[FFmpeg-devel] [PATCH v2 00/16] avcodec/vvcdec: support subpicture

[James Almer]

On 4/4/2024 12:12 AM, Nuo Mi wrote:
> On Thu, Apr 4, 2024 at 10:09 AM James Almer <jamrial at gmail.com> wrote:
> 
>> On 4/3/2024 11:05 PM, Nuo Mi wrote:
>>> On Tue, Apr 2, 2024 at 11:24 PM James Almer <jamrial at gmail.com> wrote:
>>>
>>>> On 4/2/2024 10:01 AM, Nuo Mi wrote:
>>>>> On Wed, Mar 27, 2024 at 9:01 PM Nuo Mi <nuomi2021 at gmail.com> wrote:
>>>>>
>>>>>> Changes since v1:
>>>>>> add v2-0004-avcodec-vvcdec-fix-uninitialized-last-element-of-.patch to
>>>>>> address
>>>>>> the clang-usan and clang-asan failures reported by Frank
>>>>>
>>>>>
>>>>> pushed.
>>>>
>>>> I added a test for this using SUBPIC_C_ERICSSON_1.bit
>>>> Also, i disabled the tests for OPI_B_3.bit and VPS_A_3.bit as they rely
>>>> on a non-implemented feature. And while they still output a frame or two
>>>> as is, it's not the case on all systems.
>>>>
>>> Thank you.
>>>
>>>>
>>>> There's not a lot of VVC bitstreams currently in the FATE suite, so if
>>>> you know of any that will increase the coverage (See for example
>>>>
>>>>
>> http://coverage.ffmpeg.org/index.vvc_intra.c.5d0b519a39871515a1754ee8847b6d69.html#l678
>> ,
>>>>
>>>> where vvc_predict_ibc() is never run), then please make a test for it
>>>> and I'll upload the sample.
>>>>
>>> Sure. Tracked with https://github.com/ffvvc/FFmpeg/issues/206
>>
>> FWIW i added a test to cover vvc_predict_ibc() and other functions in
>> vvc_intra.c, using IBC_B_Tencent_2.bit
>>
> Thank  you,
> We'll establish the gcov environments on GitHub and incorporate additional
> clips from the conformance test into FATE
> Achieving 100% coverage might be challenging, but we can aim for 95% and
> cover all important branches at the very least.

I don't think 100% is possible because of the error paths. Also, once 
hwaccel is introduced eventually, that will also remain uncovered. But 
yes, 95% or so should be doable.