Order of .conf processing?

Order of .conf processing?

[Gary Thomas]

What's the order of processing of the various .conf files?

It seems to me that bitbake.conf must be first as it seems
impossible to override settings in that file using a soft
assignment.  In particular, local.conf often has soft settings
but these dont' seem to change what's in bitbake.conf

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

Re: Order of .conf processing?

[Chris Larson]

On Mon, Jul 18, 2011 at 6:51 AM, Gary Thomas <gary@mlbassoc.com> wrote:
> What's the order of processing of the various .conf files?
>
> It seems to me that bitbake.conf must be first as it seems
> impossible to override settings in that file using a soft
> assignment.  In particular, local.conf often has soft settings
> but these dont' seem to change what's in bitbake.conf

BitBake only parses one global configuration file, bitbake.conf. It's
bitbake.conf that includes/requires all the other configuration files.
You can see the order by reading those inclusions in bitbake.conf.
Historically, the intention was for the configuration file load order
to be relatively parallel to the overrides order -- two different
mechanisms for implementing what is essentially a layered architecture
for the metadata, allowing, generally, more specific information to
override more general information. The configuration file order
differs slightly from that, however, as it's not possible to load the
distro or machine config files after local.conf, as local.conf is
generally where DISTRO and MACHINE are defined.
-- 
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics

Re: Order of .conf processing?

[Gary Thomas]

On 07/18/2011 07:59 AM, Chris Larson wrote:
> On Mon, Jul 18, 2011 at 6:51 AM, Gary Thomas<gary@mlbassoc.com>  wrote:
>> What's the order of processing of the various .conf files?
>>
>> It seems to me that bitbake.conf must be first as it seems
>> impossible to override settings in that file using a soft
>> assignment.  In particular, local.conf often has soft settings
>> but these dont' seem to change what's in bitbake.conf
>
> BitBake only parses one global configuration file, bitbake.conf. It's
> bitbake.conf that includes/requires all the other configuration files.
> You can see the order by reading those inclusions in bitbake.conf.
> Historically, the intention was for the configuration file load order
> to be relatively parallel to the overrides order -- two different
> mechanisms for implementing what is essentially a layered architecture
> for the metadata, allowing, generally, more specific information to
> override more general information. The configuration file order
> differs slightly from that, however, as it's not possible to load the
> distro or machine config files after local.conf, as local.conf is
> generally where DISTRO and MACHINE are defined.

In that case, isn't using ?= assignments in bitbake.conf a bit misleading?

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

Re: Order of .conf processing?

[Chris Larson]

On Mon, Jul 18, 2011 at 7:05 AM, Gary Thomas <gary@mlbassoc.com> wrote:
> On 07/18/2011 07:59 AM, Chris Larson wrote:
>>
>> On Mon, Jul 18, 2011 at 6:51 AM, Gary Thomas<gary@mlbassoc.com>  wrote:
>>>
>>> What's the order of processing of the various .conf files?
>>>
>>> It seems to me that bitbake.conf must be first as it seems
>>> impossible to override settings in that file using a soft
>>> assignment.  In particular, local.conf often has soft settings
>>> but these dont' seem to change what's in bitbake.conf
>>
>> BitBake only parses one global configuration file, bitbake.conf. It's
>> bitbake.conf that includes/requires all the other configuration files.
>> You can see the order by reading those inclusions in bitbake.conf.
>> Historically, the intention was for the configuration file load order
>> to be relatively parallel to the overrides order -- two different
>> mechanisms for implementing what is essentially a layered architecture
>> for the metadata, allowing, generally, more specific information to
>> override more general information. The configuration file order
>> differs slightly from that, however, as it's not possible to load the
>> distro or machine config files after local.conf, as local.conf is
>> generally where DISTRO and MACHINE are defined.
>
> In that case, isn't using ?= assignments in bitbake.conf a bit misleading?

First, there is one place variables can come from before the
configuration files, and that's the process environment, for variables
which aren't filtered out. Second, it could use ?= sanely after the
configuration files are included, as well.
-- 
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics

Re: Order of .conf processing?

