On 7/30/07, Ovid <[EMAIL PROTECTED]> wrote: > aspersions on a great module. The Test::Exception/Sub::Uplevel problem > causing false negatives was always a bit frustrating for me and it came > out in the last email. > > I'm sorry if I offended you :(
Apology accepted. It frustrated me so much I wound up adopting Sub::Uplevel, so I definitely get where you're coming from. > Dependences are not bad and I never said that. Dependencies which > cause false negatives are bad. That's pretty much the whole toolchain, then, and all the major testing modules. Where do you draw the line? As I suggested, functional test module dependencies are OK, but release test module dependencies are not. In general, in managing dependencies, I think you've got (at least) two choices for anything non-standard. * List it as a dependency, and suffer any upstream test failures that prevent installation * Bundle it in t/inc (as many toolchain modules seem to do) and suffer the risk that the version you bundled contains a bug I get the sense that many application deployments follow the second approach, prefering the evil they know to the evil they don't. David
