Re: [PATCH v2 6/6] selftests: enable O and KBUILD_OUTPUT

[Bamvor Zhang Jian]

Hi, Shuah

On 21 March 2017 at 21:54, Shuah Khan <shuahkh@xxxxxxxxxxxxxxx> wrote:
> On 03/21/2017 03:00 AM, Bamvor Zhang Jian wrote:
>> Hi,
>>
>> On 21 March 2017 at 16:35, Michael Ellerman <mpe@xxxxxxxxxxxxxx> wrote:
>>> Andy Lutomirski <luto@xxxxxxxxxxxxxx> writes:
>>>
>>>> On Tue, Nov 29, 2016 at 3:55 AM,  <bamvor.zhangjian@xxxxxxxxxx> wrote:
>>>>> From: Bamvor Jian Zhang <bamvor.zhangjian@xxxxxxxxxx>
>>>>>
>>>>> Enable O and KBUILD_OUTPUT for kselftest. User could compile kselftest
>>>>> to another directory by passing O or KBUILD_OUTPUT. And O is high
>>>>> priority than KBUILD_OUTPUT.
>>>>
>>>> Sorry for noticing this late, but this patch
>>>> (a8ba798bc8ec663cf02e80b0dd770324de9bafd9) is really annoying for
>>>> people who *don't* use these fancy options:
>>>
>>> Yeah sorry, it wasn't quite ready to go in.
>
> Bamovar,
>
> Please give me heads up and ask me to not commit the patch,
> if you think it isn't ready.
>
>>>
>>>> $ make -C tools/testing/selftests/x86 ldt_gdt_32
>>>> make: Entering directory '/home/luto/apps/linux/tools/testing/selftests/x86'
>>>> Makefile:44: warning: overriding recipe for target 'clean'
>>>> ../lib.mk:55: warning: ignoring old recipe for target 'clean'
>>>> make: *** No rule to make target 'ldt_gdt_32'.  Stop.
>>>> make: Leaving directory '/home/luto/apps/linux/tools/testing/selftests/x86'
>>>>
>>>> Is there any way that you can make this work again?
>>>
>>> There obviously is *a* way, but I'm not sure there's a simple and
>>> obviously correct way that is an easy fix for 4.11.
>>>
>>> I see at least 18 Makefile's in tools/testing/selftests that use
>>> $(OUTPUT)/, which would all need to be updated at least to use $(OUTPUT)
>>> (no trailing slash), and then some other changes to not propagate OUTPUT
>>> when the user didn't specify it. But hopefully someone will prove me
>>> wrong.
>> I also look at this issue. Originally, I use OUTPUT without slash in my
>> patch. People argue that it is not very clear. So, I add slash in curent
>> version.
>>>
>>> As a (poor) alternative you can do:
>>>
>>>  $ cd tools/testing/selftests/x86; make $PWD/ldt_gdt_32
>>>
>>> or just:
>>>
>>>  $ make -C tools/testing/selftests/x86
>>>
>>> cheers
>> Do we really need "make -C tools/testing/selftests/x86 ldt_gdt_32"?
>> It is useful but it will skip the top level Makefile of selftests.
>
> Being able to build individual tests is an important use-case. Please
> see kselftest.txt under Documentation directory for all the use-cases
> and new patches shouldn't break these use-cases.
Understand. I am sorry I do not know this use case before. And I read
kselftest.txt again, I do not find this use case is mentioned. So,
I add a few lines in kselftest.txt to avoid break this use case in future.
Is there any other use cases I should test?
>
> Breaking these use-cases is a regression and we have to fix it
Here is a fix. I could build individual test case in x86 and build the whole
x86 test cases successful respectively. And I do a quick test for all the
subsets. Please review the patch if it is make sense to you and Michael.
At the same time, I am buildind and testing all the testcases.

Michael: could you please do me a favor to test the powerpc part? Thanks.