[FFmpeg-devel] [PATCH] avutils/opt: fix discarded-qualifiers compiler warning

[Chad Fraleigh]

On 3/9/2021 5:30 AM, Nuo Mi wrote:
> On Tue, Mar 9, 2021 at 3:17 AM Chad Fraleigh <chadf at triularity.org> wrote:
> 
>> On 2/14/2021 10:12 AM, Nuo Mi wrote:
>>> On Mon, Feb 15, 2021 at 2:08 AM Andreas Rheinhardt <
>>> andreas.rheinhardt at gmail.com> wrote:
>>>
>>>> Nuo Mi:
>>>>> ---
>>>>>    libavutil/opt.c | 2 +-
>>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/libavutil/opt.c b/libavutil/opt.c
>>>>> index 590146b5fb..c554e9c063 100644
>>>>> --- a/libavutil/opt.c
>>>>> +++ b/libavutil/opt.c
>>>>> @@ -1735,7 +1735,7 @@ const AVClass *av_opt_child_class_iterate(const
>>>> AVClass *parent, void **iter)
>>>>>    #if FF_API_CHILD_CLASS_NEXT
>>>>>    FF_DISABLE_DEPRECATION_WARNINGS
>>>>>        if (parent->child_class_next) {
>>>>> -        *iter = parent->child_class_next(*iter);
>>>>> +        *iter = (void*)parent->child_class_next(*iter);
>>>>>            return *iter;
>>>>>        }
>>>>>    FF_ENABLE_DEPRECATION_WARNINGS
>>>>>
>>>> This doesn't look like a fix; you just silenced a warning.
>>>>
>>> Yes, but this the only way if we do not change the function signature.
>>
>> Since a major release is "coming up", now could be the time to fix the
>> signature to '(..., const void **iter)' and eliminate the non-const cast
>> kludge across the board for these cases.
> 
> Are we plan to remove all warnings before the release?
> Or more aggressively, is it feasible to treat all warnings as errors in the
> future? :)

I was thinking of just the 14 or so opt iterator functions which are 
assign to an AVClass.child_class_iterate (and related code in opt.c / 
opt.h / their iterate callers), so they no longer have to do a de-const 
(void *) cast to update *iter.

Though, it seems av_filter_iterate() and av_bsf_iterate() (maybe 
others?) would need to (/should?) be changed too, as they do the same 
non-const argument and are directly chained by child_class_iterate 
functions.