Misalignment between VTM and VVC draft for PDPC

[hongbin]

In 8.4.5.2.14 Position-dependent intra prediction sample filtering process, nScale is calculated and is used.
In SW, PDPC is disabled when nScale is less than 0. However, in Spec, there is no such constraint, which is incorrect.

The variable nScale is derived as follows:
– If predModeIntra is greater than INTRA_ANGULAR50, nScale is set equal to Min( 2, Log2( nTbH ) − Floor( Log2( 3 * invAngle − 2 ) ) + 8 ), using invAngle as specified in clause 8.4.5.2.12.
– Otherwise, if predModeIntra is less than INTRA_ANGULAR18, nScale is set equal to Min( 2, Log2( nTbW ) − Floor( Log2( 3 * invAngle − 2 ) ) + 8 ), using invAngle as specified in clause 8.4.5.2.12.
– Otherwise, nSacle is set to ( ( Log2( nTbW ) + Log2( nTbH ) − 2 ) >> 2 ).

shall be changed to

The variable nScale is derived as follows:
– If predModeIntra is greater than INTRA_ANGULAR50, nScale is set equal to Min( 2, Log2( nTbH ) − Floor( Log2( 3 * invAngle − 2 ) ) + 8 ), using invAngle as specified in clause 8.4.5.2.12.
– Otherwise, if predModeIntra is less than INTRA_ANGULAR18, nScale is set equal to Min( 2, Log2( nTbW ) − Floor( Log2( 3 * invAngle − 2 ) ) + 8 ), using invAngle as specified in clause 8.4.5.2.12.
– Otherwise, nSacle is set to ( ( Log2( nTbW ) + Log2( nTbH ) − 2 ) >> 2 ).
If nScale is less than zero, this process is terminated.

[fbossen]

I don't see a mismatch here. Both SW and text yield the same result when nScale is less than zero: the prediction array predSamples is not modified.