v4l2object only adds videometa if necessary. It performs several checks to see if the videometa is actually needed. One of these checks sees if there is padding present. However, at the moment, it only looks at the top and left borders for padding - not at the right and bottom ones.

The result is that pipelines like this one produce incorrect results:

  videotestsrc ! "video/x-raw, format=I420, width=480, height=270" ! x264enc ! v4l2videoNdec

Since 270 is not a multiple of 16, x264 will add 2 padding rows to the bottom. And since v4l2object ignores these padding rows, it will not add a videometa, even though it is actually needed to inform downstream about the correct plane offsets with the padding included.