Closed drnickisaac closed 3 years ago
Ok, I can see that the fail on testvisitsSummary()
is because I didn't update the NAMESPACE. Now done locally.
In Rstudio, in the build settings, you can tell roxygen to rebuild the namespace on build and install, this will help to avoid having that problem i the future
@drnickisaac do you need me to look at testdetection_phenology issue?
I'd like your advice on how to proceed, please. It's the obvious Travis fail. As I said above, I've had the same fail for a month or two on my local machine but it's never triggered Travis before.
Yeah I remember seeing the testdetection_phenology one before... not sure where it comes from. I'll have a 10min look
Running the tests locally I now see the error_status
issue noted earlier today.
Ignore previous comment - turns out I had not pulled down Tom's curly bracket fix.
Observation: the values returned by testdetection_phenology.r
are different on Travis vs my Mac.
Could this be associated with the random number seed?
Indeed. I'm now forcing the seed before each call this seems to fix it
Is it related to this, when I "Check Package":
x .Random.seed() masks sparta::.Random.seed()
?
No idea what that is about!
i get Error: could not find function "arrange"
I've fixed that locally now - see above.
did you push fix to your fork
Not yet: I thought I'd wait a few minutes to see if you have a brilliant solution for the random number seed. Then it only stats Travis once.
PR coming now
ok, I will test locally before pushing up.
PR in. Have a great weekend, I'm off to get a drink!
Can we first assess who decides whether to accept my PR, and when, and what it means for the Mayflies?
Really interesting: after this switch the tests still fail locally, but by a far smaller amount! For now I will push, merge PR & ask Travis's opinion. Backup option would be to increase the tolerance on these tests.
Mayflies that I am running are using your fork
Dang, there must be a random seed difference somewhere. I lowered the tolerance in that last edit, feel free to do more if you think it is acceptable. Really have to go now, sorry
Travis says Yes!
Some new test fails have tripped on Travis when merging for 0.2.11.
testdetection_phenology()
: these are the same four failures that I routinely get when compiling sparta on my Mac, but until now these have not been tripped on Travis. They relate to small but consistent differences in the parameter values coming back from this small test dataset.testvisitsSummary()
, which cannot find the functionarrange()
. I actually got this on my own machine yesterday, so I added#' @importFrom dplyr arrange
at the top ofvisitSummary()
, but somehow this is not enough.None of the errors that tripped before Christmas (#208) have resurfaced. Maybe that was a temporary R versioning problem?