Logic error in GNMGenericNetwork::LoadGraph()

[Even Rouault]

Coverity Scan raises that the following tests in GNMGenericNetwork::LoadGraph() can only evaluate to FALSE since they are in a block controlled by if(bIsBlock) but "bool bIsBlock = GNM_BLOCK_NONE == nBlockState;"

            if(nBlockState & GNM_BLOCK_SRC)
                m_oGraph.ChangeBlockState(nSrcFID, bIsBlock);
            if(nBlockState & GNM_BLOCK_TGT)
                m_oGraph.ChangeBlockState(nTgtFID, bIsBlock);
            if(nBlockState & GNM_BLOCK_CONN)
                m_oGraph.ChangeBlockState(nConFID, bIsBlock);

The warning (KS 138329 Logically dead code) looks justified.

[Bishop]

There is an logical error (fixed in r32914) as bIsBlock should be true if nBlockState != GNM_BLOCK_NONE.

bool bIsBlock = GNM_BLOCK_NONE != nBlockState;

But nBlockState can take different values or combination of values:

#define GNM_BLOCK_NONE 0x0000  // no blocking (vertex or edge)
#define GNM_BLOCK_SRC  0x0001  // the source vertex is blocked
#define GNM_BLOCK_TGT  0x0002  // the target vertext is blocked
#define GNM_BLOCK_CONN 0x0004  // the connection edge is blocked
#define GNM_BLOCK_ALL GNM_BLOCK_SRC | GNM_BLOCK_TGT | GNM_BLOCK_CONN

So bIsBlock is not always FALSE.

[Even Rouault]

Any reason to keep the ticket opened ? r32914 seems like the appropriate fix.