Bug 385592

Which digiKam version do you use exactly ?

Gilles Caulier

(In reply to caulier.gilles from comment #1)
> Which digiKam version do you use exactly ?
> 
> Gilles Caulier

From git at commit 841cd2947cccfff1b541e97e5a626ccdc2cef979

Btw. I've seen those segfaults and hangs on renaming pictures several times before. It's only the first time I digged into this.

I do not see an probleme in the ImageInfo::filePath() function. And it crashes in system or Qt. Which distribution do you use?

Maik

(In reply to Maik Qualmann from comment #4)
> I do not see an probleme in the ImageInfo::filePath() function. And it
> crashes in system or Qt. Which distribution do you use?
> 
> Maik

I'm using Gentoo with live-ebuild for digikam.

I don't think ImageInfo::filePath() is the problem in the first place. It seems to me, m_data is invalidated by some other thread, and this way a nullptr is pushed down.

(In reply to Johannes Hirte from comment #5)
> (In reply to Maik Qualmann from comment #4)
> > I do not see an probleme in the ImageInfo::filePath() function. And it
> > crashes in system or Qt. Which distribution do you use?
> > 
> > Maik
> 
> I'm using Gentoo with live-ebuild for digikam.
> 
> I don't think ImageInfo::filePath() is the problem in the first place. It
> seems to me, m_data is invalidated by some other thread, and this way a
> nullptr is pushed down.

Sorry, not a nullptr, an invalid pointer.

Created attachment 108290 [details]
ImageInfo.patch

Johannes,

can you apply this patch and test it? This crash I have never seen or reproduced here.

Maik

(In reply to Maik Qualmann from comment #7)
> Created attachment 108290 [details]
> ImageInfo.patch
> 
> Johannes,
> 
> can you apply this patch and test it? This crash I have never seen or
> reproduced here.
> 
> Maik

got this

Thread 71 "Thread (pooled)" received signal SIGBUS, Bus error.
[Switching to Thread 0x7f5584d0e700 (LWP 24038)]
malloc_consolidate (av=0x7f551c000020) at malloc.c:4489
4489    malloc.c: No such file or directory.

I can confirm, that the error changed from SIGSEGV to SIGBUS. I can't reproduce the error reliable, but when it happens, it is SIGBUS now.

(In reply to Johannes Hirte from comment #9)
> I can confirm, that the error changed from SIGSEGV to SIGBUS. I can't
> reproduce the error reliable, but when it happens, it is SIGBUS now.

Okay, I have to revert this. Now it's SIGABRT:

Thread 23 "Thread (pooled)" received signal SIGABRT, Aborted.
[Switching to Thread 0x7fb0317fa700 (LWP 6324)]
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
51      ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.

(gdb) bt
#0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
#1  0x00007fb0f2936506 in __GI_abort () at abort.c:90
#2  0x00007fb0f297a7b7 in __libc_message (action=action@entry=(do_abort | do_backtrace), fmt=fmt@entry=0x7fb0f2a8a0a0 "*** Error in `%s': %s: 0x%s ***\n") at ../sysdeps/posix/libc_fatal.c:181
#3  0x00007fb0f29822e9 in malloc_printerr (action=<optimized out>, str=0x7fb0f2a8a7c0 "malloc(): smallbin double linked list corrupted", ptr=<optimized out>, ar_ptr=<optimized out>) at malloc.c:5423
#4  0x00007fb0f2985af9 in _int_malloc (av=0x7fb014000020, bytes=<optimized out>) at malloc.c:3593
#5  0x00007fb0f2987525 in __GI___libc_malloc (bytes=64) at malloc.c:3083
#6  0x00007fb0f2f93398 in operator new(unsigned long) () from /usr/lib/gcc/x86_64-pc-linux-gnu/7.2.0/libstdc++.so.6
#7  0x00007fb0f6b12b19 in QSqlField::QSqlField (this=0x7fb0317f9580, fieldName=..., type=QVariant::ByteArray) at kernel/qsqlfield.cpp:163
#8  0x00007fb0dc966290 in qToField (field=0x7fb01413f190, tc=<optimized out>) at qsql_mysql.cpp:334
#9  0x00007fb0dc9668a3 in QMYSQLResult::record (this=<optimized out>) at qsql_mysql.cpp:836
#10 0x00007fb0f6b088b8 in QSqlQuery::record (this=this@entry=0x7fb0317f9708) at kernel/qsqlquery.cpp:915
#11 0x00007fb0f50b8d4f in Digikam::BdEngineBackend::readToList (this=this@entry=0x1b65d20, query=...) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/database/engine/dbenginebackend.cpp:838
#12 0x00007fb0f50b8edb in Digikam::BdEngineBackend::handleQueryResult (this=this@entry=0x1b65d20, query=..., values=values@entry=0x7fb0317f9750, lastInsertId=lastInsertId@entry=0x0)
    at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/database/engine/dbenginebackend.cpp:875
#13 0x00007fb0f50ba59b in Digikam::BdEngineBackend::execSql (this=this@entry=0x1b65d20, sql=..., boundValue1=..., boundValue2=..., values=values@entry=0x7fb0317f9750, lastInsertId=lastInsertId@entry=0x0)
    at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/database/engine/dbenginebackend.cpp:901
#14 0x00007fb0f50c42ca in Digikam::ThumbsDb::findByHash (this=<optimized out>, uniqueHash=..., fileSize=<optimized out>) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/database/thumbsdb/thumbsdb.cpp:131
#15 0x00007fb0f52e17bc in Digikam::ThumbnailCreator::loadThumbsDbInfo (this=this@entry=0x1ff6740, info=...) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threadimageio/thumbnailcreator.cpp:882
#16 0x00007fb0f52e3fb7 in Digikam::ThumbnailCreator::loadFromDatabase (this=this@entry=0x1ff6740, info=...) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threadimageio/thumbnailcreator.cpp:917
#17 0x00007fb0f52e5e8b in Digikam::ThumbnailCreator::load (this=this@entry=0x1ff6740, identifier=..., rect=..., pregenerate=pregenerate@entry=false) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threadimageio/thumbnailcreator.cpp:246
#18 0x00007fb0f52e60e8 in Digikam::ThumbnailCreator::load (this=this@entry=0x1ff6740, identifier=...) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threadimageio/thumbnailcreator.cpp:181
#19 0x00007fb0f52f49dd in Digikam::ThumbnailLoadingTask::execute (this=0x66baa60) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threadimageio/thumbnailtask.cpp:166
#20 0x00007fb0f52cc6fe in Digikam::LoadSaveThread::run (this=0x7fb0f5f261c0 <Digikam::(anonymous namespace)::Q_QGS_defaultObject::innerFunction()::holder>)
    at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threadimageio/loadsavethread.cpp:133
#21 0x00007fb0f5309ade in Digikam::DynamicThread::DynamicThreadPriv::run (this=0x1ff6450) at /var/tmp/portage/media-gfx/digikam-9999/work/digikam-9999/libs/threads/dynamicthread.cpp:183
#22 0x00007fb0f3388385 in QThreadPoolThread::run (this=0x30465a0) at thread/qthreadpool.cpp:99
#23 0x00007fb0f338b3d0 in QThreadPrivate::start (arg=0x30465a0) at thread/qthread_unix.cpp:368
#24 0x00007fb0eff2e8aa in start_thread (arg=0x7fb0317fa700) at pthread_create.c:465
#25 0x00007fb0f29ff0bf in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

