Custom Query (545 matches)

From lemon-users by Pierre M:

Hi,

Thank you to develop and share the Lemon graph library.

By curiosity I have ran the code analyzer from Visual Studio on the Lemon library.

And it found some strange things in lemon/list_graph.hpp:

In all the ObserverProxy implementations we can see something like:

virtual void add(const std::vector<Node>& nodes) {
          for (int i = nodes.size() - 1; i >= 0; ++i) {
            snapshot.addNode(nodes[i]);
          }
        }

The increment step of the loop must be -- and not ++ !!

This error appears 18 times and must be fixed!

I'm using Lemon 1.3 and I have checked on the current mercurials files, the error is still there in the trunk.

Regards/ Cordialement, Pierre M

The nodes that are reached by DFS are wrong and depend on the order source nodes are added. The attached code implements a simple graph n0-->n3, n1-->n2 and adds n0 and n1 to the source nodes of DFS. Depending on the order either n2 or n3 is not reached.

output

add node 1 add node 0 node 3: reached node 2: NOT reached node 1: reached node 0: reached

add node 0 add node 1 node 3: NOT reached node 2: reached node 1: reached node 0: reached

We use WIN32 define at several places in order to check the underlying macro. This however does not work with (at least recent versions of) MinGW and CodeBlocks, because it does node define WIN32.

​This link indicates that this is actually the expected behaviour. The compiler normally defines _WIN32.

What shall we do? I can imagine the following resolutions.

1. Replace WIN32 to _WIN32 everywhere in the code
2. Put something like

   #ifdef _WIN32
   #ifndef WIN32
   #define WIN32
   #endif
   #endif
   

into core.h

3. Have the platform checked by CMAKE and put the WIN32 define conditionally to lemon/config.h (generated from lemon/config.h.in​)

We may define LEMON_WIN32 instead of WIN32.

Which option do you prefer?