testing: prefer Assert over panic

testing: prefer Assert over panic



https://code.launchpad.net/~dave-cheney/juju-core/033-testing-mgo-panic/+merge/132146

(do not edit description out of merge proposal)

[rog, 2012-10-30 15:50:19]

https://codereview.appspot.com/6775079/diff/1/environs/dummy/environs.go
File environs/dummy/environs.go (right):

https://codereview.appspot.com/6775079/diff/1/environs/dummy/environs.go#newc...
environs/dummy/environs.go:30: "time"
nice

https://codereview.appspot.com/6775079/diff/1/environs/dummy/environs.go#newc...
environs/dummy/environs.go:173: testing.MgoReset(new(gocheck.C))
this seems wrong. if we haven't got a gocheck.C, panic is the only viable
alternative. there's no guarantee that new(gocheck.C) gives us a working *C
value.

one possibility would be to change MgoReset to accept an interface that
implements Fatalf, and write an implementation here that turns Fatalf into
panic.

https://codereview.appspot.com/6775079/diff/1/testing/mgo.go
File testing/mgo.go (right):

https://codereview.appspot.com/6775079/diff/1/testing/mgo.go#newcode86
testing/mgo.go:86: panic("MgoSuite tests must be run with MgoTestPackage")
if you're gonna prefer Assert over panic then calling Assert rather than panic
might be a good plan. but in general i think this all works ok, and we can leave
the panics alone.

https://codereview.appspot.com/6775079/diff/1/testing/mgo.go#newcode117
testing/mgo.go:117: panic(err)
ditto

https://codereview.appspot.com/6775079/diff/1/testing/mgo.go#newcode122
testing/mgo.go:122: panic(err)
ditto

https://codereview.appspot.com/6775079/diff/1/testing/mgo.go#newcode130
testing/mgo.go:130: panic(fmt.Errorf("Cannot drop MongoDB database %v: %v",
name, err))
ditto