make check fails for dispatch and eigs

[John W. Eaton]

On  9-Feb-2009, Michael D. Godfrey wrote:

| The most recent updates seem to have stopped the seg
| faults in dispatch, but the errors below still happen fairly
| regularly.  About one time in 4 there are no errors.

I commented out these tests and added the following note:

  ## The following tests have been disabled because creating functions
  ## on the fly causes trouble (filesystem timestamp resolution?) and so
  ## leads people to complain about the failed tests when the dispatch
  ## mechanism is working fine, but it is really the creation of the
  ## functions that is failing.  And anyway, this method of function
  ## dispatch should be considered obsolete and probably removed from
  ## Octave now that we have classes.

Does anyone disagree with this?

| While
| running make check to see about dispatch I noticed one
| failure in eigs.m.  This is below.  This error has not happened
| again.  Presumably eigs.m uses a random seed, and occasionally
| this leads to an error which exceeds the current test limit.

I increased the tolerance on all the tests in eigs.cc to 1e-11.

jwe