Hi Holger,

Thanks for the response :)

On Sun, 2007-08-12 at 01:40 +0200, Holger Freyther wrote:
> I don't think this is a recent regression. What I can say is that  
> bb.data.update_data (or whatever this method was called) is a  
> relative cheap operation. So technically there would be nothing  
> preventing us from adding it.

Agreed, its not a new problem but one I did notice something going wrong
with due to the infotroduction of SRCREV and the way it interacted with
exmap-console.

> I'm certain that we will see funny/unexpected things with the OE  
> metadata when calling update_data on immediate assignments, so if you/ 
> we decide to do it I would propose the following strategy.
> 
> Use the existing bitbake and pickle/dump each parsed file, then use  
> the new bitbake and pickle/dump each parsed file as well. Then run a  
> small algorithm (I think I have comitted it to bb.data) to diff two  
> smart_data dictionaries.
> 
> I think it is fine to change that, specially as update_data is quite  
> cheap nowadays, but I would like to see the difference in our metadata.

Agreed. We need to find a way of checking changes like this so perhaps
we should create some kind of semi-automated test for this.

Just for reference, I checked the number of immediate expansions present
in OE. We have 97 of them of which 23 are in conf/class files and 8 of
those are calls to bb.fatal. Of the ones in packages I can see at least
5 which are just plain wrong...

Looking at the list of variables, my gut feeling is we can change this
without breaking anything but I agree we need to check.

Cheers,

Richard

_______________________________________________
Bitbake-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/bitbake-dev

Reply via email to