Re: [Qemu-devel] [PULL 18/21] build: don't build hardware objects with linux-user

["Philippe Mathieu-Daudé" <philmd@redhat.com>]

On 5/21/19 10:16 PM, Laurent Vivier wrote:
> On 21/05/2019 14:54, Daniel P. Berrangé wrote:
>> On Tue, May 21, 2019 at 02:52:42PM +0200, Laurent Vivier wrote:
>>> On 21/05/2019 13:52, Daniel P. Berrangé wrote:
>>>> On Wed, May 15, 2019 at 10:50:30PM +0200, Paolo Bonzini wrote:
>>>>> From: Laurent Vivier <lvivier@redhat.com>
>>>>>
>>>>> Some objects are only needed for system emulation and tools.
>>>>> We can ignore them for the user mode case
>>>>>
>>>>> Update tests to run accordingly: conditionally build some tests
>>>>> on CONFIG_BLOCK.
>>>>>
>>>>> Some tests use components that are only built when softmmu or
>>>>> block tools are enabled, not for linux-user. So, if these components
>>>>> are not available, disable the tests.
>>>>>
>>>>> Signed-off-by: Laurent Vivier <laurent@vivier.eu>
>>>>> Message-Id: <20190401141222.30034-6-lvivier@redhat.com>
>>>>> ---
>>>>>    Makefile               |  4 +++
>>>>>    Makefile.objs          | 14 +++++---
>>>>>    tests/Makefile.include | 90
>>>>> +++++++++++++++++++++++++-------------------------
>>>>>    3 files changed, 58 insertions(+), 50 deletions(-)
>>>>>
>>>>> diff --git a/Makefile b/Makefile
>>>>> index 1851f8c..155f066 100644
>>>>> --- a/Makefile
>>>>> +++ b/Makefile
>>>>> @@ -87,6 +87,10 @@ endif
>>>>>    include $(SRC_PATH)/rules.mak
>>>>> +# notempy and lor are defined in rules.mak
>>>>> +CONFIG_TOOLS := $(call notempty,$(TOOLS))
>>>>> +CONFIG_BLOCK := $(call lor,$(CONFIG_SOFTMMU),$(CONFIG_TOOLS))
>>>>
>>>> IMHO calling this CONFIG_BLOCK is a really poor choice, as
>>>> the decision is completely unrelated to block modules. It
>>>> made really confused when trying to understand why all the
>>>> crypto or auth code had been made conditional on the block
>>>> drivers.  The block code is just one part of QEMU that
>>>> is used in tools & softmmu.
>>>>
>>>> It would be better as CONFIG_SOFTMMU_TOOLS, or
>>>> CONFIG_NOT_USER, or something else.
>>>
>>> Do you think it's worth a patch to change the name?
>>
>> Personally I'd like to see a patch to give it a better name.
> 
> I'm wondering which name to use.
> 
> CONFIG_NOT_USER is not correct because the flag can be 'y' with
> CONFIG_USER_ONLY set.
> CONFIG_SOFTMMU_TOOLS is too fuzzy.
> 
> CONFIG_HW ?
> CONFIG_IO ?
> CONFIG_DEVICE ?

CONFIG_MACHINE_HW?
CONFIG_MACHINE_IO_HELPERS?
CONFIG_AIO?

BTW looking at util/Makefile.objs there is various think you can skip
for linux-user IMO.

Regards,

Phil.