The string "boundary strength" appears twice in the body of the spec whereas "boundary filtering strength" is used seven times.

Maybe better to always use "boundary filtering strength" for consistency/searchability.

The term "coding subblock" is used in several places in the spec but is undefined. Examples of use are in 8.5.1, 8.5.3.1, 8.5.5.1.

When lossless encoding, DCT2xDCT2 is not restricted in chroma component.

In IntraSearch.cpp::xIntraCodingTUBlock, to restrict DCT2xDCT2 for luma component, following logic is applied:

...
#if JVET_AHG14_LOSSLESS
    if( !( m_pcEncCfg->getCostMode() == COST_LOSSLESS_CODING && tu.mtsIdx[compID] == 0 ) || tu.cu->bdpcmMode != 0 )
    {
      m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx(), loadTr);
    }
...
#endif

#if JVET_AHG14_LOSSLESS
  if( ( m_pcEncCfg->getCostMode() == COST_LOSSLESS_CODING && tu.mtsIdx[compID] == 0 ) && 0 == tu.cu->bdpcmMode )
  {
    uiAbsSum = 0;
    tu.getCoeffs( compID ).fill( 0 );
    TU::setCbfAtDepth( tu, compID, tu.depth, 0 );
  }
#endif

For chroma, the same logic should be applied to following part:

#if JVET_P0058_CHROMA_TS
    // encoder bugfix: Set loadTr to aovid redundant transform process
    m_pcTrQuant->transformNxN(tu, codeCompId, qpCbCr, uiAbsSum, m_CABACEstimator->getCtx(), loadTr);
#else
    m_pcTrQuant->transformNxN(tu, codeCompId, qpCbCr, uiAbsSum, m_CABACEstimator->getCtx());
#endif