Summary: | some dead processes stay in the list | ||
---|---|---|---|
Product: | [Unmaintained] ksysguard | Reporter: | Waleed Hamra <kdebugs> |
Component: | Process Controller - krunner part | Assignee: | KSysGuard Developers <ksysguard-bugs> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | alanbortu, amantia, captaincrutches, cfeck, commander.alchemy, dab1818, deabrufree, devguy.ca, ht990332, kdebugs, m01brv, markhkamp, martin.sandsmark, martin.tlustos, null, opensource, q.kde, rajindery, timshel, travisgevans |
Priority: | NOR | ||
Version: | 4.8 | ||
Target Milestone: | --- | ||
Platform: | Ubuntu | ||
OS: | Linux | ||
Latest Commit: | http://commits.kde.org/libksysguard/c2fba0b64a0a2541a40a46a9f17fdbda1ea6a498 | Version Fixed In: | |
Sentry Crash Report: | |||
Attachments: |
Dead processes amarok and java (with detailed process information)
another screenshot for KDE 4.11.0 / Kubuntu 13.04 on left new instance of ksysguard, on right old one |
Description
Waleed Hamra
2010-12-25 22:58:18 UTC
*** Bug 282974 has been marked as a duplicate of this bug. *** *** Bug 301085 has been marked as a duplicate of this bug. *** I've been seeing this for a long time in Fedora, and it's still happening in Fedora 17 with KDE 4.8.5 I can confirm this has been happening for me on kde 4.6 to kde 4.8 it seems to persist endlessly across versions. I don't understand how this is not confirmed when I see it happening to every kde user I meet. I suffer from the same issue. The only way to get rid of the ghost processes in the table is to restart ksysguard. It is already rather aged bug. I'm getting bit by this in KDE 4.8.4. I have several processes that don't even exist (they're not even in the /proc tree) persistently showing up as “disk sleep” in the CPU column. They refuse to go away. Attempting operations on them (send signal, show application window, etc.) just does nothing without an error message. Anyone happen to see this in the latest version of KDE which I don't currently have installed? I'm seeing this bug on my system running KDE 4.9.1 on Chakra Linux. The following script to generate tons of processes hasn't resulted in any of them staying in the list after terminating, so the bug might depend on processes doing something specific. #!/bin/bash for ((i=0;i!=25000;i+=1)); do # enough to fill the process table for a short while. sleep 2 & # Give time for the process to appear in the list. done exit @Waleed Hamra (or anyone that can confirm this bug, please attach a screenshot in here. The initial added image on imageshack is dead.. Created attachment 80767 [details]
Dead processes amarok and java (with detailed process information)
I can also confirm this bug for KDE 4.10.3 / Kubuntu 13.04. In the attachment there are two dead processes, namely amarok and java. The detailed process information for java shows that it is actually not running.
Created attachment 81768 [details]
another screenshot for KDE 4.11.0 / Kubuntu 13.04
Comment on attachment 81768 [details]
another screenshot for KDE 4.11.0 / Kubuntu 13.04
confirm on KDE 4.11.0 / Kubuntu 13.04
dead processes are: thunderbird, kate, kwrite, dolphin, muon-updater
Created attachment 83452 [details]
on left new instance of ksysguard, on right old one
I can confirm on Sabayon Linux with KDE 4.11.2.
Longer ksysguard runs, more processes are left on list - in screenshot below comparison between ksysguard running for ~2 days and freshly launched one.
Every process above systemd process doesn't exist.
*** Bug 332951 has been marked as a duplicate of this bug. *** *** This bug has been confirmed by popular vote. *** I got the same issue with KDE 4.13 on ArchLinux. Still in kde 4.13.1 I still have this issue in KDE 4.14.1 on openSUSE Factory *** Bug 340910 has been marked as a duplicate of this bug. *** confirm on KDE 4.14.3 (ksysguard 4.11.14, gentoo linux) Is it possible that one of the developers may have some time to take a look at this, please? This is a 5+ year old bug. Thank you very much for your time and for KDE. To debug this, we need a way to reproduce it consistently. I have seen dead processes in the list, but I have no idea how I can reliably make them stay there to compare debug output and actual behavior. (In reply to Christoph Feck from comment #21) > To debug this, we need a way to reproduce it consistently. Silly question, perhaps, but I have never seen ps give the wrong results, so would it not be possible to add some debug code that simply captured the output of ps, compared it to ksysguard's internal view and triggered a breakpoint if they differed? I think this is the code in question http://quickgit.kde.org/?p=ksysguard.git&a=blob&f=ksysguardd%2FLinux%2FProcessList.c *** Bug 355182 has been marked as a duplicate of this bug. *** Some observations: 1) systemmonitor does not have this bug. 2) it only seems to happen if ksysguard is in the background or if you have switched to another tab. *** Bug 359047 has been marked as a duplicate of this bug. *** this is still valid as of ksysguard 5.6.4-1 running arch linux packages closing and reopening ksysguard clears the list Launching two instances of okular, and then closing them, reproduces it for me. Launching ksysguardd from the command line and running "ps" doesn't seem to exhibit the same bug, so probably in the UI somehow. The problem is somewhere in Processes::processesUpdated() in libksysguard, I think. It seems to happen if a process appears and then disappears between calls to it. It's a race condition between ProcessModelPrivate::updateWindowInfo() and Processes::processesUpdated(). Git commit c2fba0b64a0a2541a40a46a9f17fdbda1ea6a498 by Martin T. H. Sandsmark. Committed on 16/07/2016 at 15:10. Pushed by sandsmark into branch 'master'. Fix race condition when new applications open When opening and closing an application quickly it sometimes got "stuck" in ksysguard because it was added by ProcessModel while Processes::processesUpdated() was running, so Processes never processed it. REVIEW: 128431 M +0 -1 processui/ProcessModel.cpp http://commits.kde.org/libksysguard/c2fba0b64a0a2541a40a46a9f17fdbda1ea6a498 *** Bug 370176 has been marked as a duplicate of this bug. *** |