|
Boost : |
From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-10-29 10:29:32
David J. Pearce wrote:
> I think I've found a bug in the transitive_closure algorithm and I'm
> wondering if people agree it is a bug and/or whether it is already known.
Luckily, I can claim that I'm completely innocent of this bug. It turns out
that libs/graph/test/transitive_closure_test.cpp was seroiusly changed in
revision 1.3. It now tests all edges in transitive closure are valid, but it
does not test that transitive closure has all edges.
When I revert back to version 1.2, the test failes, while it certainly worked
for me.
[ time passes ]
Oh.. my! There are two serious bugs which the buggy testcase did not catch.
First is the one you've found. Second is that if original graph has a
loopback, then transitive closure will not have it. Could you grab the fixed
version from
http://zigzag.cs.msu.su:7813/boost/boost/graph/transitive_closure.hpp
and try it. It passes all tests for me. I also attach the diff for reference.
Jeremy, something should be done about it? I think we need to restore the 1.2
version of the testcase. I'm not sure about code. Could you tell how to
process transitive_closure.w? I get error from nuweb.
- Volodya
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk