[R-pkg-devel] R package test breakage catch 22

[Jonathon Love]

hi,

i currently face the following situation:

my package, package A, depends on package B. i have unit tests in 
package A which expect results, which are in part as a result of 
calculations from package B.


package B has made some changes, which leads to slightly different 
results being calculated, and now package A tests fail.


now it turns out that the CRAN submission process has picked up that the 
new B, causes tests to fail on A, and so B can't be accepted by CRAN.


now i can update my tests so they expect the *new* values, but obviously 
these will fail and won't be accepted by CRAN, because the new B isn't 
on CRAN yet.


i can't update A because of old B, and B can't be updated because of old A.

how does one solve this catch-22?

with thanks

jonathon

__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel

Re: [R-pkg-devel] R package test breakage catch 22

[Johannes Ranke]

> how does one solve this catch-22?

I see these possibilities

a) reduce the precision of the tests in A, so the test passes with both 
versions of B

b) a submission of both packages at the same time, with a message in the 
comments, explaining to CRAN why it is necessary that both are published at 
the same time (and maybe in more detail why the expected test results 
changed...)

c) merge the packages into one

Cheers,

Johannes

__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel