[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[Simon Andric via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

Simon Andric  changed:

   What|Removed |Added

 CC||simonandr...@gmail.com

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[RJVB via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #13 from RJVB  ---
Created attachment 96640
  --> https://bugs.kde.org/attachment.cgi?id=96640=edit
[OS X] valgrind --track-origins=yes --smc-check=all-non-file --error-limit=no
kdevelop5

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[RJVB via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #14 from RJVB  ---
Created attachment 96641
  --> https://bugs.kde.org/attachment.cgi?id=96641=edit
[OS X] valgrind --track-origins=yes --smc-check=all-non-file --error-limit=no
kdevelop5

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[RJVB via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #10 from RJVB  ---
I've got the log, but a priori it just shows a crash while dlopen'ing
kdevclangsupport without any further explanation as to what the gdb backtrace
above already shows.

I rebuilt kdevclangsupport against the clang 3.5.1 version I grabbed from
Debian/Exp. a year ago and built via a PPA of my own. And that one seems to
work, contrary to Ubuntu's current 3.6.0 release and both the 3.6.2 and the
3.7.1 builds from the LLVM apt servers. Those last 2 fail in the same way,
which leads me to think it must be some incompatibility in the way they build
their stuff and I built KF5/KDevelop.
This particular issue is too resource demanding to continue to look into it.
I'll see if I can reproduce the crash-on-exit issue, now that I managed to work
around the crash-on-startup :)

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[Kevin Funk via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #11 from Kevin Funk  ---
For the record: Distro-provided Clang/LLVM works fine for me on Ubuntu Wily.
I'm sure I tried both the 3.6 and 3.7 versions at some point, without problems.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[RJVB via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #12 from RJVB  ---
(In reply to Kevin Funk from comment #11)
> For the record: Distro-provided Clang/LLVM works fine for me on Ubuntu Wily.

The crash I saw with Ubuntu's Clang 3.6 was different from the one I saw with
the LLVM-provided builds. Ubuntu only provide an early 3.6.0 for "Trusty", it's
not at all impossible there's a real bug in there.

Here's a wild thought: when I tried to rebuild that year old 3.5.1 LLVM
toolchain earlier today using gcc-5.3 instead of gcc-4.8 I ran into a build
failure. Gcc 5 didn't accept access to a private member of a template class in
one of LLVM's headers (in one of the class ctors...), code that is accepted
without issues by gcc-4.8 . Not that it seems likely, but suppose that is
caused by a difference in default C++ flavour in gcc-5, couldn't that also lead
to subtle incompatibilities between libraries built with an older g++ and a
host application built with g++5?

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[RJVB via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #8 from RJVB  ---
Created attachment 96525
  --> https://bugs.kde.org/attachment.cgi?id=96525=edit
valgrind --track-origins=yes -v --error-limit=no kdevelop5 (against llvm 3.7)

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[Milian Wolff via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #9 from Milian Wolff  ---
please rerun without -v but with --smc-check=all-non-file

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[Milian Wolff via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #5 from Milian Wolff  ---
please use valgrind to find out whats going on instead of guessing.

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[Kevin Funk via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

Kevin Funk  changed:

   What|Removed |Added

 Status|UNCONFIRMED |CONFIRMED
 CC||kf...@kde.org
 Ever confirmed|0   |1
Summary|crash when exiting with |crash when exiting with
   |parser running  |parser running
   ||[KDevelop::BackgroundParser
   ||::~BackgroundParser]

--- Comment #1 from Kevin Funk  ---
I think I have seen this on Linux before as well.

A valgrind run might be enlightening...

-- 
You are receiving this mail because:
You are watching all bug changes.

[kdevelop] [Bug 357630] crash when exiting with parser running [KDevelop::BackgroundParser::~BackgroundParser]

[RJVB via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=357630

--- Comment #2 from RJVB  ---
Yeah, maybe when I am sure I can reproduce it, and preferably under Linux.
Valgrind'ing CPU & memory-hungry processes has been a very efficient way to get
KPs on OS X for me so I tend to avoid it.

-- 
You are receiving this mail because:
You are watching all bug changes.