Re: [PATCH] systemd-serialgetty: Remove SERIAL_CONSOLE definition, because it depends of machine definition. Set it, here, can be make the value is not good.

[Gary Thomas <gary@mlbassoc.com>]

On 2015-05-22 07:32, Paul Eggleton wrote:
> On Friday 22 May 2015 07:28:25 Gary Thomas wrote:
>> On 2015-05-22 07:12, Paul Eggleton wrote:
>>> Hi Thomas,
>>>
>>> On Friday 22 May 2015 15:02:02 Thomas PERROT wrote:
>>>> If the machine layer definition is applied after meta-systemd also
>>>> with a default setting operator instead an affectation, this var will
>>>> be bad initialized. It's the case with meta-raspberrypi layer.
>>>
>>> 1) Where in meta-systemd is SERIAL_CONSOLE being set? I can't find it
>>> anywhere in there.
>>
>> It's actually in OE-core:
>>     meta/recipes-core/systemd/systemd-serialgetty.bb
>>>
>>> 3) When the machine configuration sets the value, that happens pretty
>>> early on in the parsing process so there's no way that this value is
>>> going to take precedence over the machine value. Have you looked at the
>>> output of bitbake -e systemd-serialgetty | less to see how the value of
>>> this variable is being set?
>
> Right, then my comment stands. Setting the value with ?= in systemd-
> serialgetty.bb is not going to override a value set by the machine
> configuration, whether the machine configuration uses ?= or = - the machine
> configuration is parsed first, and ?= means "set if not already set" - which it
> will be by the time the recipe is parsed.

I still think it's wrong to be setting such a variable in
this [obscure] recipe.

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