Proposed roadmap and design for Hob improvement

Proposed roadmap and design for Hob improvement

[Yu, Ke]

[-- Attachment #1: Type: text/plain, Size: 289 bytes --]

Hi,

As the Yocto 1.2. plan is starting (https://wiki.yoctoproject.org/wiki/Yocto_1.2_Features ), here I draft a proposal for the roadmap and design on HOB improvement task (http://bugzilla.yoctoproject.org/show_bug.cgi?id=1588) . Please review. Any feedback is welcome.

Regards
Ke

[-- Attachment #2: HOBV2-plan-0.2.pdf --]
[-- Type: application/pdf, Size: 1113822 bytes --]

Re: Proposed roadmap and design for Hob improvement

[Koen Kooi]

Op 21 okt. 2011, om 05:16 heeft Yu, Ke het volgende geschreven:

> Hi,
> 
> As the Yocto 1.2. plan is starting (https://wiki.yoctoproject.org/wiki/Yocto_1.2_Features ), here I draft a proposal for the roadmap and design on HOB improvement task (http://bugzilla.yoctoproject.org/show_bug.cgi?id=1588) . Please review. Any feedback is welcome.


The first slide says:

"It’s not an embedded Linux distribution –It creates a custom one for you."

Which is wrong, since hob is creating images, not distros

Re: Proposed roadmap and design for Hob improvement

[Yu, Ke]

> -----Original Message-----
> From: Koen Kooi [mailto:koen@dominion.thruhere.net]
> Sent: Friday, October 21, 2011 11:54 PM
> To: Yu, Ke
> Cc: Yocto Project Discussion
> Subject: Re: [yocto] Proposed roadmap and design for Hob improvement
> 
> 
> Op 21 okt. 2011, om 05:16 heeft Yu, Ke het volgende geschreven:
> 
> > Hi,
> >
> > As the Yocto 1.2. plan is starting
> (https://wiki.yoctoproject.org/wiki/Yocto_1.2_Features ), here I draft a
> proposal for the roadmap and design on HOB improvement task
> (http://bugzilla.yoctoproject.org/show_bug.cgi?id=1588) . Please review. Any
> feedback is welcome.
> 
> 
> The first slide says:
> 
> "It's not an embedded Linux distribution -It creates a custom one for you."
> 
> Which is wrong, since hob is creating images, not distros

Oh, this slogan is for Yocto project,  not for HOB, meaning that Yocto is not a distro, Yocto create custom distro. 

Regards
Ke

Re: Proposed roadmap and design for Hob improvement

[McClintock Matthew-B29882]

On Thu, Oct 20, 2011 at 10:16 PM, Yu, Ke <ke.yu@intel.com> wrote:
> Hi,
>
> As the Yocto 1.2. plan is starting (https://wiki.yoctoproject.org/wiki/Yocto_1.2_Features ), here I draft a proposal for the roadmap and design on HOB improvement task (http://bugzilla.yoctoproject.org/show_bug.cgi?id=1588) . Please review. Any feedback is welcome.

Seems like with the build packages and build image steps separated you
could skip the build step completly for most packages if we ship or
host pre-built build packages

-M

Re: Proposed roadmap and design for Hob improvement

[Yu, Ke]

> -----Original Message-----
> From: McClintock Matthew-B29882 [mailto:B29882@freescale.com]
> Sent: Saturday, October 22, 2011 5:14 AM
> To: Yu, Ke
> Cc: Yocto Project Discussion
> Subject: Re: [yocto] Proposed roadmap and design for Hob improvement
> 
> On Thu, Oct 20, 2011 at 10:16 PM, Yu, Ke <ke.yu@intel.com> wrote:
> > Hi,
> >
> > As the Yocto 1.2. plan is starting
> (https://wiki.yoctoproject.org/wiki/Yocto_1.2_Features ), here I draft a
> proposal for the roadmap and design on HOB improvement task
> (http://bugzilla.yoctoproject.org/show_bug.cgi?id=1588) . Please review. Any
> feedback is welcome.
> 
> Seems like with the build packages and build image steps separated you could
> skip the build step completly for most packages if we ship or host pre-built build
> packages
> 
> -M

Yes, current design supports this scenario, i.e. if some packages are already pre-built, HOB will notice that and won't build them again.

