Re: [PATCH 0/1] Add explicit error for missing CONFIG_ASN1

[Randy Dunlap]

On 9/22/20 12:44 PM, James Bottomley wrote:
> On Tue, 2020-09-22 at 12:38 -0700, Randy Dunlap wrote:
>> On 9/22/20 12:19 PM, James Bottomley wrote:
>>> On Tue, 2020-09-22 at 11:54 -0700, Randy Dunlap wrote:
>>>> On 9/22/20 8:53 AM, James Bottomley wrote:
>>>>> I recently ran into this as an error from 0day.  On x86 it's
>>>>> pretty
>>>>> much impossible to build a configuration where CONFIG_ASN1
>>>>> isn't
>>>>> set, so you rarely notice a problem using the ASN.1 compiler
>>>>> because something else has selected it.  However, this compiler
>>>>> is
>>>>> never built if CONFIG_ASN1 isn't set and the error you get from
>>>>> kbuild is particularly unhelpful:
>>>>>
>>>>>    make[4]: *** No rule to make target 'security/keys/trusted-
>>>>> keys/tpm2key.asn1.o', needed by 'security/keys/trusted-
>>>>> keys/built-
>>>>> in.a'.
>>>>>    make[4]: *** [scripts/Makefile.build:283:
>>>>> security/keys/trusted-
>>>>> keys/trusted_tpm2.o] Error 1
>>>>>    make[4]: Target '__build' not remade because of errors.
>>>>>
>>>>> This patch changes the above error to the much easier to
>>>>> diagnose:
>>>>>
>>>>>    scripts/Makefile.build:387: *** CONFIG_ASN1 must be defined
>>>>> for
>>>>> the asn1_compiler.  Stop.
>>>>>    make[3]: *** [scripts/Makefile.build:505:
>>>>> security/keys/trusted-
>>>>> keys] Error 2
>>>>>
>>>>> James
>>>>>
>>>>> ---
>>>>>
>>>>> James Bottomley (1):
>>>>>   Makefile.build: Add an explicit error for missing ASN.1
>>>>> compiler
>>>>>
>>>>>  scripts/Makefile.build | 5 +++++
>>>>>  1 file changed, 5 insertions(+)
>>>>
>>>> Is there a missing
>>>> 	select ASN1
>>>> somewhere?
>>>
>>> You mean in the build used to produce the errors above?  Yes, so
>>> the
>>> patch is to make the problem more explicit.
>>
>> I appreciate that the message can be improved, but it seems possible
>> that some Kconfig could also be improved.
> 
> I don't really see how.  To find the problem you have to identify a
> conditional build in the Makefile that requires the asn1 compiler but
> for which the config option doesn't have a select ASN1.  We don't
> currently preserve the "what selected this symbol" information in
> kconfig, which is what we'd need.

Well, if you have a failing .config file, I would be glad to take a
look at it...

-- 
~Randy