Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: meas_base
-
Labels:None
-
Team:External
Description
lsst.meas.base.tests.TestDataset contains two methods (makePerturbedWcs and realize) that take an optional seed for the random number generator. In both cases, the seed defaults to None, which means the RNG is seeded in some non-deterministic manner. However, this is intended for tests, which we require are deterministic. The current API therefore makes it too easy (simply neglecting to specify the optional randomSeed keyword argument) to do something bad. We should default to some deterministic value for the seed.
Attachments
Activity
Changes look fine. I added a single comment on GitHub, which you might consider — the gist is that some of the documentation you're removing is not actually wrong, so it might be worth leaving it in place (possibly with a “but be careful you understand the implications of this” caveat). Otherwise, good to go.
Thanks John Swinbank.
Reinstated the doc changes. Jenkins from before is all green. Not going to rerun Jenkins since the changes I made only involve docs, but checked that meas_base tests pass.
Merged to master.
Paul Price
added a comment - Thanks John Swinbank .
Reinstated the doc changes. Jenkins from before is all green. Not going to rerun Jenkins since the changes I made only involve docs, but checked that meas_base tests pass.
Merged to master. 
John Swinbank, would you take a look at this please?
Jenkins is just about all green.