[PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Koen Kooi]

* Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
  setting for a recipe.
* This will allow recipes to work for entire families of
  devices without having to maintain/update the compatible
  devices as new devices are added into a family

Based on 07076390358f211bd96779bec2d6eb5eaa0ad699 by Chase Maupin <chase.maupin@ti.com>

Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
---
 meta/classes/base.bbclass |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/meta/classes/base.bbclass b/meta/classes/base.bbclass
index a674f52..aaf99da 100644
--- a/meta/classes/base.bbclass
+++ b/meta/classes/base.bbclass
@@ -342,7 +342,9 @@ python () {
             import re
             this_machine = bb.data.getVar('MACHINE', d, 1)
             if this_machine and not re.match(need_machine, this_machine):
-                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
+                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
+                if (this_soc_family and not re.match(need_machine, this_soc_family)) or not this_soc_family:
+                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
 
 
         dont_want_license = bb.data.getVar('INCOMPATIBLE_LICENSE', d, 1)
-- 
1.6.6.1

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Koen Kooi]

ping

Op 28 feb 2011, om 20:21 heeft Koen Kooi het volgende geschreven:

> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>  setting for a recipe.
> * This will allow recipes to work for entire families of
>  devices without having to maintain/update the compatible
>  devices as new devices are added into a family
> 
> Based on 07076390358f211bd96779bec2d6eb5eaa0ad699 by Chase Maupin <chase.maupin@ti.com>
> 
> Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
> ---
> meta/classes/base.bbclass |    4 +++-
> 1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/meta/classes/base.bbclass b/meta/classes/base.bbclass
> index a674f52..aaf99da 100644
> --- a/meta/classes/base.bbclass
> +++ b/meta/classes/base.bbclass
> @@ -342,7 +342,9 @@ python () {
>             import re
>             this_machine = bb.data.getVar('MACHINE', d, 1)
>             if this_machine and not re.match(need_machine, this_machine):
> -                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
> +                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
> +                if (this_soc_family and not re.match(need_machine, this_soc_family)) or not this_soc_family:
> +                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
> 
> 
>         dont_want_license = bb.data.getVar('INCOMPATIBLE_LICENSE', d, 1)
> -- 
> 1.6.6.1
> 

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Mark Hatle]

On 3/1/11 1:55 PM, Koen Kooi wrote:
> ping

Do you have an example showing how this is to be used?

--Mark

> Op 28 feb 2011, om 20:21 heeft Koen Kooi het volgende geschreven:
> 
>> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>>  setting for a recipe.
>> * This will allow recipes to work for entire families of
>>  devices without having to maintain/update the compatible
>>  devices as new devices are added into a family
>>
>> Based on 07076390358f211bd96779bec2d6eb5eaa0ad699 by Chase Maupin <chase.maupin@ti.com>
>>
>> Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
>> ---
>> meta/classes/base.bbclass |    4 +++-
>> 1 files changed, 3 insertions(+), 1 deletions(-)
>>
>> diff --git a/meta/classes/base.bbclass b/meta/classes/base.bbclass
>> index a674f52..aaf99da 100644
>> --- a/meta/classes/base.bbclass
>> +++ b/meta/classes/base.bbclass
>> @@ -342,7 +342,9 @@ python () {
>>             import re
>>             this_machine = bb.data.getVar('MACHINE', d, 1)
>>             if this_machine and not re.match(need_machine, this_machine):
>> -                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
>> +                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
>> +                if (this_soc_family and not re.match(need_machine, this_soc_family)) or not this_soc_family:
>> +                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
>>
>>
>>         dont_want_license = bb.data.getVar('INCOMPATIBLE_LICENSE', d, 1)
>> -- 
>> 1.6.6.1
>>
> 
> 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Tom Rini]

On 03/01/2011 01:26 PM, Mark Hatle wrote:
> On 3/1/11 1:55 PM, Koen Kooi wrote:
>> ping
>
> Do you have an example showing how this is to be used?

For example:
http://cgit.openembedded.org/cgit.cgi/openembedded/tree/conf/machine/include/omap3.inc
http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/linux/linux-omap3.inc

-- 
Tom Rini
Mentor Graphics Corporation

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Mark Hatle]

On 3/1/11 6:14 PM, Tom Rini wrote:
> On 03/01/2011 01:26 PM, Mark Hatle wrote:
>> On 3/1/11 1:55 PM, Koen Kooi wrote:
>>> ping
>>
>> Do you have an example showing how this is to be used?
> 
> For example:
> http://cgit.openembedded.org/cgit.cgi/openembedded/tree/conf/machine/include/omap3.inc
> http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/linux/linux-omap3.inc
> 

Thanks, that makes sense to me.  I think it's useful.  (It's already in, no
objections to that)

--Mark

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Khem Raj]

On Tue, Mar 1, 2011 at 11:55 AM, Koen Kooi <koen@dominion.thruhere.net> wrote:
> ping
>

I have applied this sorry for delay

-Khem

[PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Koen Kooi]

* Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
  setting for a recipe.
* This will allow recipes to work for entire families of
  devices without having to maintain/update the compatible
  devices as new devices are added into a family

Based on 07076390358f211bd96779bec2d6eb5eaa0ad699 by Chase Maupin <chase.maupin@ti.com>

Signed-off-by: Koen Kooi <koen@dominion.thruhere.net>
---
 meta/classes/base.bbclass |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/meta/classes/base.bbclass b/meta/classes/base.bbclass
index a674f52..aaf99da 100644
--- a/meta/classes/base.bbclass
+++ b/meta/classes/base.bbclass
@@ -342,7 +342,9 @@ python () {
             import re
             this_machine = bb.data.getVar('MACHINE', d, 1)
             if this_machine and not re.match(need_machine, this_machine):
-                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
+                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
+                if (this_soc_family and not re.match(need_machine, this_soc_family)) or not this_soc_family:
+                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
 
 
         dont_want_license = bb.data.getVar('INCOMPATIBLE_LICENSE', d, 1)
-- 
1.6.6.1

[PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Chase Maupin]

* Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
  setting for a recipe.
* This will allow recipes to work for entire families of
  devices without having to maintain/update the compatible
  devices as new devices are added into a family

Signed-off-by: Chase Maupin <chase.maupin@ti.com>
---
 classes/base.bbclass |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/classes/base.bbclass b/classes/base.bbclass
index 1847491..02c7875 100644
--- a/classes/base.bbclass
+++ b/classes/base.bbclass
@@ -389,7 +389,9 @@ python () {
             import re
             this_machine = bb.data.getVar('MACHINE', d, 1)
             if this_machine and not re.match(need_machine, this_machine):
-                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
+                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
+                if this_soc_family and not re.match(need_machine, this_soc_family):
+                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
 
         need_target = bb.data.getVar('COMPATIBLE_TARGET_SYS', d, 1)
         if need_target:
-- 
1.7.0.4

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Denys Dmytriyenko]

On Tue, Aug 03, 2010 at 04:47:20PM -0500, Chase Maupin wrote:
> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>   setting for a recipe.
> * This will allow recipes to work for entire families of
>   devices without having to maintain/update the compatible
>   devices as new devices are added into a family
> 
> Signed-off-by: Chase Maupin <chase.maupin@ti.com>

Acked-by: Denys Dmytriyenko <denys@ti.com>

While SOC_FAMILY feature is relatively new on its own, it was designed and now 
being extensively used with TI machines and corresponding recipes as an 
effective override - e.g. covering many OMAP3-based machines with a single 
VAR_omap3 = "VAL" expression.

> ---
>  classes/base.bbclass |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/classes/base.bbclass b/classes/base.bbclass
> index 1847491..02c7875 100644
> --- a/classes/base.bbclass
> +++ b/classes/base.bbclass
> @@ -389,7 +389,9 @@ python () {
>              import re
>              this_machine = bb.data.getVar('MACHINE', d, 1)
>              if this_machine and not re.match(need_machine, this_machine):
> -                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
> +                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
> +                if this_soc_family and not re.match(need_machine, this_soc_family):
> +                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
>  
>          need_target = bb.data.getVar('COMPATIBLE_TARGET_SYS', d, 1)
>          if need_target:
> -- 
> 1.7.0.4
> 
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Khem Raj]

On (03/08/10 20:19), Denys Dmytriyenko wrote:
> On Tue, Aug 03, 2010 at 04:47:20PM -0500, Chase Maupin wrote:
> > * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
> >   setting for a recipe.
> > * This will allow recipes to work for entire families of
> >   devices without having to maintain/update the compatible
> >   devices as new devices are added into a family
> > 
> > Signed-off-by: Chase Maupin <chase.maupin@ti.com>
> 
> Acked-by: Denys Dmytriyenko <denys@ti.com>
> 
> While SOC_FAMILY feature is relatively new on its own, it was designed and now 
> being extensively used with TI machines and corresponding recipes as an 
> effective override - e.g. covering many OMAP3-based machines with a single 
> VAR_omap3 = "VAL" expression.

So arch -> sub-arch -> soc ->machine seems to be the heirarchy. I
would prefer to keep it per machine unless its too much clutter
and we have so many machines specific recipes.

> 
> > ---
> >  classes/base.bbclass |    4 +++-
> >  1 files changed, 3 insertions(+), 1 deletions(-)
> > 
> > diff --git a/classes/base.bbclass b/classes/base.bbclass
> > index 1847491..02c7875 100644
> > --- a/classes/base.bbclass
> > +++ b/classes/base.bbclass
> > @@ -389,7 +389,9 @@ python () {
> >              import re
> >              this_machine = bb.data.getVar('MACHINE', d, 1)
> >              if this_machine and not re.match(need_machine, this_machine):
> > -                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
> > +                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
> > +                if this_soc_family and not re.match(need_machine, this_soc_family):
> > +                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
> >  
> >          need_target = bb.data.getVar('COMPATIBLE_TARGET_SYS', d, 1)
> >          if need_target:
> > -- 
> > 1.7.0.4
> > 
> > 
> > _______________________________________________
> > Openembedded-devel mailing list
> > Openembedded-devel@lists.openembedded.org
> > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Michael 'Mickey' Lauer]

FWIW, minimal is using MACHINE_CLASS since quite a while to
reduce the need for adding the same files over and over again,
this is being used e.g. for HTC msm7 series and OpenEZX series
(see conf/machine/include).