I've put 30 pictures into an album and started renaming with the integrated renaming tool. This could take some time, but it will trigger the bug sooner or later.

And now SIGSEGV is back. The backtrace is similar to the one above. I can provide it, if you need it. Here is the workflow, how I reproduced it:

1. put 30 images into an album
2. rename the images  crashtest-001 - crashtest-030
3. rename the images  crashtest-101 - crashtest-130
4. repeat at 2.

Git commit f8d3249f580b78c0afad9dd83a929a7e12652759 by Maik Qualmann.
Committed on 12/10/2017 at 20:14.
Pushed by mqualmann into branch 'master'.

polish and add missing reference to ImageInfo lists

M  +1    -1    libs/database/item/imageinfolist.cpp
M  +1    -1    libs/database/item/imageinfolist.h
M  +6    -6    libs/database/utils/dio.cpp
M  +2    -2    libs/database/utils/dio.h
M  +3    -3    libs/database/utils/dio_p.h
M  +1    -1    libs/fileactionmanager/fileworkeriface.h
M  +1    -1    libs/fileactionmanager/metadatahub.cpp
M  +1    -1    libs/fileactionmanager/metadatahub.h
M  +1    -1    libs/fileactionmanager/metadatahubmngr.cpp
M  +1    -1    libs/fileactionmanager/metadatahubmngr.h
M  +1    -1    libs/imageproperties/disjointmetadata.cpp

https://commits.kde.org/digikam/f8d3249f580b78c0afad9dd83a929a7e12652759

Can you please test the latest git/master version?

Maik

I can now reproduce the crash...

Maik

Git commit 38ebdc7496037ad38bc356baa0242c5587ebd9c4 by Maik Qualmann.
Committed on 12/10/2017 at 21:14.
Pushed by mqualmann into branch 'master'.

try to find the renaming crash

M  +2    -2    libs/database/utils/dio.cpp

https://commits.kde.org/digikam/38ebdc7496037ad38bc356baa0242c5587ebd9c4

Git commit 89e61652264777a38578c5bf9b944a4c82173294 by Maik Qualmann.
Committed on 17/10/2017 at 18:32.
Pushed by mqualmann into branch 'master'.

add backward QHash to the ImageInfo cache to resolve changed names
Related: bug 382312, bug 384577

M  +6    -0    libs/database/item/imageinfocache.cpp
M  +1    -0    libs/database/item/imageinfocache.h
M  +2    -2    libs/database/utils/dio.cpp

https://commits.kde.org/digikam/89e61652264777a38578c5bf9b944a4c82173294

Johannes,

can you please test the latest git/master code?

Maik

Looks good, I wasn't able to reproduce a crash.

Thanks Maik to patch the code.

Thanks Johannes for the feedback.

I close this file now.

Had another segfault, this time outside of a gdb session, so I don't have a backtrace. It happened when moving pictures with dolphin into a collection during a running digikam. So I'm not sure this is the same bug. I'm trying to reproduce inside gdb and reopen this bug or open a new one, as soon as I have some more facts.

*** Bug 390287 has been marked as a duplicate of this bug. ***