Bug 1312675

Summary:	Please disable fatal warnings in dbus user session, dbus must guess about Qt5 applications
Product:	[Fedora] Fedora	Reporter:	Raphael Groner <projects.rg>
Component:	qt5-qtbase	Assignee:	Than Ngo <than>
Status:	CLOSED RAWHIDE	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	medium	Docs Contact:
Priority:	unspecified
Version:	rawhide	CC:	amigadave, awilliam, jgrulich, jreznik, kparal, lpoetter, projects.rg, rdieter, rnovacek, than, walters
Target Milestone:	---
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:	RejectedFreezeException
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2016-03-08 16:24:57 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Raphael Groner 2016-02-28 17:39:53 UTC

Description of problem:
There are weird errors (so called "fatal warnings") in Qt5 applications logs. DBus user session complains about already closed connection and aborts (condition connection == NULL). That was found while fixing failures from recent F24 mass rebuild.

Version-Release number of selected component (if applicable):
rawhide

How reproducible:
yes

Steps to Reproduce:
1. build any Qt5 application, e.g. trojita, python-qutepart or enki
2. run some tests
3.

Actual results:
error entry in log, build fails

Expected results:
build success

Additional info:
Please disable the aborts completely or discuss with upstream about the sense for this aborts in meanwhile an application is running.

Workaround:
export DBUS_FATAL_WARNINGS=0
That line is needed in every spec file of a misbehaving(?) package. We never saw such a thing with Qt4.

https://dbus.freedesktop.org/doc/api/html/dbus-internals_8c_source.html

Comment 1 Raphael Groner 2016-02-28 17:40:49 UTC

DBus user session just tries to quit too early while Qt5 is still asking if alive.

Comment 2 Raphael Groner 2016-02-28 17:50:32 UTC

Google is also not sure about that feature(?).
https://codereview.chromium.org/2238006

Comment 3 Raphael Groner 2016-02-28 17:53:39 UTC

Temporarily moving to component qt5-qtbase to get maintainers into CC.

Comment 4 Rex Dieter 2016-02-28 18:38:11 UTC

For posterity, can you post a specific example of the specific error output

Comment 5 Raphael Groner 2016-02-28 20:54:59 UTC

It's about those two logger functions:

_dbus_warn

https://dbus.freedesktop.org/doc/api/html/group__DBusInternalsUtils.html#gabea2c16d6d3fe7c4eb5b9496fc877f84

_dbus_warn_check_failed

https://dbus.freedesktop.org/doc/api/html/group__DBusInternalsUtils.html#ga63f2f8a068454b781f214ba596e313b4


The source is full of call to those helpers. For instance, take a look into:

https://dbus.freedesktop.org/doc/api/html/dbus-connection_8c_source.html

Comment 6 Fedora Blocker Bugs Application 2016-03-02 19:59:38 UTC

Proposed as a Freeze Exception for 24-alpha by Fedora user raphgro using the blocker tracking app because:

 This bug prevents UAT for Qt5 applications and propably cli tools, too. Indeed, it may confuse unexperienced or new packagers why their packages do not work. The mentioned dbus assertion is caused indirectly and has no obvious reasoning in the build logs for e.g. tests execution failure.

Comment 7 Raphael Groner 2016-03-02 20:00:08 UTC

friendly ping, any news here?

Comment 8 Colin Walters 2016-03-02 20:12:45 UTC

This is qt using the API incorrectly - please try to fix it there first.  But if for some reason this is too hard, we *can* back out the warning change.  It just should be a last resort.

Comment 9 Kamil Páral 2016-03-07 18:36:55 UTC

Discussed at today's blocker review meeting [1]. It was rejected as a freeze exception - there is not yet a clear indication that this causes practical issues in a context that would be affected by an FE (e.g. use of the KDE live image). it can be reproposed if such an impact is demonstrated

[1] https://meetbot-raw.fedoraproject.org/fedora-blocker-review/2016-03-07/

Comment 10 Rex Dieter 2016-03-07 19:17:47 UTC

I was more interested in the details outlined in the build.log's of affected packages.

I *suspect* this may be related to:
https://bugreports.qt.io/browse/QTBUG-51299

and I'll pull in the fix for that.

Comment 11 Rex Dieter 2016-03-08 16:24:57 UTC

Confirmed this fixes it for me, in local build tests and koji trojita scratch build,

https://koji.fedoraproject.org/koji/taskinfo?taskID=13275639

%changelog
* Mon Mar 07 2016 Rex Dieter <rdieter> 5.6.0-0.38.rc
- backport "crash on start if system bus is not available" (QTBUG-51299)

Comment 12 Adam Williamson 2016-03-08 16:32:36 UTC

for the record, the same build made it to F24 before freeze/bodhi enablement.