Perhaps it's time to standardize something like that.

Cheers,
-- 
:M:

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Phil Blundell]

On Wed, 2010-08-04 at 13:17 +0200, Michael 'Mickey' Lauer wrote:
> FWIW, minimal is using MACHINE_CLASS since quite a while to
> reduce the need for adding the same files over and over again,
> this is being used e.g. for HTC msm7 series and OpenEZX series
> (see conf/machine/include).
> 
> Perhaps it's time to standardize something like that.

Yeah.  I certainly don't think we want a proliferation of such
mechanisms.  If minimal is already using MACHINE_CLASS then there would
be some sense in trying to make use of the same thing.

My preference would be to make the MACHINEs in question be, for example,
"motorola-ezx-a1200", "motorola-ezx-a780" and so on.  This would then
allow you to set COMPATIBLE_MACHINE = "motorola-ezx-.*" which I think
would solve the SOC_FAMILY problem, although it wouldn't replace
MACHINE_CLASS in OVERRIDES.

Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
probably is more desirable than adding a completely new variable (i.e.
SOC_FAMILY) for that purpose.

p.

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Maupin, Chase]

> -----Original Message-----
> From: openembedded-devel-bounces@lists.openembedded.org
> [mailto:openembedded-devel-bounces@lists.openembedded.org] On Behalf Of
> Phil Blundell
> Sent: Wednesday, August 04, 2010 6:49 AM
> To: openembedded-devel@lists.openembedded.org
> Subject: Re: [oe] [PATCH] base.bbclass: add support for SOC_FAMILY in
> COMPATIBLE_MACHINES
> 
> On Wed, 2010-08-04 at 13:17 +0200, Michael 'Mickey' Lauer wrote:
> > FWIW, minimal is using MACHINE_CLASS since quite a while to
> > reduce the need for adding the same files over and over again,
> > this is being used e.g. for HTC msm7 series and OpenEZX series
> > (see conf/machine/include).
> >
> > Perhaps it's time to standardize something like that.
> 
> Yeah.  I certainly don't think we want a proliferation of such
> mechanisms.  If minimal is already using MACHINE_CLASS then there would
> be some sense in trying to make use of the same thing.

minimal doesn't use MACHINE_CLASS.  Perhaps this is getting confused with distro/micro.conf.

> 
> My preference would be to make the MACHINEs in question be, for example,
> "motorola-ezx-a1200", "motorola-ezx-a780" and so on.  This would then
> allow you to set COMPATIBLE_MACHINE = "motorola-ezx-.*" which I think
> would solve the SOC_FAMILY problem, although it wouldn't replace
> MACHINE_CLASS in OVERRIDES.

Unfortunately we can't change the MACHINE names.  They are set to correspond to the real device names and part numbers.

> 
> Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
> probably is more desirable than adding a completely new variable (i.e.
> SOC_FAMILY) for that purpose.

From what I've learned it seems that MACHINE_CLASS and SOC_FAMILY are different in what they are trying to accomplish.

SOC_FAMILY is defining a family of processors and the features that processor has.  Whereas MACHINE_CLASS is defining a type of device and its features which can use different processors.  So basically SOC_FAMILY defines common processors and MACHINE_CLASS defines common end devices.

This difference is why SOC_FAMILY exists.  We can have many processors that fit into an SOC_FAMILY that can be used across a variety of end applications from DVR to ip net camera to mobile phone.

> 
> p.
> 
> 
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Denys Dmytriyenko]

On Wed, Aug 04, 2010 at 09:11:19AM -0500, Maupin, Chase wrote:
> 
> > Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
> > probably is more desirable than adding a completely new variable (i.e.
> > SOC_FAMILY) for that purpose.
> 
> From what I've learned it seems that MACHINE_CLASS and SOC_FAMILY are 
> different in what they are trying to accomplish.
> 
> SOC_FAMILY is defining a family of processors and the features that 
> processor has.  Whereas MACHINE_CLASS is defining a type of device and its 
> features which can use different processors.  So basically SOC_FAMILY 
> defines common processors and MACHINE_CLASS defines common end devices.
> 
> This difference is why SOC_FAMILY exists.  We can have many processors that 
> fit into an SOC_FAMILY that can be used across a variety of end applications 
> from DVR to ip net camera to mobile phone.

I guess this conclusion was direved solely from the names of the respective 
variables. As technically both of them are exactly the same - a variable set 
in machine config or include file and used in OVERRIDES...

