Re: [RFC PATCH-for-10.1 00/39] single-binary: Make hw/arm/ common

[Pierrick Bouvier <pierrick.bouvier@linaro.org>]

On 4/4/25 11:04, Philippe Mathieu-Daudé wrote:
> On 4/4/25 19:12, Pierrick Bouvier wrote:
>> On 4/3/25 16:57, Philippe Mathieu-Daudé wrote:
>>> Hi,
>>>
>>> At this point this series is mostly a draft for Pierrick.
>>>
>>> After introducing the generic TargetInfo API [*], we implement
>>> the ARM variants, then use the API to remove target-specific code,
>>> allowing to eventually remove the target-specific arm_ss[] source
>>> set in meson.build, having all objects in arm_common_ss[].
>>> Various header cleanups are necessary before getting there.
>>>
>>
>> As there are some changes asked on this API, I think it's better if we
>> review this series later.
> 
> I think I misunderstood your comments in my series [*] reworking
> arm_cpu_has_feature(), I thought you wanted to see why it was for
> (completing your 'Make hw/arm/ common' effort).
> 

The assumption made in the current series is that we should remove 
dependency on cpu.h in target common code.
However, this assumption is false, and we can simply include cpu.h and 
do the same as before. For target/arm/cpu.h, it's possible because it 
was cleaned up previously to not have any target specific definition 
(removing any poisoned TARGET_ or CONFIG_ by exposing matching fields 
inconditionnally).

>> I'll still comment a few commits here that point some problems in the
>> approach.
> 
> Thanks!
> 
> [*]
> https://lore.kernel.org/qemu-devel/cecbab0f-56f2-434a-8508-8a4df2992259@linaro.org/
>