[U-Boot] [WORKING PATCH 1/2] sandbox: Use defconfig to enable features

[Joe Hershberger]

Hi Simon,

On Wed, May 6, 2015 at 5:00 PM, Simon Glass <sjg at chromium.org> wrote:
> Hi Joe,
>
> On 6 May 2015 at 11:54, Joe Hershberger <joe.hershberger at ni.com> wrote:
>> Stop using the sandbox arch Kconfig to override defaults for config
>> options. This is a bit of abuse and may be causing build problems.
>>
>> Signed-off-by: Joe Hershberger <joe.hershberger at ni.com>
>> ---
>> The errors you are seeing are the result of "CONFIG_UNIT_TEST" not being
>> defined. That should not be at all related to that test/dm/eth.c patch.
>> This is changing the way that CONFIG_UNIT_TEST is enabled to be more
>> typical in case there is some bad behavior with all this override of
>> defaults. I have seen some non-ideal behavior, so I want to move away
>> from using it.
>> Please apply this on top of u-boot-dm/env-working and try it.
>>
>>  arch/sandbox/Kconfig      | 18 ------------------
>>  configs/sandbox_defconfig |  5 +++++
>>  2 files changed, 5 insertions(+), 18 deletions(-)
>
> I should have mentioned that the errors are fixed by a later patch:

I see. I didn't realize you were saying it was a bisectability issue.
It's odd that on your machine buildman told you it broke at "test: dm:
eth: Handle failed test env cleanup"

When I bisect it the failure happens at "test: Move the unit tests to
their own menu". I actually finally got my build machine in a state
where buildman can run there and it also claims that this ("test: Move
the unit tests to their own menu") is the patch that breaks.

> 61: test: dm: Move the dm tests over to the ut command

This is the patch that fixes it for me also.

> So should I merge this change into one of your existing patches for testing?

No... I've reordered a few patches to fix the bisectability and also
make the changes simpler. I'll send a new version that passes buildman
on my machine. It's nice to finally have buildman working. I like it
better than MAKEALL, but it is more complicated with more requirements
that were surprising.

> BTW this series does have some driver model things but most of it is
> not, so perhaps Tom plans to apply it?

I asked Tom in reply to the summary email of this series. Haven't
heard back. It also has some net things (but not mostly) that
instigated the series, so I could apply it too. Waiting to hear from
Tom.

Thanks,
-Joe