Re: [Qemu-devel] [PATCH 1/2] tests: build the helper program in main build process

[Wenchao Xia <xiawenc@linux.vnet.ibm.com>]

于 2013/9/23 7:22, Wenchao Xia 写道:
> On 09/23/2013 04:36 AM, Michael Tokarev wrote:
>> 22.09.2013 12:30, Wenchao Xia пишет:
>>> This is a quick way to update helper program when qemu main code
>>> is changed or built, instead of adding new Makefile under 
>>> test/qemu-iotest.
>>
>> Why?
>>
>> []
>    I think it is a bit overkill to have a new Makefile for one single 
> .c file. If more
> file comes, we can write Makefile at that time.
>
>
>>> +# Build the help program automatically
>>> +
>>> +all: $(QEMU_IOTESTS_HELPERS-y)
>>>
>>
>> This appears to be somewhat awkward way (to do something which I
>> clearly don't see).
>>
>> First, it is not clear why this iotest helper should be built
>> by default when we hasn't been asked to perform the tests.
>   Usually we may:
> 1 configure and make.
> 2 goto tests/qemu-iotests, ./check.
>
>   It may encounter an error since build of the helper was not triggered.
>
>>
>> Second, it apparently does not achieve the (unknown) goal
>> anyway.  At least the patch does not do what the commit
>> coment says.  Commit says that the helper gets updated when
>> qemu main code is changed or built (which is main code?).
>> But when I build, say, subdir-x86_64-softmmu, which is
>> definitely some "main" code, the helper isn't being built...
>>
>   Main code may refer to the codes other than ./tests, the commit
> message is not accurate, I think it should be "build the helper by
> default".
>
>> Count me confused...  :)
>>
>> /mjt
>>
>
>
   Are you OK with this method? To add a callable Makefile by 
tests/qemu-iotests/check,
configure, variable may be needed to work on out of tree build case, so 
I try to
avoid that. This patches simply build it in default make process.
   If yes I will respin a version with better commit message.