Project

General

Profile

Bug #961

Crash in digikam while reading metadata from a .MTS movie file

Added by Robin Mills over 7 years ago. Updated over 6 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
video
Target version:
Start date:
08 Jun 2014
Due date:
% Done:

100%

Estimated time:

Description

This is topic 1737 from the forum. http://dev.exiv2.org/boards/3/topics/1737


Related issues

Related to Exiv2 - Feature #956: Support for .MTS movie formatClosed22 May 2014

Actions
Related to Exiv2 - Bug #963: Digikam is crashing for the image formats that it is not able to detect.Closed08 Jun 201411 Jun 2014

Actions

History

#1

Updated by Robin Mills over 7 years ago

I don't see a bug here. Exiv2 is throwing an exception.

1014 rmills@rmills-ubuntu:~/gnu/exiv2/trunk $ exiv2 -v ../bugs/00056.MTS 
File 1/1: ../bugs/00056.MTS
Exiv2 exception in print action for file ../bugs/00056.MTS:
../bugs/00056.MTS: The file contains data of an unknown image type
1015 rmills@rmills-ubuntu:~/gnu/exiv2/trunk $ 

#2

Updated by Robin Mills over 7 years ago

I've had a look at the trace in http://dev.exiv2.org/attachments/610/digikam_crash_dump_05.txt and it looks like an unhandled exception:

Catchpoint 1 (exception thrown), __cxxabiv1::__cxa_throw (obj=0x7fff58004600, 
130    
    tinfo=0x3a3aa980a0 <typeinfo for Exiv2::BasicError<char>>, 
131    
    dest=0x33e6e9f030 <Exiv2::BasicError<char>::~BasicError()>)
132    
    at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:63
133    
63      PROBE2 (throw, obj, tinfo);
134    
(gdb) #0  __cxxabiv1::__cxa_throw (obj=0x7fff58004600, 
135    
    tinfo=0x3a3aa980a0 <typeinfo for Exiv2::BasicError<char>>, 
136    
    dest=0x33e6e9f030 <Exiv2::BasicError<char>::~BasicError()>)
137    
    at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:63
138    
#1  0x00000033e6ee54a6 in Exiv2::ImageFactory::open(std::string const&) ()
139    
   from /usr/lib64/libexiv2.so.13

Here's the code in Exiv2:

    Image::AutoPtr ImageFactory::open(const std::string& path)
    {
        BasicIo::AutoPtr io(new FileIo(path));
        Image::AutoPtr image = open(io); // may throw
        if (image.get() == 0) throw Error(11, path);
        return image;
    }

and the error message:
error.cpp:        { 11, N_("%1: The file contains data of an unknown image type") }, // %1=path
#3

Updated by Robin Mills over 7 years ago

I've consolidated discussion of this from Issue #956 and Forum Topic #1737 into here.

Gilles has provided the following very helpful information in 1737:

From the backtrace, we can see :

#1 0x00000033e6ee54a6 in Exiv2::ImageFactory::open(std::string const&) () from /usr/lib64/libexiv2.so.13
#2 0x0000003a3a82956a in KExiv2Iface::KExiv2::load (this=0x7fff58389f20, filePath=...)
at /home/shipman/Downloads/dk/extra/libkexiv2/libkexiv2/kexiv2.cpp:298
#3 0x0000003a3bb15cf9 in Digikam::DMetadata::load (this=0x7fff58389f20, filePath=...)
at /home/shipman/Downloads/dk/core/libs/dmetadata/dmetadata.cpp:110

#1 is call from digiKam core;
#2 is call relevant in libkexiv2
#3 is call into Exiv2.

Look code from libkexiv2, and you will see that Exiv2 call is wrapped in C++ exception handler.

https://projects.kde.org/projects/kde/kdegraphics/libs/libkexiv2/repository/revisions/master/entry/libkexiv2/kexiv2.cpp#L298

And he has defended the digikam exception handling in libkexif in #956 by saying:

digiKam itself do not have C++ exception, but it never use Exiv2 directly. For code design, all pass through libkexiv2, which is a C++ wrapper. All methods from libkexiv2 use C++ exception...

So typically, digiKam must never crash due to a C++ exception from Exiv2...
__

I'm not certain the handler in libkexiv2 is bullet proof. It only checks for Exiv2 errors. So, if the Exiv2 Error code was to throw an exception (such as stack-overflow or out-of-memory) it wouldn't be caught.

It's not easy to deal with this without recreating the environment with libkexiv2 wrapper and DigiKam present. I'm not trying to "pass the buck", however maybe Gilles could augment the exception handler in libkexiv2 to determine if an exception is being thrown by something else. I'm wondering if Exiv2 dies while trying to throw error 11.

Are there exceptions that cannot be caught? For example stack overflow. Maybe we've disappeared into a black hole of some kind.

There's another issue (#960) which involves memory corruption while opening files. This evening I will investigate that and perhaps #960 and #961 are the same issue.

#4

Updated by Gilles Caulier over 7 years ago

Yes sure, libkexiv2 wrapper is dedicated to play properly with Exiv2 exception, without to touch digiKam code.

As all Exception wrapper is on private implementation, touching this will not break Binary compatibility of library.

Currently we have this king of implementation everywhere :

try {
// Exiv2 C++ exception calls here.
}
catch( Exiv2::Error& e ) {
d->printExiv2ExceptionError("Cannot load metadata from file ", e);
}

What do you want that i change here in this exception handling exactly ?

Gilles Caulier

#5

Updated by Robin Mills over 7 years ago

Thanks for getting back to me, Gilles. How about:

try {
  // code here
}
catch( Exiv2::Error& e ) {
    d->printExiv2ExceptionError("Cannot load metadata from file ", e);
}
catch (...) { 
   cout << "default exception";
}

It is possible that printExiv2ExceptionError() is throwing an exception! So the simplest code might be:

try {
  // code here
}
catch (...) { 
   cout << "default exception";
}

Robin

#6

Updated by Gilles Caulier over 7 years ago

Done with this commit in libkexiv2 :

http://commits.kde.org/libkexiv2/8a8dc535d504b70776677b933ac761bba0f7a4ae

I closed also this KDE bugzilla entry :

https://bugs.kde.org/show_bug.cgi?id=331679

Gilles Caulier

#7

Updated by Robin Mills over 7 years ago

  • Status changed from Assigned to Resolved

Thanks, Gilles.

It's always nice to work with you. I'll mark #961 and #963 as "resolved" and they'll be closed during review prior to the release of 0.25.

Robin

#8

Updated by Robin Mills over 6 years ago

  • % Done changed from 0 to 100
#9

Updated by Andreas Huggel over 6 years ago

  • Status changed from Resolved to Closed

Also available in: Atom PDF