[Coin-discuss] Re: Trouble with COIN's MPS Writer

[Andrew Makhorin]

>       I've been using COIN to solve linear programming relaxations of
>SDPs.  However, I have a few test cases that seem to crash COIN.  So, 
>before I solve the LP, I write the problem to an MPS file.  However, it 
>appears that COIN is generating an invalid problem.  Here's a summary of 
>the results that I receive when I give the file to a variety of 
>different solvers:
>...
>glpsol          invalid data card

MPS format requires fields 5 and 6 to be placed in columns 40-47 and
50-61, respectively, while in your file these fields are placed in
columns 39-46 and 49-60. Note, however, that some solvers allow
deviations from strict MPS formatting (glpsol does not allow that).

Andrew Makhorin