https://bugs.kde.org/show_bug.cgi?id=370646
--- Comment #7 from RJVB ---
I see there's been a chronological glitch in my comments :-/
I wonder: shouldn't it be more elegant to do the actual unregistering
(attributes.erase(x)) in the Attribute dtor? I'll need to check if the
Attribute dtor is act
https://bugs.kde.org/show_bug.cgi?id=370646
--- Comment #6 from RJVB ---
#1 (anonymous namespace)::NoteSharedRegistrar::~NoteSharedRegistrar()() at
.../kdepim-4.14.git/noteshared/attributes/attributeregistrar.cpp:44
#2 (anonymous namespace)::NoteSharedRegistrar::~NoteSharedRegistrar()() at
.../kd
https://bugs.kde.org/show_bug.cgi?id=370646
--- Comment #5 from RJVB ---
I don't understand how or why dynamically allocated memory can go stale when
the shared library which allocated it is unloaded, but apparently something
like that happens.
When I replace libnoteshared's attributeregistrar's
https://bugs.kde.org/show_bug.cgi?id=370646
--- Comment #4 from RJVB ---
Why this happens is unclear to me, but the line
const bool registered = dummy();
in libnoteshared's attributeregistrar.cpp is executed twice. I think that can
only happen when the library is loaded twice.
I may be saying
https://bugs.kde.org/show_bug.cgi?id=370646
--- Comment #3 from RJVB ---
I ran kontact through valgrind to get another angle. I think in the end it just
confirms what I already thought:
```
--72548-- run: /usr/bin/dsymutil "/opt/local/lib/kde4/kcm_knote.so"
--72548-- run: /usr/bin/dsymutil
"/opt
https://bugs.kde.org/show_bug.cgi?id=370646
--- Comment #2 from RJVB ---
Backtrace below.
Among the things I tried was not deleting only attributes of type
NoteDisplayAttribute. That actually made things worse, somehow, causing crashes
even when the application was starting up.
I do wonder abou
https://bugs.kde.org/show_bug.cgi?id=370646
Laurent Montel changed:
What|Removed |Added
CC||mon...@kde.org
--- Comment #1 from Laurent Mon