-- 
Denys

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Phil Blundell]

On Wed, 2010-08-04 at 09:11 -0500, Maupin, Chase wrote:
> From what I've learned it seems that MACHINE_CLASS and SOC_FAMILY are different in what they are trying to accomplish.
> 
> SOC_FAMILY is defining a family of processors and the features that processor has.  Whereas MACHINE_CLASS is defining a type of device and its features which can use different processors.  So basically SOC_FAMILY defines common processors and MACHINE_CLASS defines common end devices.
> 
> This difference is why SOC_FAMILY exists.  We can have many processors that fit into an SOC_FAMILY that can be used across a variety of end applications from DVR to ip net camera to mobile phone.

It's not completely obvious to me that, at the level of concrete
semantics, there is any real difference between those two scenarios.  If
you consider a hardware design as being a black box which contains a
bunch of features, the question of how exactly those features are
partitioned between particular ICs seems more or less irrelevant. 

It probably is true that there are multiple ways to carve up the
universe of machines into different "family" sets, and it might well
also be true that MACHINE_CLASS and SOC_FAMILY are currently using
slightly different rules for that, but I remain to be convinced that
both of them serve a useful and distinct purpose.  If they do, that
would suggest that what you really want to be using is MACHINE_FEATURES
rather than having a different variable for each possible grouping.

p.

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Denys Dmytriyenko]

On Wed, Aug 04, 2010 at 12:48:38PM +0100, Phil Blundell wrote:
> On Wed, 2010-08-04 at 13:17 +0200, Michael 'Mickey' Lauer wrote:
> > FWIW, minimal is using MACHINE_CLASS since quite a while to
> > reduce the need for adding the same files over and over again,
> > this is being used e.g. for HTC msm7 series and OpenEZX series
> > (see conf/machine/include).
> > 
> > Perhaps it's time to standardize something like that.
> 
> Yeah.  I certainly don't think we want a proliferation of such
> mechanisms.  If minimal is already using MACHINE_CLASS then there would
> be some sense in trying to make use of the same thing.
> 
> Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
> probably is more desirable than adding a completely new variable (i.e.
> SOC_FAMILY) for that purpose.

Phil,

Re: "adding a completely new variable" - SOC_FAMILY has been in use for almost 
a year now. I'm hearing about MACHINE_CLASS for the first time, although it 
appears to be slightly older than SOC_FAMILY though. But it doesn't seem to be 
used anywhere besides in few machine configs and micro.conf. There are no 
recipes actually using it, unlike SOC_FAMILY...

I'm not saying one is better than the other (actually, unifying them would 
be nice), I'm just saying it's too late to object adding SOC_FAMILY...

-- 
Denys

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Graeme Gregory]

 On 04/08/10 17:13, Denys Dmytriyenko wrote:
> On Wed, Aug 04, 2010 at 12:48:38PM +0100, Phil Blundell wrote:
>> On Wed, 2010-08-04 at 13:17 +0200, Michael 'Mickey' Lauer wrote:
>>> FWIW, minimal is using MACHINE_CLASS since quite a while to
>>> reduce the need for adding the same files over and over again,
>>> this is being used e.g. for HTC msm7 series and OpenEZX series
>>> (see conf/machine/include).
>>>
>>> Perhaps it's time to standardize something like that.
>> Yeah.  I certainly don't think we want a proliferation of such
>> mechanisms.  If minimal is already using MACHINE_CLASS then there would
>> be some sense in trying to make use of the same thing.
>>
>> Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
>> probably is more desirable than adding a completely new variable (i.e.
>> SOC_FAMILY) for that purpose.
> Phil,
>
> Re: "adding a completely new variable" - SOC_FAMILY has been in use for almost 
> a year now. I'm hearing about MACHINE_CLASS for the first time, although it 
> appears to be slightly older than SOC_FAMILY though. But it doesn't seem to be 
> used anywhere besides in few machine configs and micro.conf. There are no 
> recipes actually using it, unlike SOC_FAMILY...
>
> I'm not saying one is better than the other (actually, unifying them would 
> be nice), I'm just saying it's too late to object adding SOC_FAMILY...
>
As the person who originally added MACHINE_CLASS to openmoko and the OE,
then removed it from OE I can say it has different meaning that
SOC_FAMILY. MACHINE_CLASS was to identify a range of machines that were
90% the same but had a few differences. It was used in a few recipes
which were MACHINE_ARCH to make them use the same ARCH in these recipes
to stop them being rebuilt when switching machines.

The original use was om-gta01 and om-gta02 which had the same MACHINE_CLASS.

