if ( (bEnforceSliceRestriction && (m_pcCUAbove==NULL || m_pcCUAbove->getSlice()==NULL || m_pcCUAbove->getSCUAddr()+uiAPartUnitIdx < m_pcPic->getCU( getAddr() )->getSliceStartCU(uiCurrPartUnitIdx)))
      ||
       (bEnforceDependentSliceRestriction && (m_pcCUAbove==NULL || m_pcCUAbove->getSlice()==NULL || m_pcCUAbove->getSCUAddr()+uiAPartUnitIdx < m_pcPic->getCU( getAddr() )->getDependentSliceStartCU(uiCurrPartUnitIdx)))
      ||
       (bEnforceTileRestriction &&(m_pcCUAbove==NULL || m_pcCUAbove->getSlice()==NULL || (m_pcPic->getPicSym()->getTileIdxMap( m_pcCUAbove->getAddr() ) != m_pcPic->getPicSym()->getTileIdxMap(getAddr()))))
      )
  {
    return NULL;
  }

The function TComDataCU::getPUAbove returns NULL when the slice of current LCU is dependent slice, the slice of above LCU is dependent/independent slice and the slice of above-right LCU is independet.
It is a mismatch with the clause 6.4 in WD.

The start CU address of the slice(getSliceStartCU) is referred when deciding block availability as you can confirm in the following code.
I think that it affects to make wrong decision.

In addition, it seems that #643 ticket and this problem are caused by same reason.

It need to be correct in the view of HM stabiliy testing as well as clearing mismatch with WD.