Inter issues related to global variables

[vdrugeon]

Issue 1:
In 8.4.1 General decoding process for coding units coded in inter prediction mode, step 3 uses the number of luma coding subblocks in horizontal direction numSbX and in vertical direction numSbY as derived in step 2. However, it seems that these two variables are not derived for all possible conditions in step 2 (for example, it seems that these variables are not derived at all "if MergeTriangleFlag[ xCb ][ yCb ] is equal to 1, inter_affine_flag[ xCb ][ yCb ] and merge_subblock_flag[ xCb ][ yCb ] are both equal to 0,"). Should there be default values for these variables at the beginning of the section?

Issue 2:
In 8.5.6.1 General, the following global variables are set for the whole block: MvL0, MvL1, MvDmvrL0, MvDmvrL1, RefIdxL0, RefIdxL1, PredFlagL0, PredFlagL1, BcwIdx. This process is called for blocks coded with inter mode, except when triangle is used.
In 8.6.3.1 General, the same variables are also set for the whole block, except for MvDmvrL0 and MvDmvrL1. This process is called for blocks coded with IBC. It may not be an issue though, because MvDmvrL0 and MvDmvrL1 seem to be used only in 8.5.2.12 Derivation process for collocated motion vectors, and collocated motion vectors are not derived when the collocated block is coded with IBC.
However, in 8.5.7.3 Motion vector storing process for triangle merge mode (which is called for blocks using triangle merge mode), only the following variables are set for the whole block: MvL0, MvL1, RefIdxL0, RefIdxL1, PredFlagL0, PredFlagL1. MvDmvrL0, MvDmvrL1 and BcwIdx are missing. But at first sight, it seems that MvDmvrL0 and MvDmvrL1 may be referred to for blocks coded with triangle merge mode. Same thing for BcwIdx. Should these three variables be set to default values in 8.5.7.3 Motion vector storing process for triangle merge mode?

[jlchen]

Both reported issues are valid and will be addressed in a later released draft text.