TH3 Sooner Rather Than Later
n at rciss.us
Wed Oct 3 11:23:31 BST 2007
On 3 Oct 2007, at 11:11, Ovid wrote:
> I think we're getting distracted by side issues. We have ONE goal:
> Getting TH3 out the door. Personally, I think that parallelization is
> an appropriate target for 3.1, others seem to feel that 3.0 is OK, but
> if parallelization is going to hold things up, I'd be for a TH3
> with the parallel work following shortly thereafter.
The multiplexed parallel stuff probably just needs some test
coverage. There's already coverage for the multiplexer itself. The
forking parallel stuff should probably wait until 3.1.
> There are a
> couple of reasons for this:
> 1. We need to get TH3 out the door. People are asking me about it
> they want it. Parallel tests are wonderful gravy, but still gravy.
> 2. We're going to get bug reports for 3.0. They could be nasty. I'd
> rather not get them for our basics *and* for the parallel work.
> There's no need to bite off more than we can chew.
> We seem to be getting distracted on side issues. Right now, what
> releasing it and making it clear that the plugin/inheritance/whatever
> architecture functionality may change? The Test::Harness interface is
> stable and the prove interface is stable. This satisfies 99% of our
I'm fine with that. The only thing I'm really concerned about for TH
3.0 is the lack of test coverage for App::Prove.
> I do realize that by putting out TH3 a bit early, we could wind up
> the problem of having to change things later, but I think the wider
> exposure could expose us to the *real* problems we face rather than
> problems we're trying to solve in a vacuum.
> By this strategy, we can mark *parallel* as experimental (or
> pull it back to a branch), test on VMS and push this out. The *only*
> major caveat is that we don't break the toolchain. We're not
> likely to
> do that, right? So let's consider pulling the trigger and stop
> distracted :)
Andy Armstrong, Hexten
More information about the tapx-dev