paroli response to fso crash

Michael 'Mickey' Lauer mickey at openmoko.org
Mon Feb 9 19:25:02 CET 2009


Hi Mirko,

> we have a bit of a discussion on how to handle an fso crash while paroli
>   is running and wanted to put it out there to get your input and ideas.
>
> Quoting:
>  > ticket 13 needs testing (what happens if framework crashes during
>
> paroli use, how can we show it to the user? )

The preferred solution is: Not at all.

In a sane architecture frontends and backends should be restartable -- 
seamlessly and individually. DBus supports us by sending NameOwnerChange 
signals. If Paroli receives such a signal for frameworkd, it should wait for 
frameworkd to join the bus again. Once it's there Paroli should transparently 
reload the state into frameworkd, ideally all that transparent to the user.

> As I wrote in the ticket comments, I think that it is not that easy to
> recover from a unexpected crash from the framework. I though it would
> only be a matter of restarting the framework, but it seems that all the
> DBus connections have to be reopened as well. I am going to see if that
> is really the case.

The Python bindings for DBus support following NameOwnerChanges automatically 
for their interfaces.

-- 
:M:



More information about the devel mailing list