There seems to be a mis-match between semantics, syntax table and HM regarding whether significant_coeff_flag should be inferred for the last position in the TU.

In the WD semantics it says that significant_coeff_flag is inferred for the last position; however, this is not the case in HM-9.0.

Note that this also means the ctxIdxMap must be defined for i=15 (ctxIdxMap[15]=8), since there can be a possible to signal significant_coeff_flag at (3,3) position of 4x4 TU.

In WD9 (7.4.9.11 Residual coding semantics):
When significant_coeff_flag[ xC ][ yC ] is not present, it is inferred as follows.
– If ( xC, yC ) is the last significant location ( LastSignificantCoeffX, LastSignificantCoeffY ) in scan order or both of the following conditions are true, significant_coeff_flag[ xC ][ yC ] is inferred to be equal to 1.

In HM-9.0, significant_coeff_flag is only inferred when numNonzero=0; however, for last position numNonZero is set to 1.

if( iScanPosSig == (Int) uiScanPosLast )
{

lastNZPosInCG = iScanPosSig;
firstNZPosInCG = iScanPosSig;
iScanPosSig--;
pos[ numNonZero ] = uiBlkPosLast;
numNonZero = 1;

}

...

if( uiSigCoeffGroupFlag[ iCGBlkPos ] )

{

if( iScanPosSig > iSubPos | | iSubSet == 0 | | numNonZero )
{

uiCtxSig = TComTrQuant::getSigCtxInc( patternSigCtx, uiScanIdx, uiPosX, uiPosY, blockType, uiWidth, uiHeight, eTType );
m_pcTDecBinIf->decodeBin( uiSig, baseCtx[ uiCtxSig ] );

}
else
{

uiSig = 1;

}

}