Re: [Evolution] Proposal regarding apparent race condition on Evolution start-up

[Patrick O'Callaghan <poc usb ve>]

On Tue, 2020-02-11 at 17:31 +0000, Pete Biggs wrote:
> > ksecretservice is definitely not running. In fact I don't think it's
> > even installed on my system. However kwalletd, which is the standard
> > KDE daemon, is running as well as gnome-keyring. If that's the source
> > of the trouble I don't see why killing Evo and restarting it would make
> > a difference.
> kwalletd doesn't provide the DBUS service, it uses its own API, so it
> shouldn't be interfering.
>
> Is there a possibility that kwalletd (or at least KDE) is killing
> gnome-keyring for some other reason? Such as managing encryption keys
> unlocked during login?
>
> But it's all speculation though until you can see exactly what's
> happening!

I did a little more testing. Simply logging out and in again (BTW we
need a word for "logging out and in again" :-) doesn't seem to do it,
the reason being that g-k-r is already running. However, if I log out,
kill g-k-r and log in again, the errors pop up. I get two messages:

1) Failed to connect task list: <account>
The name is not activatable

2) Failed to connect <account>
The reported error was “Failed to authenticate: The name is not
activatable”.

At this point g-k-r is not running.

I kill Evo and restart it. This time g-k-r is running and I get a pop-
up asking for the keyring password (this didn't happen the previous
time the first time), and from then on everything is OK. If I log out
and in again, it still works (g-k-r is kept running even after I log
out).

I originally thought this was random because I often reboot, but I now
think it's deterministic.

The essence of the problem seems to me to be that Evo (or the library
Milan mentioned) expects g-k-r to be running when it starts, but for
some reason doesn't start it correctly on first login.

poc