https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #38 from Maik Qualmann ---
Ok, thanks Simon. I do not want to do another test. I am sure that the cause of
the QHash list is me remaining dead elements that are being read out via access
with iterator. Why the error occurs only or rather wit
https://bugs.kde.org/show_bug.cgi?id=392134
Maik Qualmann changed:
What|Removed |Added
Status|UNCONFIRMED |RESOLVED
Resolution|---
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #36 from Simon ---
Created attachment 112595
--> https://bugs.kde.org/attachment.cgi?id=112595&action=edit
Log&BT with digikam-6.0.0-git-20180511T042604-x86-64.appimage
The same seg fault occurred again, but there were no debug log lines
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #35 from caulier.gil...@gmail.com ---
The AppImage bundles are now updated to files.kde.org
Gilles
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #34 from caulier.gil...@gmail.com ---
Button pressed...
Q : why do you think a difference with QHash in AppImage ?
The only difference is located in OpenCV3 rules to compile : all not used
features are disable :
https://cgit.kde.org/digika
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #33 from Maik Qualmann ---
Gilles, would be nice if you press the button again. Only the AppImage is
needed.
Maik
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #32 from Maik Qualmann ---
Git commit 979c7b188435ccd38e5dbd76b492d1e8b5505da0 by Maik Qualmann.
Committed on 10/05/2018 at 18:26.
Pushed by mqualmann into branch 'master'.
QHash test, this commit is for testing purposes only and will be re
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #31 from Maik Qualmann ---
It may also be that it is more likely with the AppImage. The QHash are "salted"
with a value of a random number generator. Maybe here is something "different"
in the AppImage. I will insert the test in git/master a
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #30 from Simon ---
It's currently at 24% - the last times it never exceeded 11% before crashing. I
will leave it running, but chances are that either due to using compiled
version, different dependencies or the debug statements the "racy" se
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #29 from Maik Qualmann ---
GDB is not needed if it crashes, the last lines of the console log are of
interest.
Maik
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #28 from Maik Qualmann ---
Created attachment 112512
--> https://bugs.kde.org/attachment.cgi?id=112512&action=edit
qhashtest.patch
Okay, Simon can you please test this test patch? I expect a crash (: =))
Maik
--
You are receiving this
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #27 from caulier.gil...@gmail.com ---
Simon,
The dependencies list has been really simplified everywhere.
=> less KF5
=> no libkipi
=> no kipiplugins => in DK core now.
All become step by step more simple.
This is the goal : puzzle simpli
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #26 from Simon ---
I just tested whether I can build digikam on debian stable and it was
surprisingly doable (expected dependency hell). So I can run tests based on
patches now, unless the problem is something specific to the appimage. I don
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #25 from Maik Qualmann ---
Very interesting, Simon, if you do not mind I would like to do a second test.
If we can actually trace it back to QHash, we also need to reconsider the use
elsewhere where a QString is used as a key. I would once a
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #24 from Simon ---
I just reran the face detection with
digikam-6.0.0-git-20180507T123919-x86-64.appimage and no segfault occurred,
while with digikam-6.0.0-git-20180430T091500-x86-64.appimage it happened every
time. So from my point of view
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #23 from Maik Qualmann ---
Git commit fe647204d0bfcd30946f27ebd7bb822a1f6e1cf1 by Maik Qualmann.
Committed on 07/05/2018 at 10:23.
Pushed by mqualmann into branch 'master'.
replace for a crash test QHash with QMap in the LoadingCache
M +7
https://bugs.kde.org/show_bug.cgi?id=392134
Maik Qualmann changed:
What|Removed |Added
CC||freisi...@gmail.com
--- Comment #22 from Maik Q
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #21 from caulier.gil...@gmail.com ---
Maik,
Today i cleanup the face regions from DB, and run again the process in same
condition, excepted with GDB in background. Crash is not reproducible...
It sound like a race condition somewhere...
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #20 from caulier.gil...@gmail.com ---
qt bearer thread : http://doc.qt.io/qt-5/bearer-management.html
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #19 from caulier.gil...@gmail.com ---
"Qt http thread" appear when map view is active using googlemaps. It's the
thread used by webkit in background.
webengine (chromium based) has a dedicated process for that, not a thread. It's
more compli
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #18 from caulier.gil...@gmail.com ---
Just for info : in face management code, explicit constructor with 1 arguments
cannot be done, as Krazy report
core/utilities/facemanagement/facepipeline.h: line# 73,89 (2)
Look my commit :
https://co
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #17 from caulier.gil...@gmail.com ---
Why a pure virtual method is called ? Why the C++ exception is not caught by
digiKam. The debug trace come from certainly OpenCV.
