The Decoder interface that nsAVIFDecoder implements, calls DoDecode() twice: first with IsMetadataDecode() returning false, during which time the decoder must call PostSize() and (if appropriate PostHasTransparency()) and again with IsMetadataDecode() when the decoded image must be rendered to the surface. Currently this information is extracted from the decoded image metadata returned by dav1d or libaom. However, recent improvements to the the rust parser mean with a little additional work (basically just parsing the Image spatial extents (ispe) box and exposing it in the AvifImage struct), we can access the necessary information from the parser metadata without performing the decode at all for the IsMetadataDecode() case.

This was fixed as part of Bug 1696093