Re: [Evolution-hackers] Valgrind and evolution

[Jeffrey Stedfast <fejj novell com>]

On Mon, 2005-09-26 at 08:55 -0600, Veerapuram Varadhan wrote:
> Hi,
> 
> For the past couple of weeks, I am running valgrind on evolution and I
> am using LDTP (http://ldtp.sf.net) to automate test-cases.  LDTP
> requires accessibility to be enabled.  Initially the test runs used to
> take long hours to finish or crash.  With the much improved LDTP, I am
> able to run my test runs without much hassle.  
> 
> I have a PIV, 3.0 GHZ with 4GB RAM as the test machine for this
> memory-leaks hunting task.
> 
> I am using valgrind with following options for my memory-leaks-hunt.
> "valgrind --tool=memcheck --trace-children=yes --track-fds=yes
> --show-reachable=yes --error-limit=no -v --log-file=foo.log
> evolution-2.4"
> 
> and I am running these tests from a custom built gnome-environment.
> (built 2 weeks back)
> 
> During the valgrind run, I found traces like:
> 
> ==27176== 
> ==27176== 71592 (7408 direct, 64184 indirect) bytes in 463 blocks are
> definitely lost in loss record 251 of 274
> ==27176==    at 0x1B8FF8A6: malloc (in
> /usr/lib/valgrind/vgpreload_memcheck.so)
> ==27176==    by 0x1B900BF1: realloc (in
> /usr/lib/valgrind/vgpreload_memcheck.so)
> ==27176==    by 0x1C57C418: g_realloc (gmem.c:170)
> ==27176==    by 0x1C55EBCB: g_array_maybe_expand (garray.c:358)
> ==27176==    by 0x1C55EDC3: g_array_append_vals (garray.c:146)
> ==27176==    by 0x1D099321: gail_tree_view_real_initialize
> (gailtreeview.c:582)
> ==27176==    by 0x1C01CA7B: atk_object_initialize (atkobject.c:1272)
> ==27176==    by 0x1D0993E4: gail_tree_view_new (gailtreeview.c:680)
> ==27176==    by 0x1D072CCD: gail_tree_view_factory_create_accessible
> (gail.c:92)
> ==27176==    by 0x1C01D030: atk_object_factory_create_accessible
> (atkobjectfactory.c:85)
> ==27176==    by 0x1BED2762: gtk_widget_real_get_accessible
> (gtkwidget.c:7577)
> ==27176==    by 0x1BED26C8: gtk_widget_get_accessible (gtkwidget.c:7559)
> ==27176== 
> 
> in which, I am not finding any evolution-related calls.  This is just a
> sample trace of many such traces,
> including some bonobo, gtk, xml parser calls.
> 
> I would like to know 
> 
>   1) How to debug such traces?
>   2) How to report these leaks as bugs and under which module?

you'll first have to figure out where the leak is coming from :)

>   3) Are there any other options to valgrind that could give me a more
> larger stack trace than this?

--num-callers and set a larger value to get more stack info

>   4) Any other valgrind options available to fine-tune leak detections?

not sure if you still need --alignment=8, but you used to.

-- 
Jeffrey Stedfast
Evolution Hacker - Novell, Inc.
fejj ximian com  - www.novell.com