Bug 1841

The backtrace is in fact complete; it points to the TQTipManager destructor accessing a deleted widget.

Can you run Filelight under Valgrind and once it crashes, attach the Valgrind output to this report?

Thanks!

Tim

Created attachment 1880 [details]
xz archive: valgrind --tool=callgrind filelight

(In reply to comment #2)
> Created attachment 1880 [details]
> xz archive: valgrind --tool=callgrind filelight

Actually, I was interested in the memory checker results, so call it without the --tool option. ;-)

Created attachment 1881 [details]
valgrind filelight

>Actually, I was interested in the memory checker results, so call it without
>the --tool option.
I'm not the brightest bulb in the pack --- no output file is created when I run 'valgrind filelight'. I tried '>' redirect too.

How do I capture the output?

Attached is what I copied from the terminal.

(In reply to comment #4)
> Created attachment 1881 [details]
> valgrind filelight
> 
> >Actually, I was interested in the memory checker results, so call it without
> >the --tool option.
> I'm not the brightest bulb in the pack --- no output file is created when I run
> 'valgrind filelight'. I tried '>' redirect too.
> 
> How do I capture the output?
> 
> Attached is what I copied from the terminal.

That is exactly what I need!  You ran the correct commands.

Tim

Created attachment 1946 [details]
kcrash backtrace

I thought this report had been resolved because I was about to open a new report.

In a default Trinity, I use the konqueror Location menu and select 'Open with Filelight'. All is fine until the mouse pointer hovers over the circular graph. Then Filelight crashes. kcrash backtrace attached.

I cannot reproduce here.
I use konqueror in "file management" profile, then select any (big) folder, use "Document / Open in Filelight" menu.
Filelight shows up, I can browse the graph with my mouse, no problem. Then I close filelight.

Do you have some "strange named" (e.g. with non-standard caracters) directories or files ?

Nothing strange here that I can tell. I tested this with an existing profile and a new profile with the same crash results. Does not matter whether I use the context menu 'Open with...' or use 'Location->Open with Filelight'.

Same results on two different systems. The moment the mouse pointer reaches the very first pixel of the circle graph, the app crashes. I am using Filelight defaults.

The original kcrash backtrace was related to TQTipManager. The new backtrace indicates a different problem.

I can still duplicate the first crash.

Compilation configuration is standard in my build script:

CFLAGS=$CPUOPT \
CXXFLAGS=$CPUOPT \
./configure \
  --prefix=${PREFIX} \
  --sysconfdir=${SYSCONFDIR} \
  --libdir=${LIBDIR} \
  --mandir=${MANDIR} \
  --with-qt-dir=${QTDIR} \
  --with-qt-includes=${QT_INCLUDE_DIR} \
  --with-qt-libraries=${QT_LIB_DIR} \
  $DEBUG_AUTOTOOL_OPT || exit 1

Darrell, I can not reproduce the crash on my system.
Are the two crashes still happening on your machine?

> Are the two crashes still happening on your machine?
Yes, all systems.

I tested Filelight. I found incidentally missing tests in configure due to which are not correctly calculated the size of sparse files (fixed in git hash 1d4d0e56). But I did not notice any crash.

(In reply to Slávek Banko from comment #11)
> I tested Filelight. I found incidentally missing tests in configure due to
> which are not correctly calculated the size of sparse files (fixed in git
> hash 1d4d0e56). But I did not notice any crash.

I still see no crash here with the latest R14 sources; I use Filelight quite a bit and it has never crashed for me.

Darrell, can you post another Valgrind log with the latest GIT sources?  The backtrace in Attachment 1946 [details] is useless as it is impossible for that particular method call sequence to occur unless memory was stomped on somewhere.

Are you testing with i386 or amd64 package sets?

For now, given that the other developers still cannot replicate this problem, I'm removing it from the bug R14 list as it appears to be a corner case on one distribution.  If *anyone* else can replicate this issue please re-add it to the R14 bug list.

Thanks!

I remember these crashes (comment 10). I no longer can reproduce either crash: 1) the crash on Quit or 2) the crash when hovering the mouse pointer across the circular graph outer boundary. Something after April 28 (comment 10) fixed the problem.

Great!  T'm marking this RESOLVED FIXED then.

Thanks for reporting!