[Buildroot] [PATCH v2 00/12] RFC: ARC port

[Arnout Vandecappelle <arnout@mind.be>]

On 02/05/13 16:03, Mischa Jonker wrote:
>>>      Maybe others will have a suggestion on this point?
>>
>>    Not entirely clean either, but what about the following:
>>
>> [toolchain/gcc/Config.in]
>> config BR2_ARCH_HAS_ALL_GCC
>> 	bool
>>
>> config BR2_ARCH_HAS_GCC_4_4
>> 	bool
>> 	default y if BR2_ARCH_HAS_ALL_GCC
>>
>> [arch/Config.in]
>> config BR2_mips
>> 	select BR2_ARCH_HAS_ALL_GCC
>> ...
>>
>> [arch/Config.in.arm]
>> config BR2_arm1176jzf_s
>> 	select BR2_ARCH_HAS_ALL_GCC
>> ...
>> config BR2_cortex_a8
>> 	select BR2_ARCH_HAS_GCC_4_4
>> ...
>
> The problem with this is that you end up in Config.in.arm with a lot of "select BR2_ARCH_HAS_ALL_GCC" lines. It is more logical, but it adds a lot of clutter IMHO.

  Okay, yet another idea. I don't know if this will actually work.

[toolchain/gcc/Config.in]
config BR2_ARCH_HAS_ALL_GCC
	bool
	default y if !(BR2_ARCH_HAS_GCC_4_4 || BR2_ARCH_HAS_GCC_4_5 ...)

config BR2_ARCH_HAS_GCC_4_4
	bool

config BR2_GCC_VERSION_4_4_X
	depends on BR2_ARCH_HAS_ALL_GCC || BR2_ARCH_HAS_GCC_4_4


> Unfortunately, we cannot make "BR2_ARCH_HAS_ALL_GCC" default y, because we don't have the 'deselect' keyword, so we need to add it everywhere. For ARM and x86, the architectures that aren't supported by all GCC versions are the exception, so instead of adding additional lines for just the exceptions, we end up with extra code everywhere.
>
> I also tried to let default values depend on earlier versions:
> config BR2_ARCH_HAS_GCC_4_5
> 	bool
> 	default y if BR2_ARCH_HAS_GCC_4_4
>
> But this doesn't work for several architectures, such as BR2_sparc_sparcsfleon, that do have GCC 4.4, but no 4.5 for instance.
>
> The reverse logic helps in the sense that the BR2_HAS_NO_GCC_x_y lines will go away over time when old GCC versions get deprecated.

  Yes, looking at the alternatives, the negative logic isn't that bad...

  However, the advantage of the positive logic is that it is slightly 
more future-safe. When adding a new gcc version, you have to remember to 
also add all the negative cases. Of course, you also have to remember to 
add a new positive case, but it is less bad if you forget that.

>
> I will send both variants later today.

  As Thomas said, don't feel obliged to continue working on this refactoring.


  Regards,
  Arnout

> Mischa
>
>


-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F