Graeme

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Denys Dmytriyenko]

On Wed, Aug 04, 2010 at 05:59:30PM +0100, Graeme Gregory wrote:
>  On 04/08/10 17:13, Denys Dmytriyenko wrote:
> > On Wed, Aug 04, 2010 at 12:48:38PM +0100, Phil Blundell wrote:
> >> On Wed, 2010-08-04 at 13:17 +0200, Michael 'Mickey' Lauer wrote:
> >>> FWIW, minimal is using MACHINE_CLASS since quite a while to
> >>> reduce the need for adding the same files over and over again,
> >>> this is being used e.g. for HTC msm7 series and OpenEZX series
> >>> (see conf/machine/include).
> >>>
> >>> Perhaps it's time to standardize something like that.
> >> Yeah.  I certainly don't think we want a proliferation of such
> >> mechanisms.  If minimal is already using MACHINE_CLASS then there would
> >> be some sense in trying to make use of the same thing.
> >>
> >> Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
> >> probably is more desirable than adding a completely new variable (i.e.
> >> SOC_FAMILY) for that purpose.
> > Phil,
> >
> > Re: "adding a completely new variable" - SOC_FAMILY has been in use for almost 
> > a year now. I'm hearing about MACHINE_CLASS for the first time, although it 
> > appears to be slightly older than SOC_FAMILY though. But it doesn't seem to be 
> > used anywhere besides in few machine configs and micro.conf. There are no 
> > recipes actually using it, unlike SOC_FAMILY...
> >
> > I'm not saying one is better than the other (actually, unifying them would 
> > be nice), I'm just saying it's too late to object adding SOC_FAMILY...
> >
> As the person who originally added MACHINE_CLASS to openmoko and the OE,
> then removed it from OE I can say it has different meaning that
> SOC_FAMILY. MACHINE_CLASS was to identify a range of machines that were
> 90% the same but had a few differences. It was used in a few recipes
> which were MACHINE_ARCH to make them use the same ARCH in these recipes
> to stop them being rebuilt when switching machines.
> 
> The original use was om-gta01 and om-gta02 which had the same MACHINE_CLASS.

Graeme,

Thanks for clarifying this! Do you still see a benefit in using MACHINE_CLASS? 
Why was it removed from OE? I don't see why it can't be used alongside 
SOC_FAMILY...

-- 
Denys

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Graeme Gregory]

 On 04/08/10 18:21, Denys Dmytriyenko wrote:
> On Wed, Aug 04, 2010 at 05:59:30PM +0100, Graeme Gregory wrote:
>>  On 04/08/10 17:13, Denys Dmytriyenko wrote:
>>> On Wed, Aug 04, 2010 at 12:48:38PM +0100, Phil Blundell wrote:
>>>> On Wed, 2010-08-04 at 13:17 +0200, Michael 'Mickey' Lauer wrote:
>>>>> FWIW, minimal is using MACHINE_CLASS since quite a while to
>>>>> reduce the need for adding the same files over and over again,
>>>>> this is being used e.g. for HTC msm7 series and OpenEZX series
>>>>> (see conf/machine/include).
>>>>>
>>>>> Perhaps it's time to standardize something like that.
>>>> Yeah.  I certainly don't think we want a proliferation of such
>>>> mechanisms.  If minimal is already using MACHINE_CLASS then there would
>>>> be some sense in trying to make use of the same thing.
>>>>
>>>> Failing that though, testing COMPATIBLE_MACHINE against MACHINE_CLASS
>>>> probably is more desirable than adding a completely new variable (i.e.
>>>> SOC_FAMILY) for that purpose.
>>> Phil,
>>>
>>> Re: "adding a completely new variable" - SOC_FAMILY has been in use for almost 
>>> a year now. I'm hearing about MACHINE_CLASS for the first time, although it 
>>> appears to be slightly older than SOC_FAMILY though. But it doesn't seem to be 
>>> used anywhere besides in few machine configs and micro.conf. There are no 
>>> recipes actually using it, unlike SOC_FAMILY...
>>>
>>> I'm not saying one is better than the other (actually, unifying them would 
>>> be nice), I'm just saying it's too late to object adding SOC_FAMILY...
>>>
>> As the person who originally added MACHINE_CLASS to openmoko and the OE,
>> then removed it from OE I can say it has different meaning that
>> SOC_FAMILY. MACHINE_CLASS was to identify a range of machines that were
>> 90% the same but had a few differences. It was used in a few recipes
>> which were MACHINE_ARCH to make them use the same ARCH in these recipes
>> to stop them being rebuilt when switching machines.
>>
>> The original use was om-gta01 and om-gta02 which had the same MACHINE_CLASS.
> Graeme,
>
> Thanks for clarifying this! Do you still see a benefit in using MACHINE_CLASS? 
> Why was it removed from OE? I don't see why it can't be used alongside 
> SOC_FAMILY...
>
I removed it as it was only used for the openmoko machines and outside
of openmoko I didn't see the need for the extra maintenance work to save
10ms of build time. I was just not maintaining enough recipes that used it.

That doesn't mean that I'm against people re-introducing it if there is
a real need.

