7:22 p.m.
On Do, 2011-07-14 at 04:24 +0200, Chris Kühl wrote:
I've been working on getting my dbus-server-reorganization branch
on
par with master and I think I've almost achieved that. Currently they
are both seeing 3 failures and 2 errors. I see that the 2 errors are
probably addressed in bug #20990 as they start to appear only after
the recent autosync changes. The 3 failures seem to be due to issues
with the password property in temporary configs. I've pastebin'ed my
output in case it's helpful: http://pastebin.com/kRe0sPBK
'Method "GetReports" ... doesn't exist' is (partly?) fixed by:
commit f47d31580ddafa4a0243e809aa0ec5451e5182d3
Author: Patrick Ohly <patrick.ohly(a)intel.com>
Date: Tue Jul 12 12:46:47 2011 +0200
auto syncing: fix D-Bus API violations (part of BMC #20966)
Auto-sync sessions did not properly activate their D-Bus support and
thus couldn't be accessed via the Session D-Bus API. Must have
affected showing progress of such sessions in the GTK sync-ui.
They also weren't kept around for one minute, like the sessions
started by a client. Therefore UIs which need to retrieve information
about a completed session failed for a second reason.
Fixed by adding the necessary "activate()" and use "addTimeout()"
trick for session expiry also in the AutoSyncManager. The later was
moved into DBusServer for that.
I'm not sure whether the fix is complete, because I still see the
failure in the nightly testing, in a slightly different way than locally
(where I cannot reproduce it anymore):
http://runtests.syncevolution.org/2011-07-13-19-56_dist/head-lucid-amd64/...
http://runtests.syncevolution.org/2011-07-13-19-56_dist/head-lucid-amd64/...
[DEBUG 00:00:01] session /org/syncevolution/Session/17312472401310560063
deconstructing
process 28897: arguments to dbus_connection_get_data() were incorrect, assertion
"connection != NULL" failed in file dbus-connection.c line 5797.
This is normally a bug in some application using the D-Bus library.
Adding the auto sync tests broke some other tests, which I only noticed
in that nightly testing. Fixed by:
commit 811d10fc33f64a46ae43ce3903d637cad02c15f0
Author: Patrick Ohly <patrick.ohly(a)intel.com>
Date: Wed Jul 13 15:54:49 2011 +0200
D-Bus testing: avoid test failures due to GNOME keyring side effects
The auto-sync tests modified the GNOME keyring by setting a password
for the same URL that is also used in the password request tests,
which caused those tests to fail.
Ideally this would be fixed by running each test with a fresh GNOME
keyring, but this is hard to set up. Instead solve the problem by
keeping the URLs separate.
I said "almost" on par because I've been having issues
when building
with --disable-shared and --enable-static. When trying this I get an
error telling me to add -ldl to the LDFLAGS which I've done and it
does in fact seem to build correctly. However, with this, the backends
don't get registered and I, of course, get tons of problems when
running the tests. This issue exists in the first commit on my branch
which moved the dbus-server over to its own directory. However,
everything does seem to build and work fine with only --enable-static
specified which is what I've been using for running the tests.
Did you notice the rules which add the backend's *Register.cpp files to
the syncevo-dbus-server?
src/Makefile.am (generated from Makefile-gen.am) contains:
BACKEND_REGISTRIES = ... backends/.../*Register.cpp
This is then added to syncevo-dbus-server via:
CORE_SOURCES += $(BACKEND_REGISTRIES)
syncevo_dbus_server_SOURCES = \
...
$(CORE_SOURCES)
The easiest solution might be to keep linking the main
syncevo-dbus-server in "src", but with all code compiled in a
sub-directory into a helper library.
--
Best Regards, Patrick Ohly
The content of this message is my personal opinion only and although
I am an employee of Intel, the statements I make here in no way
represent Intel's position on the issue, nor am I authorized to speak
on behalf of Intel on this matter.