Oliver Tappe wrote:
On 2015-01-12 at 19:02:36 [+0100], Adrien Destugues<pulkomandy@xxxxxxxxx> wrote: [ ... ]It would be great to have a more complete testsuite, but for this we need to setup a buildbot running haiku so the tests can be run, and/or get our automated unit tests to run on the libbe_test platform so the other build bots can run the tests. Without an automated test suite, we can only rely on ourselves and our users to manually test that the image runs, and that everything works as expected. This is time-consuming for everyone, but the fix is not stopping to make changes to the code.Well, as you know, we do have a testsuite. The fact that it isn't being run as part of the automatic build process yet has no meaning whatsoever as long as it doesn't contain any tests covering the code that has been changed. I know writing tests can be somewhat difficult (especially for kernel code) and not many people like writing tests to begin with, but blaming the infrastructure is not really a good excuse.
I do see Adrian's point though. It is not a particularly helpful strategy to assume that old untested code works and and so should not be touched. It is probably better to replace the code with code that you can at least verify works manually. What would be even better is to verify the code via unit tests, but that is a lot of work and hasn't been done yet (patches welcome and all that). Baring unit tests, let's do our best to keep things from breaking and then unbreak them as soon as possible. Beta1 coming means that it is all the more important to verify that as much code works as possible.