Fix : Unable to provide valid hw_frames_ctx for CUDA filter chain in JavaCV GPU pipeline #2365

[marktech0813]

What I changed
If you set a hardware pixel format (e.g., AV_PIX_FMT_CUDA) on the filter, FFmpegFrameFilter now uses NV12 for the buffersrc’s pix_fmt and logs a warning. This prevents avfilter_graph_config() from failing due to a missing hw_frames_ctx.
You can still override the buffersrc arguments entirely via setVideoFilterArgs() if you need a different software format.

How to use it now

• Keep your GPU filter chain starting with hwupload_cuda (e.g., "hwupload_cuda,scale_cuda=...").
• If you call filter.setPixelFormat(AV_PIX_FMT_CUDA), the filter will automatically configure buffersrc with NV12 and continue; no hw_frames_ctx error.
• If your software input format isn’t NV12, set videoFilterArgs to the exact buffersrc args you need (it overrides the fallback), or set a software filter.setPixelFormat(...) like AV_PIX_FMT_YUV420P.

Status

• I changed FFmpegFrameFilter only; no other files touched.
• Prevents filter initialization failure when a hardware pix_fmt is set.
• Adds a hardware-format check and auto-fallback to NV12 for buffersrc; users can still fully override via setVideoFilterArgs().

[marktech0813]

Please review this PR and merge it.
thanks.