Graeme

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 04-08-10 02:19, Denys Dmytriyenko wrote:
> On Tue, Aug 03, 2010 at 04:47:20PM -0500, Chase Maupin wrote:
>> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>>   setting for a recipe.
>> * This will allow recipes to work for entire families of
>>   devices without having to maintain/update the compatible
>>   devices as new devices are added into a family
>>
>> Signed-off-by: Chase Maupin <chase.maupin@ti.com>
> 
> Acked-by: Denys Dmytriyenko <denys@ti.com>

Acked-by: Koen Kooi <k-kooi@ti.com>






> 
> While SOC_FAMILY feature is relatively new on its own, it was designed and now 
> being extensively used with TI machines and corresponding recipes as an 
> effective override - e.g. covering many OMAP3-based machines with a single 
> VAR_omap3 = "VAL" expression.
> 
>> ---
>>  classes/base.bbclass |    4 +++-
>>  1 files changed, 3 insertions(+), 1 deletions(-)
>>
>> diff --git a/classes/base.bbclass b/classes/base.bbclass
>> index 1847491..02c7875 100644
>> --- a/classes/base.bbclass
>> +++ b/classes/base.bbclass
>> @@ -389,7 +389,9 @@ python () {
>>              import re
>>              this_machine = bb.data.getVar('MACHINE', d, 1)
>>              if this_machine and not re.match(need_machine, this_machine):
>> -                raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
>> +                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
>> +                if this_soc_family and not re.match(need_machine, this_soc_family):
>> +                    raise bb.parse.SkipPackage("incompatible with machine %s" % this_machine)
>>  
>>          need_target = bb.data.getVar('COMPATIBLE_TARGET_SYS', d, 1)
>>          if need_target:
>> -- 
>> 1.7.0.4
>>
>>
>> _______________________________________________
>> Openembedded-devel mailing list
>> Openembedded-devel@lists.openembedded.org
>> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFMWS88MkyGM64RGpERAnkkAJ94TErZj0L40RwgMpHejgg8nO+CpgCeNVHI
6uB1qsq0Mj4IP48OxmIyTas=
=jTnu
-----END PGP SIGNATURE-----

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Frans Meulenbroeks]

2010/8/4 Koen Kooi <k.kooi@student.utwente.nl>

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 04-08-10 02:19, Denys Dmytriyenko wrote:
> > On Tue, Aug 03, 2010 at 04:47:20PM -0500, Chase Maupin wrote:
> >> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
> >>   setting for a recipe.
> >> * This will allow recipes to work for entire families of
> >>   devices without having to maintain/update the compatible
> >>   devices as new devices are added into a family
> >>
> >> Signed-off-by: Chase Maupin <chase.maupin@ti.com>
> >
> > Acked-by: Denys Dmytriyenko <denys@ti.com>
>
> Acked-by: Koen Kooi <k-kooi@ti.com>
>
> I have no problems with the patch as such, but would appreciate a patch for
the documentation that describes this variable.

Frans

>
>
>
> >
> > While SOC_FAMILY feature is relatively new on its own, it was designed
> and now
> > being extensively used with TI machines and corresponding recipes as an
> > effective override - e.g. covering many OMAP3-based machines with a
> single
> > VAR_omap3 = "VAL" expression.
> >
> >> ---
> >>  classes/base.bbclass |    4 +++-
> >>  1 files changed, 3 insertions(+), 1 deletions(-)
> >>
> >> diff --git a/classes/base.bbclass b/classes/base.bbclass
> >> index 1847491..02c7875 100644
> >> --- a/classes/base.bbclass
> >> +++ b/classes/base.bbclass
> >> @@ -389,7 +389,9 @@ python () {
> >>              import re
> >>              this_machine = bb.data.getVar('MACHINE', d, 1)
> >>              if this_machine and not re.match(need_machine,
> this_machine):
> >> -                raise bb.parse.SkipPackage("incompatible with machine
> %s" % this_machine)
> >> +                this_soc_family = bb.data.getVar('SOC_FAMILY', d, 1)
> >> +                if this_soc_family and not re.match(need_machine,
> this_soc_family):
> >> +                    raise bb.parse.SkipPackage("incompatible with
> machine %s" % this_machine)
> >>
> >>          need_target = bb.data.getVar('COMPATIBLE_TARGET_SYS', d, 1)
> >>          if need_target:
> >> --
> >> 1.7.0.4
> >>
> >>
> >> _______________________________________________
> >> Openembedded-devel mailing list
> >> Openembedded-devel@lists.openembedded.org
> >> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Tom Rini]

Frans Meulenbroeks wrote:
> 2010/8/4 Koen Kooi <k.kooi@student.utwente.nl>
> 
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 04-08-10 02:19, Denys Dmytriyenko wrote:
>>> On Tue, Aug 03, 2010 at 04:47:20PM -0500, Chase Maupin wrote:
>>>> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>>>>   setting for a recipe.
>>>> * This will allow recipes to work for entire families of
>>>>   devices without having to maintain/update the compatible
>>>>   devices as new devices are added into a family
>>>>
>>>> Signed-off-by: Chase Maupin <chase.maupin@ti.com>
>>> Acked-by: Denys Dmytriyenko <denys@ti.com>
>> Acked-by: Koen Kooi <k-kooi@ti.com>
>>
>> I have no problems with the patch as such, but would appreciate a patch for
> the documentation that describes this variable.

In fact, I thought there was vague agreement here from the last time we 
added or changed variable behavior.  Patch the docs at the same time.

