Re: [Xen-devel] [PATCH v5 03/10] x86: determine HAVE_AS_* just once

[Jan Beulich <jbeulich@xxxxxxxx>]

On 26.03.2020 14:42, Anthony PERARD wrote:
> On Thu, Mar 26, 2020 at 10:50:48AM +0100, Jan Beulich wrote:
>> On 25.03.2020 22:12, Andrew Cooper wrote:
>>> All the requisite infrastructure looks to be already present.
>>
>> ... there's the one open prereq question of what happens upon
>> tool chain updates. It's not clear to me if/how kconfig would
>> get invoked despite none of the recorded dependencies having
>> changed in such a case. (I'm sure you realize there's no issue
>> with this when the determination occurs out of a makefile.)
> 
> We might need one small change for this to happen, it is to add a
> comment in .config which display the output of `$(CC) --version | head
> -1`. Simple :-).
> If the output of `$(CC) --version` changes, kconfig will run again. That
> would be enough to detect tool chain updates, right?

I'm afraid it's not that simple: For one I'm not sure that line
would indeed change when a distro issues a gcc update. Even the
minor version may not change in this case; recall as an example
the backport of the compiler support backing INDIRECT_THUNK.
And then gcc isn't the tool chain - it may well be that e.g. gas
gets updated (supporting new insns or directives) without gcc
getting touched at all. Plus finally I don't think a comment
like you suggest would do - while processing it kconfig would
find that $(CC) gets used, but aiui it would then record just
$(CC) as needing tracking, not the output of the command. But
maybe I'm lacking some further detail here.

> Have a look at "include/config/auto.conf.cmd" to find out how kconfig is
> forced to run again.

Oh, that's good to know. Thanks for the pointer.

Jan