[DTrace-devel] [PATCH 4/4] Move raise() tests to unittest/actions/raise with updates
Eugene Loh
eugene.loh at oracle.com
Wed Jun 9 22:35:29 PDT 2021
On 6/10/21 1:01 AM, Kris Van Hees wrote:
> On Thu, Jun 10, 2021 at 12:35:53AM -0400, Eugene Loh wrote:
>> Reviewed-by: Eugene Loh <eugene.loh at oracle.com>
>>
>> On 6/8/21 11:37 PM, Kris Van Hees wrote:
>>> Tests have been updated to not depend on trace() with a string argument.
>> I just don't get it. There is no point. What could happen? A test
>> fails for a reason we did not anticipate? That would be good -- a
>> godsend, actually -- not bad. These movements to narrow our test
>> coverage do not move us closer to a quality tool.
> What could happen is exactly what has been happening - these raise tests
> have remained marked as xfail even though the raise() action was implemented
> a while ago. And the reason they remained xfail is because they were failing
> even though raise() did actually work. They failed because they depended on
> a feature that didn't work yet. So, these tests that were meant to validate
> that raise() works as expected were not effective due to a side-effect.
Basically, each of our "unittests" relies on multiple features. That may
seem like a nuisance when bringing up a new implementation, but it's
nice to have once there is even a little maturity in the tool. So,
narrowing the tests strikes me as a short-sighted workaround.
> And no, making these tests not depend on trace(string) does not in any way
> affect the quality of our testing because we have tests that validate that
> trace(string) works correctly.
Well, assuming that we can foresee how things will fail. No engineer is
that prescient.
Anyhow, I gave my R-b.
More information about the DTrace-devel
mailing list