[patch] Distinguish between OOM and Key error

[John (J5) Palmieri]

On Fri, 2006-09-08 at 16:25 -0400, Havoc Pennington wrote:
> John (J5) Palmieri wrote:
> > diff -u -r1.45 activation.c
> > --- bus/activation.c	6 Sep 2006 00:14:06 -0000	1.45
> > +++ bus/activation.c	8 Sep 2006 18:05:50 -0000
> > @@ -285,7 +285,8 @@
> >    if (!bus_desktop_file_get_string (desktop_file,
> >                                      DBUS_SERVICE_SECTION,
> >                                      DBUS_SERVICE_NAME,
> > -                                    &name))
> > +                                    &name,
> > +                                    error))
> >      {
> >        dbus_set_error (error, DBUS_ERROR_FAILED,
> >                        "No \""DBUS_SERVICE_NAME"\" key in .service file\n");
> > @@ -295,7 +296,8 @@
> >    if (!bus_desktop_file_get_string (desktop_file,
> >                                      DBUS_SERVICE_SECTION,
> >                                      DBUS_SERVICE_EXEC,
> > -                                    &exec))
> > +                                    &exec,
> > +                                    error))
> >      {
> >        dbus_set_error (error, DBUS_ERROR_FAILED,
> >                        "No \""DBUS_SERVICE_EXEC"\" key in .service file\n");
> 
> This looks like it sets a new FAILED error over the top of the error 
> returned from bus_desktop_file_get_string, should just leave the error 
> set by the invoked function (should _DBUS_ASSERT_ERROR_IS_SET I think)

Ya, I just missed taking out the dbus_set_error there. 

> Then, should some code be checking whether the error is an OOM error or 
> not? In principle on OOM the .service file should be retried later, 
> while for a bad .service file it should never be retried.

Should we have a wait for memory call and then retry?  Do we do this
indefinitely?

-- 
John (J5) Palmieri <johnp at redhat.com>