Philipp von Weitershausen [:philikon] Reporter
	Description • 13 years ago

I've found that the hack in bug 725750 does not always work. More than 50% of the time the phone seems to not find any networks upon the first scan. It must be some sort of race condition.

I'm going to propose that as a temporary hack, we should just re-scan periodically if we can't find the preferred default network right away. Attached is a simple patch that does that. (Sorry, it's not nicely HG formatted, I just threw it together; also it's based on my changes in bug 717122.)

	Chris Jones [:cjones] inactive; ni?/f?/r? if you need me
	Comment 1 • 13 years ago

Comment on attachment 597696 [details] [diff] [review]
v1

This is going to burn battery, including when the device is idle.  Getting the logic for scanning right is going to be a big endeavor though.

I would be happier compromising if the scan interval backs off exponentially, capping at 5-10mins or so.  Then you can start at a lower initial interval too.

	Blake Kaplan (:mrbkap) (inactive)
	Comment 2 • 13 years ago

Comment on attachment 597696 [details] [diff] [review]
v1

Along  what Chris said, I'd like to know why this is necessary. At least for me, I see the first scan race with the driver and fail leading to us not seeing any networks on the first pass but then wpa_supplicant automatically triggers a second scan which does work. Are you not seeing the second scan happen at all?

	Philipp von Weitershausen [:philikon] Reporter
	Comment 3 • 13 years ago

(In reply to Blake Kaplan (:mrbkap) from comment #2)
> Along  what Chris said, I'd like to know why this is necessary. At least for
> me, I see the first scan race with the driver and fail leading to us not
> seeing any networks on the first pass but then wpa_supplicant automatically
> triggers a second scan which does work. Are you not seeing the second scan
> happen at all?

I don't think I was, no. I would see

  I/Gecko   ( 2609): -*- nsWifiWorker component: Available networks: {}

from the debugging line I added in the patch and then it would just not do anything. I agree that it'd be nice to fix the underlying race. And I'm not particularly keen on this approach, either.

Unfortunately, I'm traveling today and tomorrow and won't be near a Mozilla Guest network before Monday, so I won't have the chance to double check.

	Blake Kaplan (:mrbkap) (inactive)
	Comment 4 • 13 years ago

The combination of patches in bug 727680 should fix the race correctly.

	Philipp von Weitershausen [:philikon] Reporter
	Comment 5 • 13 years ago

Sweet!