Re: [Bug-glpk] time limit odd case

[Andrew Makhorin]

Brady,

Thank you for the bug report.

> My student and I noticed some incorrect behavior (in our opinion) for
> the time limit parameter.  When solving a MILP, the time limit is only 
> checked while solving the root LP relaxation and then before solving 
> each node in the b&b tree.
> 
> We came across an odd case where one of the nodes takes a long time to 
> solve (maybe it's even a numerical issue).  At any rate, the time limit 
> does not work correctly, because it isn't checked until after the node 
> is solved.

That check is used because the time to solve a node subproblem is
assumed to be much less than the time to solve the entire problem.
The time limit is just an estimation, so probably there is no need
to trace it exactly.

Andrew Makhorin