make *config regression: pkg-build

make *config regression: pkg-build

[Meelis Roos]

Just tried to run 'make menuconfig' on v4.18-10568-g08b5fa819970 and 
found a bad surprise:

'make *config' requires 'pkg-config'. Please install it.
make[1]: *** [scripts/kconfig/Makefile:219: scripts/kconfig/.mconf-cfg] Error 1

This is clearly a regression - I have libncurses devele pakcage 
installed in the default system location (as do 99%+ on actuall 
develeopers proobably) and in this case, pkg-config is useless. 
pkg-config is needed only when libraries and headers are installed in 
non-default locations but it is bad to require installation of 
pkg-config on all the machines where make menuconfig would be possibly 
run (for example, I have a aboult 100 machine kernel testbed with 
self-hosted kernel compilation and machine-specific kernel 
configurations that ocassionally need tweaking).

I notice 4.18 complained it can not find pkg-config but still worked. 
This is clearly better than now.

If we want to support developers with libraries in non-default 
locations, why not - but the common case of system include path should 
work without any trouble or warnings. For exaple, test if compilation 
against ncurses works, and if not retry it with pkg-config (and error 
out if it does not give working result).

-- 
Meelis Roos (mroos@linux.ee)

Re: make *config regression: pkg-build

[Randy Dunlap]

On 08/19/2018 04:15 AM, Meelis Roos wrote:
> Just tried to run 'make menuconfig' on v4.18-10568-g08b5fa819970 and 
> found a bad surprise:
> 
> 'make *config' requires 'pkg-config'. Please install it.
> make[1]: *** [scripts/kconfig/Makefile:219: scripts/kconfig/.mconf-cfg] Error 1
> 
> This is clearly a regression - I have libncurses devele pakcage 
> installed in the default system location (as do 99%+ on actuall 
> develeopers proobably) and in this case, pkg-config is useless. 
> pkg-config is needed only when libraries and headers are installed in 
> non-default locations but it is bad to require installation of 
> pkg-config on all the machines where make menuconfig would be possibly 
> run (for example, I have a aboult 100 machine kernel testbed with 
> self-hosted kernel compilation and machine-specific kernel 
> configurations that ocassionally need tweaking).
> 
> I notice 4.18 complained it can not find pkg-config but still worked. 
> This is clearly better than now.
>
> If we want to support developers with libraries in non-default 
> locations, why not - but the common case of system include path should 
> work without any trouble or warnings. For exaple, test if compilation 
> against ncurses works, and if not retry it with pkg-config (and error 
> out if it does not give working result).
> 

Hi,
This is due to my patch: 4ab3b80159d4db63b902ef635d4b5e882911b2da.

Yamada-san, any ideas?  Feel free to just revert it.

thanks,
-- 
~Randy

Re: make *config regression: pkg-build

[Masahiro Yamada]

Hi.

2018-08-20 0:37 GMT+09:00 Randy Dunlap <rdunlap@infradead.org>:
> On 08/19/2018 04:15 AM, Meelis Roos wrote:
>> Just tried to run 'make menuconfig' on v4.18-10568-g08b5fa819970 and
>> found a bad surprise:
>>
>> 'make *config' requires 'pkg-config'. Please install it.
>> make[1]: *** [scripts/kconfig/Makefile:219: scripts/kconfig/.mconf-cfg] Error 1
>>
>> This is clearly a regression - I have libncurses devele pakcage
>> installed in the default system location (as do 99%+ on actuall
>> develeopers proobably) and in this case, pkg-config is useless.
>> pkg-config is needed only when libraries and headers are installed in
>> non-default locations but it is bad to require installation of
>> pkg-config on all the machines where make menuconfig would be possibly
>> run (for example, I have a aboult 100 machine kernel testbed with
>> self-hosted kernel compilation and machine-specific kernel
>> configurations that ocassionally need tweaking).
>>
>> I notice 4.18 complained it can not find pkg-config but still worked.
>> This is clearly better than now.
>>
>> If we want to support developers with libraries in non-default
>> locations, why not - but the common case of system include path should
>> work without any trouble or warnings. For exaple, test if compilation
>> against ncurses works, and if not retry it with pkg-config (and error
>> out if it does not give working result).
>>
>
> Hi,
> This is due to my patch: 4ab3b80159d4db63b902ef635d4b5e882911b2da.
>
> Yamada-san, any ideas?  Feel free to just revert it.
>

Sorry for delay, I chose a partial revert
because I want to keep checking pkg-config for {g,x}config.

I posted a patch.


-- 
Best Regards
Masahiro Yamada