Wandering through the triple-r reports last night, I noticed that
people who are using the very latest RubyUnit (0.4.5) were getting a
bunch of type errors. This is because Rubicon has code like

   assert("cat" =~ animal)

which returns (say) '0' to indicate a match. '0' corresponds to true
(as does anything apart from nil and false), to the assert used to
pass and everyone was happy. Now, RubyUnit checks that boolean
parameters are really 'true' or 'false', so this type of assert now
throws an exception.

I fixed Rubicon last night to it now works with RubyUnit
again. However, this leaves us with 9 runs in triple-r that between
them contribute some 145 errors (out of 187) to the results, errors
that are not in Ruby itself.

So, would it be OK to go through and remove these errors from the
results. Or should I remove the whole run? Or should I forget about it
and leave the errors in the database?


Cheers


Dave