image.bbclass vs core-image.bbclass

image.bbclass vs core-image.bbclass

[Gary Thomas]

Why are some ROOTFS_POSTPROCESS_COMMANDs being set in image.bbclass
and others in core-image.bbclass?  If I build an image using only
image.bbclass, I miss the settings from core-image.bbclass (which
is somewhat misnamed IMO since it's heavier than image.bbclass)?

Is there some reason not to have all of the ROOTFS_POSTPROCESS_COMMANDs
just in image.bbclass alone?

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: image.bbclass vs core-image.bbclass

[Paul Eggleton]

Hi Gary,

On Friday 17 July 2015 04:56:43 Gary Thomas wrote:
> Why are some ROOTFS_POSTPROCESS_COMMANDs being set in image.bbclass
> and others in core-image.bbclass?  If I build an image using only
> image.bbclass, I miss the settings from core-image.bbclass (which
> is somewhat misnamed IMO since it's heavier than image.bbclass)?
> 
> Is there some reason not to have all of the ROOTFS_POSTPROCESS_COMMANDs
> just in image.bbclass alone?

The existence of this class is kind of a legacy from when parts of Poky became 
OE-Core - originally core-image.bbclass was called poky-image.bbclass, and 
what was in it was specific to Poky. We had to bring it over though because all 
of our example images, which we need to have for verification (if nothing 
else), inherited from it and still do. We've made minor adjustments to core-
image.bbclass since then but there are still things in there that are clearly 
"distro" type definitions that don't make sense for everyone; so far nobody has 
really stepped up to find any better common items or reasonable defaults 
(perhaps there aren't any, though I doubt that).

There is a bug open assigned to me to try to sort this out, but to be honest 
I've been struggling with how to best to do it:

  https://bugzilla.yoctoproject.org/show_bug.cgi?id=5424

I'm open to any suggestions, because I do think the dichotomy between these 
classes ought to be resolved if it can be done practically.

Specifically on the ROOTFS_POSTPROCESS_COMMANDs, those do look like they ought 
to somehow be in image.bbclass if they can be added in a manner that doesn't 
interfere with people's ability to create images that aren't rootfses.

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: image.bbclass vs core-image.bbclass

[Gary Thomas]

On 2015-07-17 05:14, Paul Eggleton wrote:
> Hi Gary,
>
> On Friday 17 July 2015 04:56:43 Gary Thomas wrote:
>> Why are some ROOTFS_POSTPROCESS_COMMANDs being set in image.bbclass
>> and others in core-image.bbclass?  If I build an image using only
>> image.bbclass, I miss the settings from core-image.bbclass (which
>> is somewhat misnamed IMO since it's heavier than image.bbclass)?
>>
>> Is there some reason not to have all of the ROOTFS_POSTPROCESS_COMMANDs
>> just in image.bbclass alone?
>
> The existence of this class is kind of a legacy from when parts of Poky became
> OE-Core - originally core-image.bbclass was called poky-image.bbclass, and
> what was in it was specific to Poky. We had to bring it over though because all
> of our example images, which we need to have for verification (if nothing
> else), inherited from it and still do. We've made minor adjustments to core-
> image.bbclass since then but there are still things in there that are clearly
> "distro" type definitions that don't make sense for everyone; so far nobody has
> really stepped up to find any better common items or reasonable defaults
> (perhaps there aren't any, though I doubt that).
>
> There is a bug open assigned to me to try to sort this out, but to be honest
> I've been struggling with how to best to do it:
>
>    https://bugzilla.yoctoproject.org/show_bug.cgi?id=5424
>
> I'm open to any suggestions, because I do think the dichotomy between these
> classes ought to be resolved if it can be done practically.
>
> Specifically on the ROOTFS_POSTPROCESS_COMMANDs, those do look like they ought
> to somehow be in image.bbclass if they can be added in a manner that doesn't
> interfere with people's ability to create images that aren't rootfses.

It seems that many of the ROOTFS_POSTPROCESS_COMMANDs in image.bbclass already
assume that a rootfs is being built.

To me the ROOTFS_POSTPROCESS_COMMANDs that are in core-image.bbclass
don't seem any more invasive than the ones in image.bbclass.  For starters,
I'd like to see them moved to image.bbclass.  It's also quite strange that
the read-only-rootfs hook is defined in image.bbclass but only invoked
from core-image.bbclass??  [That's the one that lead me down this road]

Any objections to a patch that does that?

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: image.bbclass vs core-image.bbclass

[Paul Eggleton]

On Friday 17 July 2015 05:30:12 Gary Thomas wrote:
> On 2015-07-17 05:14, Paul Eggleton wrote:
> > Hi Gary,
> > 
> > On Friday 17 July 2015 04:56:43 Gary Thomas wrote:
> >> Why are some ROOTFS_POSTPROCESS_COMMANDs being set in image.bbclass
> >> and others in core-image.bbclass?  If I build an image using only
> >> image.bbclass, I miss the settings from core-image.bbclass (which
> >> is somewhat misnamed IMO since it's heavier than image.bbclass)?
> >> 
> >> Is there some reason not to have all of the ROOTFS_POSTPROCESS_COMMANDs
> >> just in image.bbclass alone?
> > 
> > The existence of this class is kind of a legacy from when parts of Poky
> > became OE-Core - originally core-image.bbclass was called
> > poky-image.bbclass, and what was in it was specific to Poky. We had to
> > bring it over though because all of our example images, which we need to
> > have for verification (if nothing else), inherited from it and still do.
> > We've made minor adjustments to core- image.bbclass since then but there
> > are still things in there that are clearly "distro" type definitions that
> > don't make sense for everyone; so far nobody has really stepped up to
> > find any better common items or reasonable defaults (perhaps there aren't
> > any, though I doubt that).
> > 
> > There is a bug open assigned to me to try to sort this out, but to be
> > honest> 
> > I've been struggling with how to best to do it:
> >    https://bugzilla.yoctoproject.org/show_bug.cgi?id=5424
> > 
> > I'm open to any suggestions, because I do think the dichotomy between
> > these
> > classes ought to be resolved if it can be done practically.
> > 
> > Specifically on the ROOTFS_POSTPROCESS_COMMANDs, those do look like they
> > ought to somehow be in image.bbclass if they can be added in a manner
> > that doesn't interfere with people's ability to create images that aren't
> > rootfses.
>
> It seems that many of the ROOTFS_POSTPROCESS_COMMANDs in image.bbclass
> already assume that a rootfs is being built.

You could well be right yes.
 
> To me the ROOTFS_POSTPROCESS_COMMANDs that are in core-image.bbclass
> don't seem any more invasive than the ones in image.bbclass.  For starters,
> I'd like to see them moved to image.bbclass.  It's also quite strange that
> the read-only-rootfs hook is defined in image.bbclass but only invoked
> from core-image.bbclass??  [That's the one that lead me down this road]
> 
> Any objections to a patch that does that?

None on my end, certainly...

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre