I know this isn't 100% related to csound development but I want to learn
more about the troubles developing a multiplatform system such as csound
and how one tests the various versions that come out. I'm particularly
interesting in what testing occurs, what is is attempted and how it is
attempted.

I was wondering specifically for CSound are there any automated tests or
scores you run per version, to make sure it's compatible? Had anything
like that been tried previously? What were the problems?

How valuable would regression tests for csound be? Has a major problem
been having to have a human ear listen to the output, seeing as audio
isn't always the system everytime its rendered.

How do you test or benchmark the performance of the system in realtime
on various platforms?

I'd love to hear speculation about the software one would want to help
perform regression testing on csound.

abram

-- 
abez ------------------------------------------
http://www.abez.ca/ Abram Hindle (abez@abez.ca)
------------------------------------------ abez

For a while I used the Conder tests as reversion tests.  I had a
script that ran them and then did cmp on the output and the previous
output.  Took a lot of file space and eventually I stopped as I could
not accommodate it.  In Csound5 I do not think we have got that far
yet; too many problems remain.

I usually only tested on the Linux system.  On windows and the Mac I
did spot tests only.  Again lack of file space and very slow
processors (166MHz on Windoes) .  When Csound5 gets to a goo enough
state I will resurrect as I have a new disk.

==John ffitch