Is it time to start regression testing yet?

[Jordan K. Hubbard]

On Jun 6, 2009, at 1:01 PM, Daniel J. Luke wrote:

> There have been a few attempts at this in the past - what was wrong  
> with them again? (I can remember at least three - your chroot build  
> scripts, wbb4's builds that auto generated a website that listed  
> failures and successes, and blb's mpab http://trac.macports.org/wiki/MPAB 
> )

They all "kinda worked" but not well enough that nobody wanted to  
actually host them anywhere since they were more clearly science  
experiments, abandoned in early childhood, rather than any serious  
attempt at regression testing.

> Perhaps there's just a little more work that needs to happen to get  
> things into a state where we can work out the 'homing' situation?

Sure.  Any of the above (I think mpab might be the best starting  
point) could be carried forward to the point where the output they  
produced was actually useful to some percentage of Portfile  
maintainers, which I think is the reasonable litmus test for "enough  
success to host".

> I actually had a hacked up version of your scripts running with disk  
> images (with the ports getting installed onto shadow files - I was  
> thinking it might be possible to build packages from that output  
> too ...) but got distracted by something shiny ;-)

That seems to happen to everyone in this space.  I think it's because  
testing is so exciting and capable of holding one's interest for  
extended periods. :)

> ... but packaging will all be fixed once we get apkg, right? ;-)

Absolutely!  :)  However, I'd be happy if we merely threw the output  
away completely right now and did it only for the testing value.

- Jordan