-- 
Tom Rini
Mentor Graphics Corporation

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Phil Blundell]

On Tue, 2010-08-03 at 16:47 -0500, Chase Maupin wrote:
> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>   setting for a recipe.
> * This will allow recipes to work for entire families of
>   devices without having to maintain/update the compatible
>   devices as new devices are added into a family

I don't suppose you'd consider refactoring your MACHINE names in order
to allow a simple regex in COMPATIBLE_MACHINE?  I don't have a strong
objection to this patch but, conceptually, it seems a bit ugly.

p.

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Khem Raj]

On Wed, Aug 4, 2010 at 2:54 AM, Phil Blundell <philb@gnu.org> wrote:
> On Tue, 2010-08-03 at 16:47 -0500, Chase Maupin wrote:
>> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>>   setting for a recipe.
>> * This will allow recipes to work for entire families of
>>   devices without having to maintain/update the compatible
>>   devices as new devices are added into a family
>
> I don't suppose you'd consider refactoring your MACHINE names in order
> to allow a simple regex in COMPATIBLE_MACHINE?

that would be a good approach too. Has it been
considered.

I don't have a strong
> objection to this patch but, conceptually, it seems a bit ugly.



>
> p.
>
>
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Maupin, Chase]

> -----Original Message-----
> From: openembedded-devel-bounces@lists.openembedded.org
> [mailto:openembedded-devel-bounces@lists.openembedded.org] On Behalf Of
> Khem Raj
> Sent: Wednesday, August 04, 2010 2:49 PM
> To: openembedded-devel@lists.openembedded.org
> Subject: Re: [oe] [PATCH] base.bbclass: add support for SOC_FAMILY in
> COMPATIBLE_MACHINES
> 
> On Wed, Aug 4, 2010 at 2:54 AM, Phil Blundell <philb@gnu.org> wrote:
> > On Tue, 2010-08-03 at 16:47 -0500, Chase Maupin wrote:
> >> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
> >>   setting for a recipe.
> >> * This will allow recipes to work for entire families of
> >>   devices without having to maintain/update the compatible
> >>   devices as new devices are added into a family
> >
> > I don't suppose you'd consider refactoring your MACHINE names in order
> > to allow a simple regex in COMPATIBLE_MACHINE?
> 
> that would be a good approach too. Has it been
> considered.

As I mentioned before, we are matching the machine names to the name of the actual device.  Unfortunately, not every device that falls into an SOC family has a similar name and cannot all be dictated by us.  For example in omap3 SOC_FAMILY there is:

beagleboard
omap3evm
nokia900
dm37x-evm
archos5
bug20

and many many more.  These machines are defined by a variety of companies, some from TI, some community, Nokia, Archos, etc.  I don't think it is likely that we will get agreement on the naming between all the groups.  I know that just within TI the naming is driven by marketing names for the devices.

> 
> I don't have a strong
> > objection to this patch but, conceptually, it seems a bit ugly.
> 
> 
> 
> >
> > p.
> >
> >
> >
> > _______________________________________________
> > Openembedded-devel mailing list
> > Openembedded-devel@lists.openembedded.org
> > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
> >
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Khem Raj]

On Wed, Aug 4, 2010 at 1:24 PM, Maupin, Chase <chase.maupin@ti.com> wrote:
>> -----Original Message-----
>> From: openembedded-devel-bounces@lists.openembedded.org
>> [mailto:openembedded-devel-bounces@lists.openembedded.org] On Behalf Of
>> Khem Raj
>> Sent: Wednesday, August 04, 2010 2:49 PM
>> To: openembedded-devel@lists.openembedded.org
>> Subject: Re: [oe] [PATCH] base.bbclass: add support for SOC_FAMILY in
>> COMPATIBLE_MACHINES
>>
>> On Wed, Aug 4, 2010 at 2:54 AM, Phil Blundell <philb@gnu.org> wrote:
>> > On Tue, 2010-08-03 at 16:47 -0500, Chase Maupin wrote:
>> >> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>> >>   setting for a recipe.
>> >> * This will allow recipes to work for entire families of
>> >>   devices without having to maintain/update the compatible
>> >>   devices as new devices are added into a family
>> >
>> > I don't suppose you'd consider refactoring your MACHINE names in order
>> > to allow a simple regex in COMPATIBLE_MACHINE?
>>
>> that would be a good approach too. Has it been
>> considered.
>
> As I mentioned before, we are matching the machine names to the name of the actual device.  Unfortunately, not every device that falls into an SOC family has a similar name and cannot all be dictated by us.  For example in omap3 SOC_FAMILY there is:

and omap3 family is armv7-a or cortex-a8 based can it have different
CPUs and be still categorized as omap3
is there omap3 specific you need which is not covered with armv7-a or cortex-a8