[Richard Purdie]

On Mon, 2011-07-18 at 07:12 -0700, Chris Larson wrote:
> On Mon, Jul 18, 2011 at 7:05 AM, Gary Thomas <gary@mlbassoc.com> wrote:
> > On 07/18/2011 07:59 AM, Chris Larson wrote:
> >>
> >> On Mon, Jul 18, 2011 at 6:51 AM, Gary Thomas<gary@mlbassoc.com>  wrote:
> >>>
> >>> What's the order of processing of the various .conf files?
> >>>
> >>> It seems to me that bitbake.conf must be first as it seems
> >>> impossible to override settings in that file using a soft
> >>> assignment.  In particular, local.conf often has soft settings
> >>> but these dont' seem to change what's in bitbake.conf
> >>
> >> BitBake only parses one global configuration file, bitbake.conf. It's
> >> bitbake.conf that includes/requires all the other configuration files.
> >> You can see the order by reading those inclusions in bitbake.conf.
> >> Historically, the intention was for the configuration file load order
> >> to be relatively parallel to the overrides order -- two different
> >> mechanisms for implementing what is essentially a layered architecture
> >> for the metadata, allowing, generally, more specific information to
> >> override more general information. The configuration file order
> >> differs slightly from that, however, as it's not possible to load the
> >> distro or machine config files after local.conf, as local.conf is
> >> generally where DISTRO and MACHINE are defined.
> >
> > In that case, isn't using ?= assignments in bitbake.conf a bit misleading?
> 
> First, there is one place variables can come from before the
> configuration files, and that's the process environment, for variables
> which aren't filtered out. Second, it could use ?= sanely after the
> configuration files are included, as well.

It does this second part pretty clearly:

[...]
require conf/xxx.conf
require conf/yyy.conf

TUNE_ARCH ?= "INVALID"
TUNE_FEATURES ?= "${TUNE_FEATURES_tune-${TUNENAME}}"

##################################################################
# Weak variables (usually to retain backwards compatibility)
##################################################################

DL_DIR ?= "${TOPDIR}/downloads"
[...]

so most of the ?= variables are after the includes.

Looking at the file, there are some good candidates for either = or ??=
though, including I suspect the console ones which are likely what
prompted this?

Cheers,

Richard

Re: Order of .conf processing?

[Gary Thomas]

On 07/18/2011 08:28 AM, Richard Purdie wrote:
> On Mon, 2011-07-18 at 07:12 -0700, Chris Larson wrote:
>> On Mon, Jul 18, 2011 at 7:05 AM, Gary Thomas<gary@mlbassoc.com>  wrote:
>>> On 07/18/2011 07:59 AM, Chris Larson wrote:
>>>>
>>>> On Mon, Jul 18, 2011 at 6:51 AM, Gary Thomas<gary@mlbassoc.com>   wrote:
>>>>>
>>>>> What's the order of processing of the various .conf files?
>>>>>
>>>>> It seems to me that bitbake.conf must be first as it seems
>>>>> impossible to override settings in that file using a soft
>>>>> assignment.  In particular, local.conf often has soft settings
>>>>> but these dont' seem to change what's in bitbake.conf
>>>>
>>>> BitBake only parses one global configuration file, bitbake.conf. It's
>>>> bitbake.conf that includes/requires all the other configuration files.
>>>> You can see the order by reading those inclusions in bitbake.conf.
>>>> Historically, the intention was for the configuration file load order
>>>> to be relatively parallel to the overrides order -- two different
>>>> mechanisms for implementing what is essentially a layered architecture
>>>> for the metadata, allowing, generally, more specific information to
>>>> override more general information. The configuration file order
>>>> differs slightly from that, however, as it's not possible to load the
>>>> distro or machine config files after local.conf, as local.conf is
>>>> generally where DISTRO and MACHINE are defined.
>>>
>>> In that case, isn't using ?= assignments in bitbake.conf a bit misleading?
>>
>> First, there is one place variables can come from before the
>> configuration files, and that's the process environment, for variables
>> which aren't filtered out. Second, it could use ?= sanely after the
>> configuration files are included, as well.
>
> It does this second part pretty clearly:
>
> [...]
> require conf/xxx.conf
> require conf/yyy.conf
>
> TUNE_ARCH ?= "INVALID"
> TUNE_FEATURES ?= "${TUNE_FEATURES_tune-${TUNENAME}}"
>
> ##################################################################
> # Weak variables (usually to retain backwards compatibility)
> ##################################################################
>
> DL_DIR ?= "${TOPDIR}/downloads"
> [...]
>
> so most of the ?= variables are after the includes.
>
> Looking at the file, there are some good candidates for either = or ??=
> though, including I suspect the console ones which are likely what
> prompted this?

