On 11/10/06, Stefano Bagnara <[EMAIL PROTECTED]> wrote:
Bernd Fondermann wrote:
> -1. reason: from my experience, commenting out unit tests leads to a
> degeneration of the test base as a whole, on the long run making the
> unit tests more or less useless. a small improvement or fix does not
> justify leaving our tests compromised.
Norman opened a JIRA issue assigned to Noel for this.
I think it is a good alternative to the other you're proposing.
ok then, lets see if this helps. with the JIRA I am now -0 on the change.
It is better to have unit tests run together with the main build anyway.
It is more convenient to notice a problem before a commit and make the
main code change and the unit test change together in one unit of
work.
> o let us simply have the breaking test until it gets fixed - bad,
> because some tests are not run anymore but maybe we need some more
> days until it gets fixed, which is fine.
Having a broken test may hide other problems arising from newer commits.
... and heightens the pressure to fix the failing test. ;-)
CI and Nightly builds no more give us useful output if we have a broken
test.
That said if you don't retire the veto or don't apply any of the other
solution I will try to follow the first solution (remove the patch that
broke the test). I don't like this solution but I prefer it over having
failing builds because of the failing test.
Succeeding tests (because of missing/commented tests) suggest that
everything is good while it isn't.
If the build fails, we are confronted with the fact that there is a problem.
Bernd
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