I use openCV 3.4.0, compiled myself with the minimum required for digiKam
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #16 from caulier.gil...@gmail.com ---
digiKam has crashed finally after one hour :
digikam.metaengine: Loading image history ""
digikam.metaengine: DateTime => Exif.Photo.DateTimeOriginal =>
QDateTime(2007-10-09 13:52:02.000 CEST Qt::TimeS
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #15 from caulier.gil...@gmail.com ---
digiKam running with Face detection multicore :
https://www.flickr.com/photos/digikam/41155846361/in/dateposted/
Look the htop view and number of threads running in parallel.
Q : why i cannot see the t
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #14 from caulier.gil...@gmail.com ---
ok, i switch to multicore.
Note that it's currently 15% of scanning with one core. Nothing special, it
run. I can switch to another view without problem. Memory allocation is around
250-300Kb and still s
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #13 from Maik Qualmann ---
My guess is that it's possible that a QHash from a QString is not unique. In
Qt4, non-ANSI characters could cause this. Although the algorithm has been
revised in Qt5...
Maik
--
You are receiving this mail becau
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #12 from Maik Qualmann ---
Gilles,
Please use all cores, the parallel face worker are limited to 3 threads. If I
remember it correctly...
Maik
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #11 from caulier.gil...@gmail.com ---
MAik,
I applied your patch and now i rebuild the faces detection data from scratch.
There are 38000 items in my collection. This will take time with only one core
to use (instead 8)...
Gilles
--
You
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #10 from nroycea+...@gmail.com ---
Nope, I grepped for non-ansi characters in the filenames and came up with no
results.
Also, I intentionally disabled the use of all-cores for face recog/detect.
--
You are receiving this mail because:
You
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #9 from caulier.gil...@gmail.com ---
I will try while this week end.
Gilles
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #8 from Maik Qualmann ---
Created attachment 111566
--> https://bugs.kde.org/attachment.cgi?id=111566&action=edit
hashtomap.patch
Gilles,
You could always reproduce a crash if you ran face recognition on all cores.
Can you please try it
https://bugs.kde.org/show_bug.cgi?id=392134
Maik Qualmann changed:
What|Removed |Added
Summary|SIGSEGV While Scanning |SIGSEGV While Scanning
|Faces
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #7 from Maik Qualmann ---
In the past, we have seen problems with the function
SharedLoadingTask::notifyNewLoadingProcess(). After a long debugging I have a
theory, a possible problem with QHash. Do you have non-ANSI characters in your
file
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #6 from nroycea+...@gmail.com ---
with debug:
*
...
digikam.general: Check for finish: 51 packages, 0 infos to filter,
hasFinished() false
Thread 1673 "Thread (pooled)" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0
https://bugs.kde.org/show_bug.cgi?id=392134
Maik Qualmann changed:
What|Removed |Added
CC||metzping...@gmail.com
--- Comment #5 from Maik
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #4 from nroycea+...@gmail.com ---
RAM 8GB
System at 1.8GB when scan begins
Bouncing around 2.1-2.3GB and then higher ranges
Peaking at 3.2GB
Essentially, the program is using less than 1800MB
At the point of crashing, it was using 24% RAM:
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #3 from caulier.gil...@gmail.com ---
What's the amount of memory used while scanning faces ? It's grow step by step
until a critical stage ?
Gilles Caulier
--
You are receiving this mail because:
You are watching all bug changes.
https://bugs.kde.org/show_bug.cgi?id=392134
--- Comment #2 from nroycea+...@gmail.com ---
I just tried the digikam-5.8.0-01-x86-64 version and that crashed as well.
GDB looks to be useless for this portable type of program, but the program
finished with:
*
digikam.facesengine: detectMultiScale
https://bugs.kde.org/show_bug.cgi?id=392134
caulier.gil...@gmail.com changed:
What|Removed |Added
CC||caulier.gil...@gmail.com
--- Comment
39 matches
Mail list logo
