[U-Boot] [PATCH 4/8] spi: Kconfig: Mark MXS_SPI has BROKEN

[Marek Vasut]

On 5/14/19 4:45 PM, Tom Rini wrote:
> On Tue, May 14, 2019 at 07:46:06PM +0530, Jagan Teki wrote:
>> On Mon, Apr 29, 2019 at 4:54 PM Tom Rini <trini at konsulko.com> wrote:
>>>
>>> On Mon, Apr 29, 2019 at 01:06:57PM +0200, Marek Vasut wrote:
>>>> On 4/29/19 1:02 PM, Jagan Teki wrote:
>>>>> On Mon, Apr 29, 2019 at 4:28 PM Marek Vasut <marex at denx.de> wrote:
>>>>>>
>>>>>> On 4/29/19 12:40 PM, Jagan Teki wrote:
>>>>>>> On Mon, Apr 29, 2019 at 3:24 PM Marek Vasut <marex at denx.de> wrote:
>>>>>>>>
>>>>>>>> On 4/28/19 10:47 PM, Jagan Teki wrote:
>>>>>>>>> Mark MXS_SPI has BROKEN, this so the resulting build shows
>>>>>>>>> warning for broken configuration enabled and associated code
>>>>>>>>> will remove in v2019.07 release.
>>>>>>>>
>>>>>>>> But the code is not broken, it works fine, right ?
>>>>>>>
>>>>>>> Yes the code work fine, but dm conversion not done yet. Now the
>>>>>>> CONFIG_BROKEN can mark the driver configs which doesn't convert dm
>>>>>>
>>>>>> That makes no sense, if the code works, it works and thus it is NOT
>>>>>> broken. Find some other way to flag it as non-DM-converted, not this hack.
>>>>>
>>>>> np, It depends on whether we consider non-dm code(for legacy SPI
>>>>> drivers) is broken or not, let Tom comment on this.
>>>>
>>>> Again, the code works, it's not broken. That's very clear.
>>>> DM conversion is a separate matter, so let's not overload CONFIG_BROKEN
>>>> for such things. The Makefile warning seems sufficient to me.
>>>
>>> The code is broken because it doesn't work with DM and is going to be
>>> removed due to lack of maintenance soon.  My problem with this series,
>>> but I need to check that we can do what I want to do and still keep
>>> travis green, is have this (and the other non-converted and thus
>>> non-maintained drivers) depend on, not select BROKEN.
>>
>> Yes, the idea here is to build the broken one by selecting BROKEN
>> so-that the travis-CI is green and have warning on the build,
>> otherwise it is another rework of selecting BROKEN for those
>> non-converted drivers if they depends.
>>
>>>
>>> The Makefile warnings tell people the code is going to be removed and
>>> I'm trying to have the window be:
>>> 1) LOUD warning about needing work
>>> 2) 1 release past those loud deadlines of the code there, but not built
>>> unless you enable BROKEN (as yes, I see code that's not updated to use
>>> DM as being broken)
>>> 3) Code removed.
>>
>> ie what this series does.  Driver (for example) will select BROKEN,
>> that means it is still working bu would like to remove in next
>> versions and so..
> 
> What I'm after is
> https://github.com/trini/u-boot/commits/WIP-BROKEN-test which is close,
> but not quite there.  We no longer build the drivers marked as BROKEN by
> default, and the fallout from that has been:
> - Some missing dependencies (I'm trying to fix that now) leading to
>   build problems.
> - Needing to yank CONFIG_SPI for example from configs that had been
>   using a now-not-built driver as the subsystem fails to link if there's
>   no controller.  I don't think that's a problem however.
> - Needing to migrate a few more symbols to Kconfig in order to be able
>   to properly express "You need to depend on BROKEN to use this option
>   now".

I still disagree with marking drivers that need DM conversion, but are
otherwise working perfectly fine, as broken. That's just misleading and
in fact incorrect. Invent some new, more descriptive Kconfig option please.

-- 
Best regards,
Marek Vasut