http://bugs.meego.com/show_bug.cgi?id=1007

--- Comment #33 from pohly <[email protected]> 2010-08-26 07:34:42 PDT ---
> Libecal in our "Debian Testing" installation is 2.28.3, which has already been
> replaced by 2.30.2.1-1. I'll update our chroot, perhaps that resolves some of
> these errors.

It also revealed additional problems.

First, valgrind 3.6 (currently in Debian unstable) is needed to avoid reports
about glibc 2.11 (in Debian testing). In installed 3.6 and most of the reports
went away.

Our suppressions also needed to be updated due to different stack back traces.
I started to use "frame wildcards", a new feature added in valgrind 3.4.0:

{
   g_type_init
   Memcheck:Leak
   ...
   fun:g_type_init_with_debug_flags
}

Second, libecal/ebook seem to leak real memory when calling dbus_g_* functions.
This code will be replaced in Evolution 2.32 with the D-Bus code in glib
itself, so I only added very broad suppressions.

Third, libecal/ebook with D-Bus as transport have a problem with apps which
don't call g_main_loop_run() for a loop in the default GMainContext: the D-Bus
backend terminates after 10 seconds of idle time. The corresponding "destroyed"
signal is not delivered without the main loop, so libecal (and probably
libebook) continue using the dead instance of the backend, leading to:
libecal-WARNING **: Cannot get cal from factory: The name :1.2 was not provided
by any .service files

I worked around this in our nightly testing by selecting Client::Source tests
manually and reordering them so that ical20/itodo20/text are run directly one
after the other, thus avoiding the 10 second idle time (hopefully...). As a
side effect, we no longer run kind of redundant tests (file_ical20 is used, but
not file_vcard30).

This worked for Client::Source, but not Client::Sync. Need a better solution
for that. Restart client-test for each test? Ping backends regularly while it
runs?

------------------

I noticed that "prebuilt" tests (= compiled on Hardy, run inside Debian
Testing) does not use valgrind. That would be useful (covers real scenario) and
should succeed just as well as running the natively compiled binaries. Will add
it.

-- 
Configure bugmail: http://bugs.meego.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching someone on the CC list of the bug.
_______________________________________________
Syncevolution-issues mailing list
[email protected]
http://lists.syncevolution.org/listinfo/syncevolution-issues

Reply via email to