Bug 1356677
Summary: | Liveusb creator can't start (crash) using nvidia driver/libraries | ||||||||
---|---|---|---|---|---|---|---|---|---|
Product: | [Fedora] Fedora | Reporter: | jeremy9856 | ||||||
Component: | qt5-qtbase | Assignee: | Than Ngo <than> | ||||||
Status: | CLOSED WONTFIX | QA Contact: | Fedora Extras Quality Assurance <extras-qa> | ||||||
Severity: | high | Docs Contact: | |||||||
Priority: | unspecified | ||||||||
Version: | 25 | CC: | awilliam, ed.greshko, helio, jgrulich, jreznik, leigh123linux, lewk, mbriza, ovasik, rdieter, than | ||||||
Target Milestone: | --- | ||||||||
Target Release: | --- | ||||||||
Hardware: | All | ||||||||
OS: | Linux | ||||||||
Whiteboard: | |||||||||
Fixed In Version: | Doc Type: | If docs needed, set a value | |||||||
Doc Text: | Story Points: | --- | |||||||
Clone Of: | Environment: | ||||||||
Last Closed: | 2017-07-18 12:20:34 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: | |||||||||
Attachments: |
|
Description
jeremy9856
2016-07-14 16:37:31 UTC
Can you please confirm that it works with nouveau? If so, it's unlikely that we can do much about it, since proprietary drivers are...proprietary...we can't look at the source and figure out what's going on. I just tried with Nouveau and it worked. I just have this output in the terminal : file:///usr/lib64/qt5/qml/QtQuick/Controls/ComboBox.qml:560:5: QML Menu: Binding loop detected for property "__selectedIndex" I posted on Nvidia forum. I hope they could help : https://devtalk.nvidia.com/default/topic/950867/linux/fedora-24-liveusb-creator-can-t-start-crash-/ I don't think it's a nvidia issue as it works fine here with 367.27. That's interesting. I have a GT430, with the 367.27 too, from Negativo17 installed like that : $ sudo dnf install nvidia-driver dkms-nvidia nvidia-driver-libs.i686 I use this in /etc/environment but I don't think it's related : CLUTTER_PAINT=disable-clipped-redraws:disable-culling CLUTTER_VBLANK=True Can you post a backtrace please? su dnf install gdb python-debuginfo gdb -ex r --args python /usr/bin/liveusb-creator When it crashes type bt that will generate a backtrace (post it here). Created attachment 1181262 [details]
Backtrace
Here is a backtrace. It asked to install a truck load of packages. I managed to install everything except this :
dnf debuginfo-install libglvnd-0.1.0-3.8277115.fc24.x86_64
activation du dépôt fedora-debuginfo
activation du dépôt rpmfusion-free-updates-debuginfo
activation du dépôt rpmfusion-nonfree-debuginfo
activation du dépôt rpmfusion-nonfree-updates-debuginfo
activation du dépôt rpmfusion-free-debuginfo
activation du dépôt updates-debuginfo
Vérification de l'expiration des métadonnées effectuée il y a 0:11:02 sur Mon Jul 18 21:39:50 2016.
Le paquet glibc-debuginfo-2.23.1-8.fc24.x86_64 est déjà installé, ignorer
Le paquet nss-softokn-debuginfo-3.25.0-1.0.fc24.x86_64 est déjà installé, ignorer
Le paquet libX11-debuginfo-1.6.3-3.fc24.x86_64 est déjà installé, ignorer
Le paquet libxcb-debuginfo-1.11.1-2.fc24.x86_64 est déjà installé, ignorer
Le paquet libXau-debuginfo-1.0.8-6.fc24.x86_64 est déjà installé, ignorer
Le paquet libXext-debuginfo-1.3.3-4.fc24.x86_64 est déjà installé, ignorer
Dépendances résolues.
Rien à faire.
Terminé !