Yes, I tried to override TERMCMD using ?= in local.conf and that didn't
work.  Using = fixed it though.

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

Re: Order of .conf processing?

[Khem Raj]

On Mon, Jul 18, 2011 at 7:34 AM, Gary Thomas <gary@mlbassoc.com> wrote:
> On 07/18/2011 08:28 AM, Richard Purdie wrote:
>>
>> On Mon, 2011-07-18 at 07:12 -0700, Chris Larson wrote:
>>>
>>> On Mon, Jul 18, 2011 at 7:05 AM, Gary Thomas<gary@mlbassoc.com>  wrote:
>>>>
>>>> On 07/18/2011 07:59 AM, Chris Larson wrote:
>>>>>
>>>>> On Mon, Jul 18, 2011 at 6:51 AM, Gary Thomas<gary@mlbassoc.com>
>>>>> wrote:
>>>>>>
>>>>>> What's the order of processing of the various .conf files?
>>>>>>
>>>>>> It seems to me that bitbake.conf must be first as it seems
>>>>>> impossible to override settings in that file using a soft
>>>>>> assignment.  In particular, local.conf often has soft settings
>>>>>> but these dont' seem to change what's in bitbake.conf
>>>>>
>>>>> BitBake only parses one global configuration file, bitbake.conf. It's
>>>>> bitbake.conf that includes/requires all the other configuration files.
>>>>> You can see the order by reading those inclusions in bitbake.conf.
>>>>> Historically, the intention was for the configuration file load order
>>>>> to be relatively parallel to the overrides order -- two different
>>>>> mechanisms for implementing what is essentially a layered architecture
>>>>> for the metadata, allowing, generally, more specific information to
>>>>> override more general information. The configuration file order
>>>>> differs slightly from that, however, as it's not possible to load the
>>>>> distro or machine config files after local.conf, as local.conf is
>>>>> generally where DISTRO and MACHINE are defined.
>>>>
>>>> In that case, isn't using ?= assignments in bitbake.conf a bit
>>>> misleading?
>>>
>>> First, there is one place variables can come from before the
>>> configuration files, and that's the process environment, for variables
>>> which aren't filtered out. Second, it could use ?= sanely after the
>>> configuration files are included, as well.
>>
>> It does this second part pretty clearly:
>>
>> [...]
>> require conf/xxx.conf
>> require conf/yyy.conf
>>
>> TUNE_ARCH ?= "INVALID"
>> TUNE_FEATURES ?= "${TUNE_FEATURES_tune-${TUNENAME}}"
>>
>> ##################################################################
>> # Weak variables (usually to retain backwards compatibility)
>> ##################################################################
>>
>> DL_DIR ?= "${TOPDIR}/downloads"
>> [...]
>>
>> so most of the ?= variables are after the includes.
>>
>> Looking at the file, there are some good candidates for either = or ??=
>> though, including I suspect the console ones which are likely what
>> prompted this?
>
> Yes, I tried to override TERMCMD using ?= in local.conf and that didn't
> work.  Using = fixed it though.

?= means you still wanted others places to override it if any
local.conf settings are mostly like what you want to have override over others
so using = would be right here.