Regards
Ke

Re: Proposed roadmap and design for Hob improvement

[McClintock Matthew-B29882]

On Sun, Oct 23, 2011 at 10:12 PM, Yu, Ke <ke.yu@intel.com> wrote:
>> Seems like with the build packages and build image steps separated you could
>> skip the build step completly for most packages if we ship or host pre-built build
>> packages
>>
>> -M
>
> Yes, current design supports this scenario, i.e. if some packages are already pre-built, HOB will notice that and won't build them again.

It would be nice to be able to skip the package generation if the RPM
already exists somewhere. This could provide significant speedup above
what is already provided by sstate-cache. Although other issues might
need to be solved....

-M

Re: Proposed roadmap and design for Hob improvement

[Mark Hatle]

On 10/25/11 11:08 AM, McClintock Matthew-B29882 wrote:
> On Sun, Oct 23, 2011 at 10:12 PM, Yu, Ke <ke.yu@intel.com> wrote:
>>> Seems like with the build packages and build image steps separated you could
>>> skip the build step completly for most packages if we ship or host pre-built build
>>> packages
>>>
>>> -M
>>
>> Yes, current design supports this scenario, i.e. if some packages are already pre-built, HOB will notice that and won't build them again.
> 
> It would be nice to be able to skip the package generation if the RPM
> already exists somewhere. This could provide significant speedup above
> what is already provided by sstate-cache. Although other issues might
> need to be solved....

sstate-cache already caches the results of the package builds (not just RPM).
This is needed to be able to populate both the sysroot and the package set for
the image generation step.

To me what is a bigger deal is to break up the image generation into an external
step.. so I can simply start with a set of packages and build them.  (since
sysroots and similar are not needed in this situation.)

--Mark

> -M
> _______________________________________________
> yocto mailing list
> yocto@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/yocto

Re: Proposed roadmap and design for Hob improvement

[Joshua Lock]

On 25/10/11 09:08, McClintock Matthew-B29882 wrote:
> On Sun, Oct 23, 2011 at 10:12 PM, Yu, Ke <ke.yu@intel.com> wrote:
>>> Seems like with the build packages and build image steps separated you could
>>> skip the build step completly for most packages if we ship or host pre-built build
>>> packages
>>>
>>> -M
>>
>> Yes, current design supports this scenario, i.e. if some packages are already pre-built, HOB will notice that and won't build them again.
> 
> It would be nice to be able to skip the package generation if the RPM
> already exists somewhere. This could provide significant speedup above
> what is already provided by sstate-cache. Although other issues might
> need to be solved....

I think this falls in line with the proposed design. If we create the
images from package feeds we needn't build when the package feeds are
already available.

Of course with ambiguous terms like *somewhere* I'm not sure we're
thinking along exactly the same lines but we're definitely looking to
speed up and re-use in as many places as possible.

Cheers,
Joshua
-- 
Joshua Lock
        Yocto Project "Johannes factotum"
        Intel Open Source Technology Centre

Re: Proposed roadmap and design for Hob improvement

[McClintock Matthew-B29882]

On Tue, Oct 25, 2011 at 1:09 PM, Joshua Lock <josh@linux.intel.com> wrote:
> Of course with ambiguous terms like *somewhere* I'm not sure we're
> thinking along exactly the same lines but we're definitely looking to
> speed up and re-use in as many places as possible.

Well, somewhere is 'tmp/deploy/rpms' or $BB_RPM_MIRROR or RPMs from
sstate-cache.

-M