If The FfmpegVideoDecoder fails to create an image it will report an out of memory error. This can be misleading, as in the test introduced by bug 1542539 -- in this case the file is malformed (0 by 0 dimensions reported for some frames) and fails validation in CreateAndCopyData. In this scenario we're not running out of memory, but actively aborting our attempt to decode.

It would be useful to differentiate this case and report a different error. This would be helpful for debugging this path, and would make our test logs cleaner. The reason I found this bug is that someone suggested to me that the test could be breaking later tests by causing an OOM -- this turns out not to be the case, but it would be nice to have our tests logs indicate this more clearly.