[daip] VBGLU bug

[Eric Greisen]

We are thinking about a serious re-write of the task as well.
Currently, it reads data set 1 and writes the output.  Then it reads
the output in one sample at a time and for each new time fills a large
buffer with all samples from the 2nd data set for that time.  Then it
selects the samples and re-writes the output.  If the is a 3rd and a
4th input file it repeats the process.

This means that any times not in the first data set are lost and I
think also means that the task is vulnerable to other failures, one of
which we have now debugged.

One could write the task to make a list of all samples in each data
set and then re-read them in order writing out the output only once
containing all times that occur in any input data set.  I suspect that
this could be faster, with a better result.

It might help to keep those 2 data sets around - or have new ones
later - so that you can help us test any new version.  I am about to
go on vacation for 12 days so it will not happen all that soon.

Thanks for your persistence

Eric Greisen