Bug 518264 - Discover fails to launch on KDE Linux.
Summary: Discover fails to launch on KDE Linux.
Status: RESOLVED FIXED
Alias: None
Product: Discover
Classification: Applications
Component: discover (other bugs)
Version First Reported In: 6.6.80
Platform: Other Linux
: NOR minor
Target Milestone: ---
Assignee: Plasma Bugs List
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2026-03-28 16:13 UTC by kovic
Modified: 2026-08-25 10:17 UTC (History)
4 users (show)

See Also:
Latest Commit:
Version Fixed/Implemented In:
Sentry Crash Report:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description kovic 2026-03-28 16:13:49 UTC
STEPS TO REPRODUCE
 Attempt launching Discover, either through the GUI or the CLI with `plasma-discover`

OBSERVED RESULT
Attempting to launch through the GUI simply results in a loading animation that lasts ~9 or so seconds, and then fails to launch
Attempting to launch through the CLI returns 
`QThreadStorage: entry 2 destroyed before end of thread 0x558419476260
QThreadStorage: entry 1 destroyed before end of thread 0x558419476260`


EXPECTED RESULT
Discover launches.

SOFTWARE/OS VERSIONS
Linux/KDE Plasma: KDE Linux, Linux 6.19.6-zen1-1-zen
KDE Plasma Version: 6.6.80 
KDE Frameworks Version: 6.25.0
Qt Version: 6.10.2 

ADDITIONAL INFORMATION
Restarting the system fixes the issue, killing Discover's process from System Monitor does aswell, although, when this issue was first replicateable around 2-3 months ago, Discover would create an unkillable process.
The issue however re-arises after some time.
Comment 1 Adam Khaled 2026-04-04 11:34:15 UTC
I experienced the same issue on Fedora 43 with plasma-discover 6.6.3. It occurs after long uptime and is caused by a stuck plasma-discover process from a previous session. Running pkill -9 plasma-discover before launching fixes it without a reboot. I also have a stack trace from coredumpctl showing the crash occurring in LazyIconResolver::resolve → KIconLoader::queryIcons → deduplicateIconsByName if that's useful for debugging
Stack trace of thread 3255:
#0  0x00007f0154c813cc __pthread_kill_implementation (libc.so.6 + 0x743cc)
#1  0x00007f0154c2715e raise (libc.so.6 + 0x1a15e)
#2  0x00007f0157e220d6 _ZN6KCrash19defaultCrashHandlerEi (libKF6Crash.so.6 + 0x50d6)
#3  0x00007f0154c27290 __restore_rt (libc.so.6 + 0x1a290)
#4  0x00007f015540379d _ZN9QtPrivate14compareStringsE11QStringViewS0_N2Qt15CaseSensitivityE (libQt6Core.so.6 + 0x20379d)
#5  0x00007f0155425738 _ZN9QtPrivate20QStringList_containsEPK5QListI7QStringE11QStringViewN2Qt15CaseSensitivityE (libQt6Core.so.6 + 0x225738)
#6  0x00007f01559650c5 _ZN12_GLOBAL__N_1L22deduplicateIconsByNameERK5QListI7QStringE.lto_priv.0 (libKF6IconThemes.so.6 + 0xa0c5)
#7  0x00007f0155970b3c _ZNK11KIconLoader10queryIconsEv (libKF6IconThemes.so.6 + 0x15b3c)
#8  0x00007f015809ebd0 _ZN14AppStreamUtils18kIconLoaderHasIconERK7QString (libDiscoverCommon.so + 0x80bd0)
#9  0x00007f00f573273f _ZN21AppPackageKitResource11resolveIconEv (packagekit-backend.so + 0x3d73f)
#10 0x00007f01580840d1 _ZN16LazyIconResolver7resolveEv (libDiscoverCommon.so + 0x660d1)
#11 0x00007f015808411d _ZN16LazyIconResolver11customEventEP6QEvent (libDiscoverCommon.so + 0x6611d)
Comment 2 TraceyC 2026-05-18 21:36:19 UTC
(In reply to kovic from comment #0)
> STEPS TO REPRODUCE
>  Attempt launching Discover, either through the GUI or the CLI with
> `plasma-discover`

Do you happen to know if this happens when updates are being installed? That's the reported trigger for bug 519683

Also, is there a crash for Discover in Crashed Process Viewer? If so, we'll need a backtrace.

In bug 513569, the problem was linked to running with specific backends, in a specific order. At the time, it wasn't able to be reproduced on KDE Linux.
Can you try these and see what happens?

This may fail to bring up Discover
  killall -9 plasma-discover (flatpak first)
  plasma-discover --backends flatpak,fwupd

This may bring up Discover successfully (fwupd first)
  killall -9 plasma-discover
  plasma-discover --backends fwupd,flatpak

Let us know your results. Thanks!
Comment 3 kovic 2026-05-20 08:26:06 UTC
(In reply to TraceyC from comment #2)

Hello, fortunately I've not been able to replicate this bug for a while (around a month, now)

It seems the bug that I was experiencing is more or less identical to bug 519683, with the same trigger

No crash msg from what I can recall.

