Hey,

Issue #1: I believe we will start working on this in the next sprint (starting 
on 16th). I hope we can get a basic version out quickly, so that some 
adaptation can actually be made in a reasonable way. Fine tuning afterwards.

Issue #3: I'll propose fixing the contextfw sensor structure to the next sprint 
as well.

Issue #4: An init script has been done by another team for their purposes. 
Should be included in the main tree soon.

// Timo 

>-----Original Message-----
>From: Zhang, Xing Z [mailto:[email protected]] 
>Sent: Friday, June 11, 2010 9:25 AM
>To: [email protected]; Rongas Timo
>Cc: [email protected]; [email protected]; 
>[email protected]; [email protected]; 
>[email protected]
>Subject: RE: issue of enabling accelerometer in sensor framework
>
>thank you your quick reply
>
>> 
>> HI Timo:
>>         I am working on enabling our accelerometer in Nokia's sensor
>> framework. Here are some issues i met:
>> 
>> 1. hard coding of adaptor name. accelerometer chain and 
>accelerometer hard
>> codes the adaptor name to
>> "accelerometer". If all adaptors have same name, sensord 
>failed to distinguish
>> right one to load.
>> 
>> You are correct, that's an issue. We are currently trying to 
>figure out a nice
>> solution which would allow sensorfw to always load the 
>correct plugin. There
>> will either be a configuration file providing links between metanames
>> ("accelerometer") and real adaptors ("accelerometer-<devicename>") or
>> somekind of automatic detection of available sensors. In any 
>case, we need to
>> add the use of meta-adaptors/plugins to manage the dependencies.
>> 
>> 2. non-industry data format. For accelerometer, the generic 
>data format is [-Xg,
>> Xg] in float, while Nokia SF
>> uses integer for each axis value. If the framework will 
>persist in integer, there
>> should be a document describes
>> how should another driver map its data. Otherwise 
>accelerometer adopters
>> can't feed their data into systems (e.g filter system)
>
>[Wing] OK, the mg is good. I just thought its hardware 
>specific format before.
>
>> 
>> The hardware drivers we have had in use have provided values 
>as integers, in
>> [-XmG, XmG]. Thus integers were used for sensorfw as well. I 
>have no strong
>> feelings about this though.
>> 
>> 3. context sensor requires too many sensors. Context sensor requires
>> accelerometer, compass, orientation sensor, which
>> means at least needs accelerometer adaptor and magnetic 
>adaptor. For me
>> only have an accelerometer, how can I play
>> with context sensor to provide screen rotation property?
>> 
>> The design of the contextfw sensor should be altered so that 
>it does not require
>> everything it could possibly use to be present. And it 
>definitely should not list
>> itself as a provider for some property it can not calculate. 
>Definitely needs
>> fixing.
>> 
>> In the meantime, it shouldn't be too hard to remove the 
>compass related parts
>> from the context sensor. Comment out the dependency to 
>compassplugin in
>> contextplugin and the construction of CompassBin in 
>ContextSensor. That
>> *should* get it running nicely without having a magnetometer around.
>> 
>> 4. while Nokia SF require root to run, there should be a 
>init.d service file to
>> make it start during system boot. I didn't find it in source.
>> 
>> There used to be one, but it was taken out when it became 
>unnecessary for us.
>> Don't see any harm in having one around though, so if 
>there's a need I guess
>> we could put it back.
>
>[Wing] for others, what time will you expect they are fixed? 
>Especially issue 1,
>it is hard worked around. If I hard code my adaptor in 
>sensord, the N900 adaptor
>is blocked 
>
>> 
>> // Timo
>> 
>> --
>> Timo Rongas <[email protected]>
>
_______________________________________________
MeeGo-dev mailing list
[email protected]
http://lists.meego.com/listinfo/meego-dev

Reply via email to