>
> beagleboard
> omap3evm
> nokia900
> dm37x-evm
> archos5
> bug20
>
> and many many more.  These machines are defined by a variety of companies, some from TI, some community, Nokia, Archos, etc.  I don't think it is likely that we will get agreement on the naming between all the groups.  I know that just within TI the naming is driven by marketing names for the devices.
>
>>
>> I don't have a strong
>> > objection to this patch but, conceptually, it seems a bit ugly.
>>
>>
>>
>> >
>> > p.
>> >
>> >
>> >
>> > _______________________________________________
>> > Openembedded-devel mailing list
>> > Openembedded-devel@lists.openembedded.org
>> > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>> >
>>
>> _______________________________________________
>> Openembedded-devel mailing list
>> Openembedded-devel@lists.openembedded.org
>> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 04-08-10 23:37, Khem Raj wrote:
> On Wed, Aug 4, 2010 at 1:24 PM, Maupin, Chase <chase.maupin@ti.com> wrote:
>>> -----Original Message-----
>>> From: openembedded-devel-bounces@lists.openembedded.org
>>> [mailto:openembedded-devel-bounces@lists.openembedded.org] On Behalf Of
>>> Khem Raj
>>> Sent: Wednesday, August 04, 2010 2:49 PM
>>> To: openembedded-devel@lists.openembedded.org
>>> Subject: Re: [oe] [PATCH] base.bbclass: add support for SOC_FAMILY in
>>> COMPATIBLE_MACHINES
>>>
>>> On Wed, Aug 4, 2010 at 2:54 AM, Phil Blundell <philb@gnu.org> wrote:
>>>> On Tue, 2010-08-03 at 16:47 -0500, Chase Maupin wrote:
>>>>> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
>>>>>   setting for a recipe.
>>>>> * This will allow recipes to work for entire families of
>>>>>   devices without having to maintain/update the compatible
>>>>>   devices as new devices are added into a family
>>>>
>>>> I don't suppose you'd consider refactoring your MACHINE names in order
>>>> to allow a simple regex in COMPATIBLE_MACHINE?
>>>
>>> that would be a good approach too. Has it been
>>> considered.
>>
>> As I mentioned before, we are matching the machine names to the name of the actual device.  Unfortunately, not every device that falls into an SOC family has a similar name and cannot all be dictated by us.  For example in omap3 SOC_FAMILY there is:
> 
> and omap3 family is armv7-a or cortex-a8 based can it have different
> CPUs and be still categorized as omap3
> is there omap3 specific you need which is not covered with armv7-a or cortex-a8

DSP, SGX, IVA, DSS, AESE, ISP, take your pick.

regards,

Koen
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFMWf4jMkyGM64RGpERAnM2AKCR2UEEdmgB/WFr0F9UZ55DqffJ0ACgrROX
7OIjSOU9ki67zIi8c1Xx66k=
=nZ7Z
-----END PGP SIGNATURE-----

Re: [PATCH] base.bbclass: add support for SOC_FAMILY in COMPATIBLE_MACHINES

[Khem Raj]

On (05/08/10 01:56), Koen Kooi wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 04-08-10 23:37, Khem Raj wrote:
> > On Wed, Aug 4, 2010 at 1:24 PM, Maupin, Chase <chase.maupin@ti.com> wrote:
> >>> -----Original Message-----
> >>> From: openembedded-devel-bounces@lists.openembedded.org
> >>> [mailto:openembedded-devel-bounces@lists.openembedded.org] On Behalf Of
> >>> Khem Raj
> >>> Sent: Wednesday, August 04, 2010 2:49 PM
> >>> To: openembedded-devel@lists.openembedded.org
> >>> Subject: Re: [oe] [PATCH] base.bbclass: add support for SOC_FAMILY in
> >>> COMPATIBLE_MACHINES
> >>>
> >>> On Wed, Aug 4, 2010 at 2:54 AM, Phil Blundell <philb@gnu.org> wrote:
> >>>> On Tue, 2010-08-03 at 16:47 -0500, Chase Maupin wrote:
> >>>>> * Add support for using SOC_FAMILY in the COMPATIBLE_MACHINES
> >>>>>   setting for a recipe.
> >>>>> * This will allow recipes to work for entire families of
> >>>>>   devices without having to maintain/update the compatible
> >>>>>   devices as new devices are added into a family
> >>>>
> >>>> I don't suppose you'd consider refactoring your MACHINE names in order
> >>>> to allow a simple regex in COMPATIBLE_MACHINE?
> >>>
> >>> that would be a good approach too. Has it been
> >>> considered.
> >>
> >> As I mentioned before, we are matching the machine names to the name of the actual device.  Unfortunately, not every device that falls into an SOC family has a similar name and cannot all be dictated by us.  For example in omap3 SOC_FAMILY there is:
> > 
> > and omap3 family is armv7-a or cortex-a8 based can it have different
> > CPUs and be still categorized as omap3
> > is there omap3 specific you need which is not covered with armv7-a or cortex-a8
> 
> DSP, SGX, IVA, DSS, AESE, ISP, take your pick.

I see. Do we have knobs to toggle their use in some applications ?

> 
> regards,
> 
> Koen
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.5 (Darwin)
> 
> iD8DBQFMWf4jMkyGM64RGpERAnM2AKCR2UEEdmgB/WFr0F9UZ55DqffJ0ACgrROX
> 7OIjSOU9ki67zIi8c1Xx66k=
> =nZ7Z
> -----END PGP SIGNATURE-----
> 
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel