[Bf-committers] Mantaflow landing and unit tests.

Ray Molenkamp ray at lazydodo.com
Fri Jan 3 03:49:31 CET 2020

There's a couple of things we can look at:

We can take human element out of it, running the tests on a proposed patch
can and probably should be done automatically. I heard there were plans for
this, I hope they materialize sooner rather than later. I think this is one of Nathans projects

We are running the tests now on a nightly basis which is great, however
nobody seems to be looking at the results otherwise we would have figured
out the tests were failing 3 weeks ago and either disabled the test or fixed
the problem. I kinda feel that looking at the tests on a daily basis should
be on someones daily morning checklist, perhaps Nathan and/or Dalai can find
someone for this task?

as for commit messages, we have good guidance there [1], having people try a
little harder will probably go a long way there.


[1] https://wiki.blender.org/wiki/Style_Guide/Commit_Messages

On 2020-01-02 7:04 p.m., Sebastián Barschkis wrote:
> Hi Ray,
> you’re right, the Cycles tests need to be updated. They need to make use of the new fluid modifier.
> Sergey pointed this out to me right after (within 12h) I landed the commits, i.e. the issue is known.
> So yes, blame it on me, I overlooked this part. I will look into this as soon as possible.
> I agree that this is not ideal. But apart from better commit messages, how can we do better at this time?
> Happy new year and best wishes,
> Sebastián
>> On 2. Jan 2020, at 22:31, Ray Molenkamp <ray at lazydodo.com> wrote:
>> All,
>> Hate to be a heckler for running the unit tests, but please:
>> When you land and/or review something big, RUN THE UNIT TESTS!
>> When the monster 10 patch mantaflow patch landed, it broke
>> 6 cycles unit tests on all platforms, 5 with different renders
>> than the reference images and one full on blender crash [1]
>> bda4a284d20164fec2433f7c40f49fc903319400 [2] fixed the render
>> differences and turned the crash into a render difference [3]
>> Unrelated side note: whats with the less than helpful
>> commit message on this commit? it may as well have been
>> committed with the message 'something fluid' or 'Tuesday'
>> would have just as enlightening for other developers.
>> To summarize:
>> - The person submitting the patches has not run the tests
>> - The reviewers have not run the tests
>> - Less than optimal commit messages
>> - 18!! days after landing, there is still a failing test
>> Holiday season or not: I think we can and should do better than this
>> --Ray
>> [1] https://i.imgur.com/LE3baOg.png
>> [2] https://developer.blender.org/rBbda4a284d20164fec2433f7c40f49fc903319400
>> [3] https://i.imgur.com/5we0hEv.png
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> https://lists.blender.org/mailman/listinfo/bf-committers
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> https://lists.blender.org/mailman/listinfo/bf-committers

More information about the Bf-committers mailing list