On Sat, 2010-11-20 at 22:24 +0100, Miloslav Trmač wrote:
Adam Williamson píše v So 20. 11. 2010 v 13:14 -0800:
On Sat, 2010-11-20 at 20:30 +0100, Miloslav Trmač wrote:
If the "something else" is more important than testing the update, testing the updates truly is a waste of time.
I don't disagree with anything you say, but the question of what's more important than testing an update is key. If an update's worth doing, it's worth testing.
.. and if it is worth testing, it is worth testing more, right?
Erm, I didn't say that, don't put words into my mouth.
Yet one has to stop sooner or later.
Sure. That's why FESCo quite heavily toned down the testing requirements from the initial draft: the initial draft required all updates to have +3 karma to be pushed, remember. That was considerably reduced to +1/+1 for critpath, and no specific requirement for non-critpath.
Let me give you a specific example: I maintain quite a few leaf packages. The packages have an automated test suite. I test the code changes as applied the main branch. I test the final update RPMs rebuilt locally my system.
Given all this testing, I'm not going to spend time testing the particular builds on all supported distributions - the overhead would often take more time than all of the testing above, and it is much less likely to find any problems. The only really remaining risks are compiler problems (which are extremely rare) and dependency problems (which are just as rare for these particular packages). Mirek
It's already been mentioned in this thread that we could probably adjust the policy for packages which have automated test suites.