>>>I'm trying to rein-in my coworkers from building and releasing our production EXE's without doing full regression testing first. They do a build then when they try a couple things and it doesn't bomb, they think we are good to go. I come from a different school. Any thoughts on this? Oh, I have seen one error about VFP9 thinking that a file that is not opened exclusively needs to be, but they think that may just be an environmental setting, but maybe I'll save that for another thread.
>>
>>We don't really have the tools to test the UI's completely. What I find helps is to do unit testing as Craig B. said. Of course, if you're not making discrete modules, unit testing is harder.
>>
>>I also have a person take the production exe, install and run it on a clean machine, install and run it on a machine simulating an upgrade and finally only that person can say if it's good to go to customers.
>>
>>My wife and I do this for my customers. It reduces the number of times I get a call from customers that it didn't install and/or didn't work. The only drawback is my wife gets to see my errors. ;)
>
>I'm specifically referring to going from VFP8 to VFP9. They haven't actually released anything yet, but it seems to be that the entire system needs to be "touched" in testing before anyone can say it is good to upgrade. Just don't see how someone would release in this scenario without full regression testing.

Do you have a known set of test data? I mean a set of data where the individual amounts and totals and resulting reports are guaranteed correct? Data generated from users isn't easy to eyeball for correctness.

If not, make it and use that set of data to compare the resulting reports, do some edits etc. Compare the old and new results.

HTH