Discover launches normally without any problems whether I use the argument 'flatpak,fwupd' or 'fwupd,flatpak'
Comment 4 TraceyC 2026-05-20 15:38:53 UTC
Thanks for letting us know. I'm glad to hear things are working for you now!
Comment 5 kovic 2026-08-03 06:58:53 UTC
Issue has resurfaced, as of KDE Linux 2026-07-28

Running pkill -9 plasma-discover before launching does in fact fix it without rebooting.
Comment 6 Nate Graham 2026-08-03 14:16:30 UTC
If that helped, it means the last-running instance of Discover got stuck and didn't fully quit.
Comment 7 Nate Graham 2026-08-12 17:39:23 UTC
Which also means https://invent.kde.org/plasma/discover/-/merge_requests/1355 will likely fix it.
Comment 8 Bug Janitor Service 2026-08-14 16:54:46 UTC
A possibly relevant merge request was started @ https://invent.kde.org/plasma/discover/-/merge_requests/1393
Comment 9 Oliver Beard 2026-08-15 14:10:26 UTC
Git commit a8b5538958594d90406aa0294d3934bcd06e7607 by Oliver Beard, on behalf of Aleix Pol.
Committed on 15/08/2026 at 14:03.
Pushed by olib into branch 'master'.

DiscoverObject: Address quitting with concurrent KJobs

We have a mechanism that disables random eventloops (KJobs) from
quitting Discover outside of the user's interaction (see
QCoreApplication::setQuitLockEnabled in discover/main.cpp).

It so happens that we would at times have KJobs running also _after_
Discover's window closes. See how on the reported bugs they often
describe it happening early in the startup process.

This patch restores the Q*Application::quitLockEnabled property so that
once Discover is with its task and we let go and QCoreApplication can
let these jobs finish and eventually quit.

It changes the DummyBackend to issue a KJob at the start to be able to
reproduce this problem and eventually its fix.
Related: bug 497419, bug 522245, bug 522815

M  +2    -0    discover/DiscoverObject.cpp
M  +22   -0    libdiscover/backends/DummyBackend/DummyBackend.cpp

https://invent.kde.org/plasma/discover/-/commit/a8b5538958594d90406aa0294d3934bcd06e7607
Comment 10 Oliver Beard 2026-08-15 14:31:51 UTC
Git commit 0d0a1d35c715838f585590d54a1a9d5b14cbd151 by Oliver Beard.
Committed on 15/08/2026 at 14:11.
Pushed by olib into branch 'Plasma/6.7'.

DiscoverObject: Address quitting with concurrent KJobs

We have a mechanism that disables random eventloops (KJobs) from
quitting Discover outside of the user's interaction (see
QCoreApplication::setQuitLockEnabled in discover/main.cpp).

It so happens that we would at times have KJobs running also _after_
Discover's window closes. See how on the reported bugs they often
describe it happening early in the startup process.

This patch restores the Q*Application::quitLockEnabled property so that
once Discover is with its task and we let go and QCoreApplication can
let these jobs finish and eventually quit.

It changes the DummyBackend to issue a KJob at the start to be able to
reproduce this problem and eventually its fix.
Related: bug 497419, bug 522245, bug 522815


(cherry picked from commit a8b5538958594d90406aa0294d3934bcd06e7607)

Co-authored-by: Aleix Pol <aleixpol@kde.org>

M  +2    -0    discover/DiscoverObject.cpp
M  +22   -0    libdiscover/backends/DummyBackend/DummyBackend.cpp

https://invent.kde.org/plasma/discover/-/commit/0d0a1d35c715838f585590d54a1a9d5b14cbd151
Comment 11 Oliver Beard 2026-08-25 10:17:39 UTC
Git commit 6bb252afbfef7bc413824ee153285e498b669801 by Oliver Beard.
Committed on 15/08/2026 at 14:10.
Pushed by olib into branch 'Plasma/6.6'.

DiscoverObject: Address quitting with concurrent KJobs

We have a mechanism that disables random eventloops (KJobs) from
quitting Discover outside of the user's interaction (see
QCoreApplication::setQuitLockEnabled in discover/main.cpp).

It so happens that we would at times have KJobs running also _after_
Discover's window closes. See how on the reported bugs they often
describe it happening early in the startup process.

This patch restores the Q*Application::quitLockEnabled property so that
once Discover is with its task and we let go and QCoreApplication can
let these jobs finish and eventually quit.

It changes the DummyBackend to issue a KJob at the start to be able to
reproduce this problem and eventually its fix.
Related: bug 497419, bug 522245, bug 522815


(cherry picked from commit a8b5538958594d90406aa0294d3934bcd06e7607)

Co-authored-by: Aleix Pol <aleixpol@kde.org>

M  +2    -0    discover/DiscoverObject.cpp
M  +22   -0    libdiscover/backends/DummyBackend/DummyBackend.cpp

https://invent.kde.org/plasma/discover/-/commit/6bb252afbfef7bc413824ee153285e498b669801