On 9/21/25 12:17, Alyssa Ross wrote:
> Demi Marie Obenour <demiobenour@gmail.com> writes:
> 
>> On 9/21/25 12:13, Alyssa Ross wrote:
>>> Demi Marie Obenour <demiobenour@gmail.com> writes:
>>>
>>>> On 9/21/25 04:55, Alyssa Ross wrote:
>>>>> Demi Marie Obenour <demiobenour@gmail.com> writes:
>>>>>
>>>>>> On 9/19/25 07:55, Alyssa Ross wrote:
>>>>>>> Demi Marie Obenour <demiobenour@gmail.com> writes:
>>>>>>>
>>>>>>>> This detected a missing prototype.
>>>>>>>>
>>>>>>>> No functional change.
>>>>>>>>
>>>>>>>> Signed-off-by: Demi Marie Obenour <demiobenour@gmail.com>
>>>>>>>> ---
>>>>>>>>  tools/meson.build    | 2 ++
>>>>>>>>  tools/start-vmm/ch.h | 1 +
>>>>>>>>  2 files changed, 3 insertions(+)
>>>>>>>>
>>>>>>>> diff --git a/tools/meson.build b/tools/meson.build
>>>>>>>> index 9cebd03e323531fca7600cacf120161a98de16c5..8262f3e01d7bd56561306d7dd4650a22ca40ebe7 100644
>>>>>>>> --- a/tools/meson.build
>>>>>>>> +++ b/tools/meson.build
>>>>>>>> @@ -9,6 +9,8 @@ project('spectrum-tools', 'c',
>>>>>>>>    })
>>>>>>>>  
>>>>>>>>  add_project_arguments('-Wno-error=attributes', language : 'c')
>>>>>>>> +add_project_arguments('-Werror=missing-prototypes', language : 'c')
>>>>>>>> +add_project_arguments('-Werror=missing-declarations', language : 'c')
>>>>>>>>  
>>>>>>>>  if get_option('host')
>>>>>>>>    add_languages('rust')
>>>>>>>> diff --git a/tools/start-vmm/ch.h b/tools/start-vmm/ch.h
>>>>>>>> index 7230913ef0abf41a4f712ac4a543c7f7fdecec0f..5431365e6e2894cdebae22a9a44e2ccf1222e0d2 100644
>>>>>>>> --- a/tools/start-vmm/ch.h
>>>>>>>> +++ b/tools/start-vmm/ch.h
>>>>>>>> @@ -8,3 +8,4 @@ struct net_config {
>>>>>>>>  	char id[18];
>>>>>>>>  	uint8_t mac[6];
>>>>>>>>  };
>>>>>>>> +struct net_config net_setup(const char name[static 1], int name_len);
>>>>>>>
>>>>>>> Why do we need to declare this in a C header?  It's only used from Rust.
>>>>>>
>>>>>> Ideally the Rust declarations would be generated from the C ones
>>>>>> using bindgen.  Also, this catches genuine bugs on the C side.
>>>>>
>>>>> What bugs?  Implicit definitions are already disallowed, aren't they?
>>>>
>>>> Function declared in one file, defined in another file with different
>>>> prototype.  This makes it undefined behavior to call.
>>>
>>> I don't see how this really enforces that?  Fundamentally the problem
>>> there is conflicting prototypes, not missing ones.  If you had another
>>> compilation unit that didn't include the header with the prototype, and
>>> instead make its own declaration, you'd still have the same problem.  If
>>> we have a norm of not declaring functions outside of the header file
>>> that corresponds to and is included by the implementation file, that
>>> prevents that from happening regardless of whether we have these errors,
>>> assuming that the compiler would still catch the prototype not matching
>>> the implementation.
>>
>> It also catches cases where a function or variable should have been marked
>> static.
> 
> Okay, I /am/ bad at remembering to mark functions as static, and I guess
> there's not really a better way to avoid that, so happy to take it with
> that rationale.  Happy for me to just adjust the commit message to
> mention that?

Sure!
-- 
Sincerely,
Demi Marie Obenour (she/her/hers)