Bug 339751 - large tif files crash thumbnail generator (out of memory / sig11)
Summary: large tif files crash thumbnail generator (out of memory / sig11)
Status: RESOLVED FIXED
Alias: None
Product: digikam
Classification: Applications
Component: Thumbs-Engine (show other bugs)
Version: 4.3.0
Platform: Debian unstable Linux
: NOR crash
Target Milestone: ---
Assignee: Digikam Developers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-10-07 14:00 UTC by Peter Gervai
Modified: 2019-08-15 15:23 UTC (History)
2 users (show)

See Also:
Latest Commit:
Version Fixed In: 6.3.0
Sentry Crash Report:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Peter Gervai 2014-10-07 14:00:41 UTC
I have found bug #314260 which described the problem: a directory contained a dozen large (20+MB) TIFs (generated by hugin btw) and it crashed digikam at startup. It was a bitch to figure out the problem since neither console nor debug contained the exact filenames (nor hashes nor anything useful), so first I thought it's about the pretty large thumbnails.db. Later it turned out that digikam crashed by scanning or displaying a subdirectory so I clicked them one-by-one (hadda be fast, like an action game) until the crasher was found.

digikam(5561)/digikam (core): Cannot allocate buffer of size 2408382400 
was one of the messages which probably mean that digikam wanted to use some extreme amounts of memory to do whatever thumbnailing. I believe I'd prefer _not_ to see thumbnails instead of startup crashing so probably some pre-allocation checks are in high demand.

Reproducible: Always

Steps to Reproduce:
1. 30+ TIFs in 20+MB size thrown in the middle of a large image collection subdir tree
2. start digikam
3. crash in 5 seconds



I used the Debian package, neither git nor debug version as I don't have right now the bandwidth to gather all the stuff to compile it.
Comment 1 Christoph Feck 2014-10-07 17:29:51 UTC
Are you able to view them in Gwenview?
Comment 2 Peter Gervai 2014-10-07 19:12:36 UTC
I use geeqie generally and it was able to show them. Never heard about gwen :)
But honestly I was so angry that I've deleted the files on the spot, so if you relaly believe it's related to the specific content I can recreate them with hugin. 

(Technically life isn't that simple: the files have been there for 2 months now and they ought to be already indexed without a crash. I mean, yeah, digicam crashes all the time (it proceeds quit with a crash most of the time) so I cannot be sure which one was it but startup crash wasn't on the menu until now. I am puzzled why it didn't crash so far and why started now, I believe I haven't upgraded it for months now. Nor kde libs. I can accept this bug to be closed after this confession, since it's partly based on unexplainable mistery.)
Comment 3 caulier.gilles 2014-10-07 20:39:27 UTC
Run digiKam into GDB as explained here  :

https://www.digikam.org/contrib

And report the crash trace...

Note : I use also Hugin to make big panorama (200/300Mb TIFF), and there is no crash here to render thumbnails, preview and load image in editor.

Gilles Caulier
Comment 4 caulier.gilles 2014-10-07 20:52:40 UTC
example of 241Mb TIFF panorama show in digiKam :

https://www.flickr.com/photos/digikam/15449257226/sizes/o/
Comment 5 Peter Gervai 2014-10-07 21:32:48 UTC
(In reply to Gilles Caulier from comment #3)
> Run digiKam into GDB as explained here  :
> 
> https://www.digikam.org/contrib
> 
> And report the crash trace...

Crash trace was pretty useless. It kept being pretty useless with exiv2 and kexiv2 debug symbols, and I couldn't conjure digikam debug symbols since, as I have mentioned, I don't have the means right now. (lib debug took 50 minutes to retrieve. go figure.)

Basically it died in libkexiv2's thumbnail code, maybe it's the culprit, now that I started to think about it.

> Note : I use also Hugin to make big panorama (200/300Mb TIFF), and there is
> no crash here to render thumbnails, preview and load image in editor.

I know, me too. 


But look what I found: the straces from yesterday. And they SIGSEGV'd on a different file (this bug was originally reproduced by not SIGSEGV but the mem allocation error, but they were oscillating a day before, probably depending on the selected albums of the collection). And now I went there and checked the file. Bang, sigsegv. Here's a trace which is _very_ similar to the one with the mem problem above:

#0  0x00007fffedd396e5 in Exiv2::PreviewManager::getPreviewImage (this=0x7fff7803c390, properties=...) at preview.cpp:1157
#1  0x00007ffff5c12902 in KExiv2Iface::KExiv2Previews::data (this=0x7fff837fd710, index=0) at ../../libkexiv2/kexiv2previews.cpp:170
#2  0x00007ffff5c12a8b in KExiv2Iface::KExiv2Previews::image (this=<optimized out>, index=<optimized out>) at ../../libkexiv2/kexiv2previews.cpp:187
#3  0x00007ffff551807b in Digikam::ThumbnailCreator::createThumbnail(Digikam::ThumbnailInfo const&, QRect const&) const () from /usr/lib/digikam/libdigikamcore.so.4.3.0
#4  0x00007ffff551bcf9 in Digikam::ThumbnailCreator::load(QString const&, QRect const&, bool) const () from /usr/lib/digikam/libdigikamcore.so.4.3.0
#5  0x00007ffff551cbe2 in Digikam::ThumbnailCreator::load(QString const&) const () from /usr/lib/digikam/libdigikamcore.so.4.3.0
#6  0x00007ffff552b31f in ?? () from /usr/lib/digikam/libdigikamcore.so.4.3.0
#7  0x00007ffff5500b8e in Digikam::LoadSaveThread::run() () from /usr/lib/digikam/libdigikamcore.so.4.3.0
#8  0x00007ffff55352ee in Digikam::DynamicThread::DynamicThreadPriv::run() () from /usr/lib/digikam/libdigikamcore.so.4.3.0
#9  0x00007ffff1c528a4 in ?? () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#10 0x00007ffff1c5f3ff in ?? () from /usr/lib/x86_64-linux-gnu/libQtCore.so.4
#11 0x00007fffecb370a4 in start_thread () from /lib/x86_64-linux-gnu/libpthread.so.0
#12 0x00007ffff0e80c2d in clone () from /lib/x86_64-linux-gnu/libc.so.6

The image is pretty large so it's available here: http://vilya.grin.hu/t/digikrash01.png I believe this is the one but it's not easy to tell since digikam crashes when entering the collection.
Comment 6 Christoph Feck 2014-10-07 21:43:27 UTC

*** This bug has been marked as a duplicate of bug 339144 ***
Comment 7 caulier.gilles 2019-08-15 15:23:46 UTC
Fixed with bug #339144