On 3/19/2025 1:36 PM, Zhao Zhili wrote:
> 
> 
>> On Mar 19, 2025, at 23:38, Sean McGovern <gseanmcg@gmail.com> wrote:
>>
>> Hi Zhao,
>>
>> On Mon, Mar 10, 2025 at 10:35 PM Zhao Zhili
>> <quinkblack-at-foxmail.com@ffmpeg.org> wrote:
>>>
>>> From: Zhao Zhili <zhilizhao@tencent.com>
>>>
>>> ---
>>> tests/fate/flvenc.mak                  |   4 +
>>> tests/ref/fate/enhanced-flv-hevc-hdr10 | 100 +++++++++++++++++++++++++
>>> 2 files changed, 104 insertions(+)
>>> create mode 100644 tests/ref/fate/enhanced-flv-hevc-hdr10
>>
>> Sorry for not addressing this earlier -- this fails on big-endian
>> PowerPC (both 32- and 64-bit). Please have a look at FATE.
>> The test should probably explicitly request that the sample is little-endian.
> 
> OK. It’s due to ffprobe shows the native endian yuv420p10 as yuv420p10le on little endian
> arch.
> 
> I can filter out the pix_fmt entry, but it’s a general problem in test: there is no way to separate
> native endian from explicit big/little endian in pixel format.

See what's done for fate-mxf-probe-applehdr10 in mxf.mak, which had the 
same issue.