Closed GegznaV closed 3 years ago
OK, I see that we do not build binaries for Windows and Mac at all, and the drat
repo contains only source packages. And these source packages have all the files needed to build a complete R package. So the problem is likely not what I anticipated before.
I'll have to delve into this issue more deeply.
Current builds pass without fixing anything I wrote above:
So now I think that the problem was somewhere else so I'll close this issue for now as irrelevant.
Right now the windows R 3.6 build is commented out and that may be why the checks pass. This is related to whether we want to support any R before 4.0 (on the Trello list). I am working on this and will report at the next video meeting.
Reactivated Windows + R 3.6 and all checks pass, must have been a transient issue.
This error is from
hySpc.read.jdx
repo:but it will arise in any package that depends on the development version of
hyperSpec
as binaries of this version are not present in ourdrat
repo calledpkg-repo
. Some issues may also arise while installing from the source by usingremotes::instal_github()
as the source on GitHub does not contain some files (like .Rd ones) which we do not track. So there are some points to consider:1) What CI strategy should we use? a) How many previous R versions should we use for the R CMD check? b) For how many old R versions should we build
hyperSpec
binaries? 2) Use simpler code vs. start tracking helper files: a) Should we still git-ignore some helper files (like.Rd
files) and create some more advanced code for CI procedures (which potentially run longer and make CI servers consume more energy); b) or start tracking the files and useremotes::install_github()
for easier installation ofhyperSpec
(devel version).There might be some other solutions that I'm not currently aware of.