Hi Marcel, On 19.04.2012 12:36, Marcel Holtmann wrote: >>>> This fixes a small regression with the AutoConnect changes on ethernet. >>>> If the user has used ConnMan in the past and updates to the recent version >>>> the settings files wont contain the AutoConnect key value pair. Then >>>> we do not enable the auto connect. >>>> --- >>>> src/service.c | 2 ++ >>>> 1 file changed, 2 insertions(+) >>>> >>>> diff --git a/src/service.c b/src/service.c >>>> index 8f97c09..ed64133 100644 >>>> --- a/src/service.c >>>> +++ b/src/service.c >>>> @@ -407,6 +407,8 @@ static int service_load(struct connman_service >>>> *service) >>>> service->identifier, "AutoConnect", &error); >>>> if (error == NULL) >>>> service->autoconnect = autoconnect; >>>> + else >>>> + service->autoconnect = TRUE; >>> >>> shouldn't this be using the default value or the settings from main.conf >>> depending on the technology? >> >> service_load() is called after the update from the main.conf. So it >> overwrites the main.conf settings. Sounds reasonable for me, that the >> per service settings have higher precedence then the main.conf ones. > > that is correct, but you are currently overwriting it with TRUE if it is > not found in the storage file. > > Are you actually sure that your fix is doing the right thing. If by > accident the storage has AutoConnect=false, we can do nothing to fix > this here. Bad luck.
First, this patch should actually only be needed for those, who have used ConnMan in the past and upgrade now. When service_save() is called the AutoConnect key will be written. A missing key should only happen once, when ConnMan loads the setting file after the upgrade. In this case, the key is missing and we should get an error != NULL, right? In all other cases the key will be present. At least that is what I think will happen. Not 100% sure. If we really want to be on the safe side, we just document it as known upgrade issue. What do you think? cheers, daniel _______________________________________________ connman mailing list [email protected] http://lists.connman.net/listinfo/connman
