After i had spent more than a week with hopeless debugging i found that the true solution was lost due to a bug in my source code: an array was treated as the indexing started from one but for that particular array indexing started from zero. As a result the LP solver got incorrect input data, and that is why it discarded the good solution (once again: due to my fault). It was very unfortunate that the incorrect data were very close to the good data values, i found the bug accidentally.
As a by product: i implemented the algorithm of A. Neumaier and O. Shcherbina, Safe bounds in linear and mixed-integer programming, Math. Programming A 99 (2004), 283-296; using GLPK and C-XSC. I plan to implement the enhanced version (faster) discussed in the paper in a few weeks and i would like to make that code publicly available. And of course, the enhanced version would not rely on C-XSC or other interval library. Thank you all again for you kind help. Ali _______________________________________________ Help-glpk mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-glpk
