Re: [Xen-devel] [XEN PATCH v2 2/6] xen: Have Kconfig check $(CC)'s version

[Jan Beulich <jbeulich@xxxxxxxx>]

On 08.01.2020 15:47, Anthony PERARD wrote:
> On Mon, Jan 06, 2020 at 03:34:43PM +0100, Jan Beulich wrote:
>> On 06.01.2020 15:01, Anthony PERARD wrote:
>>> On Fri, Jan 03, 2020 at 05:42:18PM +0100, Jan Beulich wrote:
>>>> Wouldn't both better
>>>> have a "depends on CC_IS_*" line instead? This would then also
>>>> result (afaict) in no CONFIG_CLANG_VERSION in .config if building
>>>> with gcc (and vice versa), instead of a bogus CONFIG_CLANG_VERSION=0.
>>>
>>> It sounds attracting to remove variables from .config, but it is equally
>>> attracting to always have a variable set. It can be used
>>> unconditionally when always set (without risking invalid syntax for
>>> example).
>>
>> Hmm, yes, as long as we don't have (by mechanical conversion) or gain
>> constructs like
>>
>> #if CONFIG_GCC_VERSION < 50000 /* must be gcc 4.x */
>>
>> Plus - what's CONFIG_CC_IS_{GCC,CLANG} good for then? The same can
>> then be achieved by comparing CONFIG_{GCC,CLANG}_VERSION against zero.
> 
> Sure, but it is much easier to understand what "ifdef CONFIG_CC_IS_GCC"
> is actually checking than it is to understand what
> "[ $CONFIG_GCC_VERSION -ne 0 ]" is for. In the second form, it isn't
> immediatly obvious for humans that we are simply checking which compiler
> is in use.

And I wasn't really suggesting to drop the CC_IS_* ones. What I
dislike is the duplication resulting from the *_VERSION ones not
having a "depends on CC_IS_*".

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel