helps too.
--
You are receiving this mail
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #11 from Luis Cáceres ---
It crashes seemingly randomly. Sometimes it takes just a few minutes, sometimes
it takes several hours.
For completion, I've been having the issue on all plasma versions 5.22.x. My
desktop effects on Fedora 34
https://bugs.kde.org/show_bug.cgi?id=438097
Aleix Pol changed:
What|Removed |Added
CC||aleix...@kde.org
--- Comment #10 from Aleix Pol
https://bugs.kde.org/show_bug.cgi?id=438097
Aleix Pol changed:
What|Removed |Added
CC||ha...@dayfamilyweb.com
--- Comment #9 from Aleix
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #8 from Lewis Lakerink ---
Possibly related issue: https://bugs.kde.org/show_bug.cgi?id=440094
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=438097
Luis Cáceres changed:
What|Removed |Added
CC||lacacere...@gmail.com
--- Comment #7 from Luis
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #6 from Lewis Lakerink ---
Ok, the issue with core dumps was that my HDD was almost full.
I collected another trace today with the same problem, this time kwin was built
from the 5.22 branch - commit:
https://bugs.kde.org/show_bug.cgi?id=438097
Lewis Lakerink changed:
What|Removed |Added
Version|git master |5.22.1
--
You are receiving this mail
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #5 from Lewis Lakerink ---
Annoyingly, I'm still seeing weird behaviour after time with xwayland and
sometimes kwin crashing and restarting the session - however, I no longer get
coredumps for kwin, and I've no idea why that is. The only
https://bugs.kde.org/show_bug.cgi?id=438097
Nate Graham changed:
What|Removed |Added
CC||n...@kde.org
--
You are receiving this mail
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #4 from Lewis Lakerink ---
What can I do to help provide more details/debug info that might help with
this?
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=438097
--- Comment #3 from Vlad Zahorodnii ---
As is, I don't see how kwin can crash in AnimationEffect::postPaintScreen().
EffectWindow should always be valid. Unless xwayland crashes for some reason
and no EffectsHandler::windowDeleted signal is emitted.
https://bugs.kde.org/show_bug.cgi?id=438097
Lewis Lakerink changed:
What|Removed |Added
Summary|[AnimationEffect] |[AnimationEffect]
46 matches
Mail list logo
