Re: [Qemu-devel] [PATCH 02/16] qapi: Drop superfluous qapi_enum_parse() parameter max

[Markus Armbruster]

Eric Blake <address@hidden> writes:

> On 08/24/2017 01:35 PM, Markus Armbruster wrote:
>> Eric Blake <address@hidden> writes:
>> 
>>> On 08/24/2017 03:45 AM, Markus Armbruster wrote:
>>>> The lookup tables have a sentinel, no need to make callers pass their
>>>> size.
>>>>
>>>> Fun: the header has it in the wrong position.  Good riddance.
>>>>
>
>>>> +++ b/include/qapi/util.h
>>>> @@ -12,7 +12,7 @@
>>>>  #define QAPI_UTIL_H
>>>>  
>>>>  int qapi_enum_parse(const char * const lookup[], const char *buf,
>>>> -                    int max, int def, Error **errp);
>>>> +                    int def, Error **errp);
>>>
>>> I'm not sure what you meant by wrong position; were you thinking that
>>> lookup/max should be immediately adjacent (since max is a property of
>>> the lookup[] parameter), and sticking 'buf' in between the two is what
>>> meant 'max' was in the wrong position?
>> 
>> Compare the declaration above with the definition below:
>> 
>>     diff --git a/qapi/qapi-util.c b/qapi/qapi-util.c
>>     index 46eda7d..ee7594f 100644
>>     --- a/qapi/qapi-util.c
>>     +++ b/qapi/qapi-util.c
>>     @@ -16,7 +16,7 @@
>>      #include "qapi/util.h"
>> 
>>      int qapi_enum_parse(const char * const lookup[], const char *buf,
>>     -                    int max, int def, Error **errp)
>>     +                    int def, Error **errp)
>>      {
>>          int i;
>> 
>> Declaration has max before def, definition has it the other way round.
>
> Huh?  On current master (commit 248b2373), the two look like they match
> to me:
>
> $ git grep -A1 qapi_enum_parse'.*const look'
> include/qapi/util.h:int qapi_enum_parse(const char * const lookup[],
> const char *buf,
> include/qapi/util.h-                    int max, int def, Error **errp);
> --
> qapi/qapi-util.c:int qapi_enum_parse(const char * const lookup[], const
> char *buf,
> qapi/qapi-util.c-                    int max, int def, Error **errp)
>
>
>> 
>> Such errors are one reason I prefer to have documentation next to
>> definitions, which are authoritative, rather than declarations, which
>> may or may not match the definition.
>> 
>>> The change itself is reasonable, even if the commit message needs a
>>> tweak to answer my question.
>> 
>> Care to suggest a wording?
>
> At this point, I find the claim to be bogus, so I suggest you delete the
> 'Fun:' paragraph.

Sure.

>>> Reviewed-by: Eric Blake <address@hidden>
>> 
>> Thanks!