Re: [Edmonds-alg-users] Edmond algorithm

[Ali T. <ali...@gm...>]

Hi Andrea,

Thanks for reporting this! I believe you are correct in both the
description of the bug and the required fix.

I no longer use sourceforge to keep my repositories. I have moved to
github. Please switch to using the following git repository instead:
https://github.com/atofigh/edmonds-alg

I have already committed a patch with your suggested fix on the master
branch. Could you do me a favor and run all the tests for me (This code was
written a long time ago and I currently do not have boost installed)? You
just need to compile the test.cpp file and run the test.py script in the
test/ directory. If all tests pass, I will then create a new release tag.

Thanks again and happy new year to you too!
/ali

On Tue, Dec 29, 2015 at 7:55 AM, Andrea Marino <ma...@di...> wrote:

> Dear All,
>
> Thanks for this nice implementation of the Edmond Algorithm. We have been
> lucky to have found it and we are really happy to use it.
>
> I think there is a small problem sometimes giving segmentation fault. The
> problem is when the number of nodes is computed, namely in the function
> void operator()() in the file src/edmonds_optimum_branching_impl.hpp
> The number of nodes is computed as the 1+the maximum index among all the
> target of all the edges. This is not correct if the maximum has indegree 0.
> I think it is sufficient to add after the line
>
> max_vertex_idx = std::max(max_vertex_idx, index[target(e, g)]);
>
> also
>
> max_vertex_idx = std::max(max_vertex_idx, index[source(e, g)]);
>
> It is certainly a very small problem, but since our input had this bad
> shape, we spent a while to find out this.
>
> Thanks again for the nice job.
> With my very best regards,
> Happy new year
> Andrea Marino
>
>
>
>
>
> ------------------------------------------------------------------------------
>
> _______________________________________________
> Edmonds-alg-users mailing list
> Edm...@li...
> https://lists.sourceforge.net/lists/listinfo/edmonds-alg-users
>
>