Re: [PATCH] build: make sure build fails when running kconfig fails

[Jan Beulich <jbeulich@xxxxxxxx>]

On 21.02.2024 16:22, Anthony PERARD wrote:
> On Tue, Feb 20, 2024 at 01:20:16PM +0100, Jan Beulich wrote:
>> On 20.02.2024 12:18, Anthony PERARD wrote:
>>> On Tue, Feb 20, 2024 at 09:43:56AM +0100, Jan Beulich wrote:
>>>> Because of using "-include", failure to (re)build auto.conf (with
>>>> auto.conf.cmd produced as a secondary target) won't stop make from
>>>> continuing the build. Arrange for it being possible to drop the - from
>>>> Rules.mk, requiring that the include be skipped for tools-only targets.
>>>> Note that relying on the inclusion in those cases wouldn't be correct
>>>> anyway, as it might be a stale file (yet to be rebuilt) which would be
>>>> included, while during initial build, the file would be absent
>>>> altogether.
>>>>
>>>> Fixes: 8d4c17a90b0a ("xen/build: silence make warnings about missing 
>>>> auto.conf*")
>>>> Reported-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
>>>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>>>
>>> Just to make sure, this patch is a workaround to a harmless bug in older
>>> version of GNU Make which print spurious error messages, and said bug as
>>> been fixed in GNU Make 4.2, right? Bug report:
>>>
>>>     bug #102: Make prints an incorrect error for missing includes
>>>     https://savannah.gnu.org/bugs/?func=detailitem&item_id=102
>>
>> Assuming you mean the change referenced in Fixes: - yes, that's my
>> understanding. (Whether the referenced make bug is the one I can't tell,
>> though. But it looks like it is.)
>>
>>>> --- a/xen/Makefile
>>>> +++ b/xen/Makefile
>>>> @@ -375,6 +375,7 @@ $(KCONFIG_CONFIG): tools_fixdep
>>>>  # This exploits the 'multi-target pattern rule' trick.
>>>>  # The syncconfig should be executed only once to make all the targets.
>>>>  include/config/%.conf include/config/%.conf.cmd: $(KCONFIG_CONFIG)
>>>> +  $(Q)rm -f include/config/$*.conf
>>>
>>> Maybe this should say "include/config/auto.conf" instead of using "$*".
>>> "syncconfig" is going to generate "auto.conf" and not "$*.conf". And it
>>> would make easier to find the "rm" command via "grep".
>>
>> Generally I advocate for using $* wherever possible in pattern rules. If,
> 
> Well, it might look like a pattern rule, but it isn't one. It's a
> "groupes targets" rule that works on version of Make that doesn't have
> this brand new feature. It would be written as:
>     include/config/auto.conf include/config/auto.conf.cmd&: $(KCONFIG_CONFIG)

Assuming new enough make.

> Second, the other command in the recipe doesn't use "$*", so it's not
> possible to use this rule with a value of $* other than "auto".

True. "auto" isn't used there either, though.

To emphasize this isn't really a pattern rule, how about shrinking the
stem along the lines of what I've done in

https://sourceware.org/git/?p=binutils-gdb.git;a=commitdiff;h=79d635fc64ba9ca97c24c0c52c9924dba77e9527

e.g. making it just the (first) dot here?

>> however, replacing that is the only way to get an ack for this change, then
>> I would (hesitantly) do so.
> 
> Yes please, even if I don't like the patch. I hope it isn't going to
> byte us later if something in "tools/" happens to start depending on
> Kconfig (only fixdep and kconfig need to not depend on kconfig at the
> moment). But the patch currently works and looks ok. So, with "$*"
> replace by "auto", you can add my "Reviewed-by: Anthony PERARD
> <anthony.perard@xxxxxxxxxx>".

Thanks.

Jan