[U-Boot] [PATCH] LCD: Add an option to skip registration as an stdio output

[Hannes Schmelzer <hannes@schmelzer.or.at>]

On 12/02/2015 03:08 PM, Ayotte, Stephane wrote:
> On December-02-15 12:30 AM, Hannes Schmelzer wrote:
>
>> On 12/01/2015 10:31 PM, Ayotte, Stephane wrote:
>>
>> Hi Stephane,
>>> Thanks for the info, Hannes. I agree with you that I could have met my
>> requirements (of showing  "applicative" output to the LCD and
>> "technical/debug" output to the serial port) with
>> CONFIG_SYS_CONSOLE_OVERWRITE_ROUTINE. To be totally honest, I had not
>> seen this config and was attracted to how the video device did it.
>>> For the sake of argumentation only, I will raise the question whether
>> CONFIG_SYS_CONSOLE_OVERWRITE_ROUTINE breaks NETCONSOLE. If I
>> understand correctly, this config will force the console to use only the serial
>> device. I could imagine use cases for netconsole and LCD.
>>> I'm asking the question, really, as this is just from my reading of the code. I
>> haven't tested with netconsole.
>> I also read the code again and my understanding is.
>> First the environment is asked for stdin, stdout, stderr devices - afterwards
>> overwrite_console is asked how about searching for some other devices.
>> If overwrite_console returns != 0 the devs inputdev,outputdev,errdev are left
>> uninitialized / values from environment (see line 802...805 in
>> common/console.c).
>>
>> A few lines afterwards the fallback takes care about if NULL and assigns "serial"
>> to them.
>>
>> So there is no indication for me breaking netconsole.
>> Further i'm also using netconsole on my B&R boards, its working fine.
> Thanks for the clarification. In that case, you are right that we end up with two different solutions.
> It's too bad your clarification came just a little after the pull request.
>
>>> My feeling is that there should be a more generic way of saying whether a
>> device should register to stdio or not. My preference would go to an API over
>> individual configs for each device. Neither of the two solutions are ideal nor
>> completely implemented as this time.
>>> Anatolij, based on Hannes info, if you want to backtrack on this, I'm open both
>> ways.
>> I would suggest to have not too much ways for one goal. Maybe it would be
>> enough work a bit on documentation how to quiet console on LCD.
> I agree with you 100% on not having multiple ways. What would be the procedure at this point?
> Submit a reverse patch?
i would suggest reverting the commit and send some patch with the revert.
What do Anatolij or Tom say?
>
> Regards,
> Stephane
best regards,
Hannes