(In reply to Adam Williamson from comment #1) > Can you please confirm that it works with nouveau? If so, it's unlikely that > we can do much about it, since proprietary drivers are...proprietary...we > can't look at the source and figure out what's going on. looking at the backtrace provided it would appear to be a QT5 issue, can you reassign it? liveusb-creator crashes on f25 here (gdb) bt #0 0x00007ffff6d947fe in __strstr_sse2 () at /lib64/libc.so.6 #1 0x00007fffdea2292a in QSGRenderContext::initialize(QOpenGLContext*) () at /lib64/libQt5Quick.so.5 #2 0x00007fffdea415d9 in QSGRenderThread::run() () at /lib64/libQt5Quick.so.5 #3 0x00007fffe8ae3a1a in QThreadPrivate::start(void*) () at /lib64/libQt5Core.so.5 #4 0x00007ffff77e26ca in start_thread () at /lib64/libpthread.so.0 #5 0x00007ffff6e0ce2f in clone () at /lib64/libc.so.6 Maybe we should open an issue on Qt bug tracker ? I'm convinced this is more of a driver issue than a Qt bug. Generally, fixing bugs related to closed-source drivers is not feasible for both maintainers and the original developers of the graphical software. FWIW I just tried on current F25 with an NVIDIA GeForce 9600 GT and nouveau, and it launches OK. Yeah with nouveau it work but it doesn't work with Nvidia proprietary drivers. See also: https://bugs.kde.org/show_bug.cgi?id=348812 and https://bugreports.qt.io/browse/QTBUG-50368 Triaging component, and adjusting summary And, as discussed on the user's list, a workaround for this issue seems to be modifying /usr/bin/liveusb-creator to add a single line at the start of the code. import os import sys from OpenGL import GL <<<<----- added line from liveusb import _ This allows it to run on both of my systems using nVidia drivers. I suspect that may simply be a workaround for what's fundamentally a driver bug (opengl not getting initialized properly) So folks don't have to go searching, the interesting part of backtrace from comment #7 : (gdb) bt #0 0x0000000000000000 in () #1 0x00007fffe46c9b8d in QOpenGLContextPrivate::maxTextureSize() (this=0x7fffc01af660, params=0x7fffd126c070, pname=4096, level=0, target=32868) at ../../src/gui/opengl/qopenglfunctions_1_0.h:401 #2 0x00007fffe46c9b8d in QOpenGLContextPrivate::maxTextureSize() (this= 0x555555f406d0) at kernel/qopenglcontext.cpp:381 #3 0x00007fffe48c7ffd in QTextureGlyphCache::populate(QFontEngine*, int, unsigned int const*, QFixedPoint const*) (this=this@entry=0x7fffc01b6d20, fontEngine=<optimized out>, numGlyphs=<optimized out>, glyphs=<optimized out>, positions=<optimized out>) at painting/qtextureglyphcache.cpp:202 #4 0x00007fffe012dba0 in QSGTextMaskMaterial::populate(QPointF const&, QVector<unsigned int> const&, QVector<QPointF> const&, QSGGeometry*, QRectF*, QPointF*, QMargins const&) (this=<optimized out>, p=..., glyphIndexes=..., glyphPositions=..., geometry=geometry@entry=0x7fffc01b6c40, boundingRect=boundingRect@entry=0x7fffd126c380, baseLine=<optimized out>, margins=...) at /usr/src/debug/qtdeclarative-opensource-src-5.6.1/src/quick/scenegraph/qsgdefaultglyphnode_p.cpp:450 which doesn't match what's in comment #9 (9's issue is more a symptom of opengl not working, in my experience) actually, comment #9 doesn't appear to be a crash at all, next time, get backtrace from all threads using: thread apply all bt Created attachment 1204880 [details] backtrace from all threads > actually, comment #9 doesn't appear to be a crash at all, next time, get backtrace from all threads using: thread apply all bt Ok so I just did a backtrace from all threads using "thread apply all bt". Liveusb creator don't seem to crash while backtrace but display an empty (black) window. By the way mediawriter 3.97, that seem really really similar to liveusb-creator, works. liveusb-creator still crash, mediawriter not. this issue only appears when using nvidia proprietary drivers, and it's unlikely that we can do much about it, since proprietary drivers are close source. close WONTFIX |