[Bug 28782] Migrate from dbus-glib to glib's GDBus

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=28782

--- Comment #27 from Simon McVittie <simon.mcvittie at collabora.co.uk> ---
(In reply to comment #25)
> > Thanks for the reminder. I'm vaguely inclined to reinstate turning off
> > exit-on-close to preserve existing behaviour, and so CMs can be valgrinded
> > (valground?).
> 
> Why would valgrind close the connection? I don't understand why it would
> affect it. Note that g_test_dbus_down() will disable exit-on-close before
> stopping the test dbus-daemon.

It's not that; the issue is that if you're using valgrind to spot memory leaks,
it's better to terminate the GMainLoop, unref everything and exit gracefully,
rather than committing suicide with _exit(0) (libdbus) or raise(SIGTERM)
(GDBus) which leaves the last known state of lots of objects as "reachable".

> > On one hand I want to get a minimum viable port of the whole stack working
> > first; on the other hand, getting closer to how it used to work is likely to
> > make that go quicker.
> > [ACTION smcv] reinstate existing behaviour
> 
> Thinking about it, if dbus-daemon crash the CM probably want to first close
> the server connection before leaving. I don't know if our CMs does that, but
> could be safer to reinstate that, yes.

I don't think they do, but the kernel will close the TCP sockets for them,
which should be enough in most protocols.

> Please open bugs for all the "later" points, or keep this bug open to handle
> them, but I'm fine with merging that branch as-is already. Good job!

I'm going to hold off on this until I have at least the CMs, and preferably
Folks, passing tests. My gdbus3 branch has a rebase onto next-tap and next-abi,
plus one more commit, which turned out to be necessary for Rakia.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.