Re: BSP Packagegroup

[Ann Thornton <Ann.Thornton@freescale.com>]

[-- Attachment #1: Type: text/plain, Size: 4088 bytes --]


I am thinking of packagegroups as a way to make life easier. Graphics in 
particular is complicated, with some things working in one place and not 
another.  Graphics packagegroups can handle a lot of that so a common 
image recipe can easily work in multiple environments and new recipes 
can be created more easily without having to know every detail.  Having 
different levels allows an easy way to choose how much to include in a 
particular image. Packagegroups are not specific to BSP or SDK or 
whatever.  They can be used wherever desired.

Ann

On 7/11/2015 9:24 AM, Daiane Angolini wrote:
> On Thu, Jul 9, 2015 at 1:29 PM, Ann Thornton <Ann.Thornton@freescale.com> wrote:
>> For packagegroup divisions, how about
>> graphics
>> multimedia
>> networking
>> tools
>> (probably others I haven't thought of)
> When we think about a set of packagegroups intended to give users a
> set of useful packages or an example on how to group applications more
> is more. As much as possible is good enough. (meta-fsl-demos)
>
> However, when you think about a BSP packagegroup, less is more. As
> less as possible, as much closer to only critical packages, the
> minimal set, is the optimal packagegroup. (meta-fsl-arm)
>
> Do we need BSP packagegroups?
>
> In case we do, I cannot see the need of a BSP packagegroup for graphic
> (GPU) or tools.
>
> I would accept "tools" to be split into other sets like "audio", but
> definitively not "tools".
>
> In fact, I can only see the need of a VPU and a CAN package groups.
> Maybe audio. But I'm trying to stress the BSP packagegroup idea here,
> something I'm not completely convinced of.
>
>
> Daiane
>> Each of those groups might be further divided into minimal, core, demos,
>> extended as needed.
>>
>> Each packagegroup could check DISTRO-FEATURES, etc so that they would be as
>> generic as possible.
>>
>> Then recipes could include the level of detail desired and they would work
>> across product lines.
>>
>> Ann Thornton
>>
>>
>> On 7/9/2015 9:56 AM, Daiane Angolini wrote:
>>
>> For me, packagegroup is only a set of packages wrapped together to
>> make my life easier.
>>
>> Should BSP provide packagegroups to ease the addition (and removal) of
>> set o BSP packages, or their “functional” dependency. For example an
>> application such as aplay is needed to stress the audio functionality,
>> even though there is no dependency from alsa driver from kernel with
>> alsa-utils. Should BSP provide packagegroups?
>>
>> I think offering packagegroup options to enable BSP pieces may really
>> ease the BSP usage, however I main point here is how far should BSP
>> go. What is the limit between a BSP packagegroup and a "demo"
>> packagegroup (as we does in meta-fsl-demos)?
>>
>> Thinking about a package group to provide BSP packages related with
>> VPU, in my opinion it should have:
>>
>> * VPU firmware
>> * VPU lib
>>
>> In case I’m using gstreamer, I would like a packagegroup like:
>>
>> * VPU firmware
>> * VPU lib
>> * gstreamer plugins for VPU (gstreamer-imx or gst1.0-fsl-plugin)
>>
>> In case I’m using gstreamer with kernel mainline:
>>
>> * VPU firmware
>> * gstreamer
>>
>>
>> Should mp3 encoder (such as lame) be part of a BSP packagegroup? And
>> in GPU case? Would DEPENDS and PROVIDES be enough to include needed
>> packages?
>>
>> Should meta-fsl-arm (or meta-freescale) provide a bluetooth BSP
>> packagegroup even though there is no special hardware acceleration
>> provided by meta-fsl-arm for bluetooth?
>>
>>
>> Daiane
>>
>>
>>
>> --
>> Ann Thornton
>>
>> Microcontrollers Software and Applications
>> Freescale Semiconductors
>> email: Ann.Thornton@freescale.com
>>
>> --
>> _______________________________________________
>> meta-freescale mailing list
>> meta-freescale@yoctoproject.org
>> https://lists.yoctoproject.org/listinfo/meta-freescale
>>


-- 
Ann Thornton

/Microcontrollers Software and Applications
Freescale Semiconductors
email: Ann.Thornton@freescale.com/

[-- Attachment #2: Type: text/html, Size: 5063 bytes --]