On Tue, Aug 30, 2011 at 12:47 PM, Koen Kooi <[email protected]> wrote: > > Op 30 aug. 2011, om 18:46 heeft Jason Kridner het volgende geschreven: > >> On Mon, Aug 29, 2011 at 7:08 PM, Joel A Fernandes <[email protected]> >> wrote: >>> On Mon, Aug 29, 2011 at 4:03 PM, Jason Kridner <[email protected]> >>> wrote: >>>>>>> >>>>>>> This script is BSP specific and shouldn't live in the OE-core layer. >>>>>> >>>>>> The only issue is this script is used from within the IMAGE_CMD_sdimg >>>>>> code which lives in OE-core (meta/classes/image_types. >>>>> >>>>> classes shouldn't be calling external scripts >>>>> >>>> >>>> Is the right approach to add parameters to the IMAGE_CMD_sdimg class >>>> such that it can be used generically to produce SD card images, >>>> instead of trying to move this to meta-ti? Should it perhaps be a bit >>>> closer to what is being done by the Linaro image tools [1]? >>>> >>>> [1] https://wiki.linaro.org/Source/ImageBuilding >>>> >>> >>> Don't the Linaro scripts need to run on the target? >> >> No. >> >>> Does it fit well >>> with OE(-core) ? >> >> I don't know, but I hoped that others would comment to help figure it >> out. I think it is a common challenge not just for ARM architectures. >> I'm not sure if the approach is general enough or not, but it does go >> beyond the BeagleBoard. >> >> My primary concerns about leaving it in the BSP are: >> 1) that there is some room for non-BeagleBoard specific optimizations >> to the card layout and >> 2) there may be improvements to the tools that make it easier to >> create images for systems with different boot requirements. >> >> Also, I think we might want to move to an ext3 partition only in the >> future or other such layout optimizations. I'd like that to be >> something that can be parameterized by the BSP. > > I think you're missing the point: > > The *script* needs to go into the BSP, you are free to extend the bbclasses > with code to deal with sd cards internally in OE-core. > >
I think there are 2 threads to this discussion. One talks about mkcard script (which I think Koen is asking to move to BSP), and the other is the IMAGE_CMD_sdimg variable from image_types.bbclass which holds the code to create the image (which Jason is referring to). I am dropping mkcard script as it requires root access (due to kparted and device mapper), instead I'm just using loop and avoiding device mapper. As for the IMAGE_CMD_sdimg , I'm thinking a clean way is to extend image_types.bbclass from within the BSP layer (which I think Koen also is suggesting). Maybe I'm just summarizing what everyone is saying, but just making sure everyone is on the same page. Thanks, Joel _______________________________________________ Openembedded-core mailing list [email protected] http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
