[ANN]Draft rough plan for 3.6!
dway at riskmetrics.com
Wed Apr 16 00:09:03 UTC 2003
Marcus Denker wrote:
> On Tue, Apr 15, 2003 at 04:04:59PM +0200, Marcus Denker wrote:
> > Yes, there are two reasons: 1. This will be a nice starting point for
> > adding more tests. 2. even a simple test will catch some trivial errors...
> > I think for many packages we could come up with some pretty simple
> > tests. lets see... we have the folowing removals:
> > Macro benchmarks: just run them
> Ok, the "Benchmarks Tests" is now on SqueakMap.
> This is really trivial... just made a subclass of TestCase
> "MacroBenchmarksTest", added a method "testRun":
> self shouldnt: [MacroBenchmarks run] raise: Error.
> Done. Same for all the other Benchmark-classes.
> This is *realy* simple, but I think these kind of tests, run
> automatically on some kind of "Integration Server"
> can help us a lot, especially if you consider how simple this is.
I just wanted to add that this scheme of having a test package for each
removed "Squeak official" package sounds like an excellent idea. Probably
eventually (in a future version of SqueakMap) there would be a special link
from a package to its test package, or the test package would be a dependent,
or something similar. But for now, a regular extra package sounds great.
Mmm, would we want to make the original package maintainer also be the
maintainer of the test package? If possible, that might be best, even if you
end up writing some of the initial test packages.
- Doug Way
More information about the Squeak-dev