[Bioc-devel] Debugging when test case passes in interactive session but fails R CMD check

Hi Cory,

Hello.

I've got a test case that completes in an interactive session but fails
when run via R CMD check. Has anyone run into this situation, and--if
so--found a way to reproduce the error in an interactive session?

Yes. When you run R CMD check, you are actually running a script
called "check" which can be found somewhere in R_HOME (exact location
varies depending on platform), which looks like this:

echo 'tools:::.check_packages()' | R_DEFAULT_PACKAGES= LC_COLLATE=C
"${R_HOME}/bin/R" --no-restore --slave --args ${args}

So you can simulate that with:

R_DEFAULT_PACKAGES= LC_COLLATE=C R --no-restore  --vanilla

Then in R:

tools:::.check_packages("myPkg") # you need to be in the right directory

Also, I'd be curious whether
R -f myPkg/tests/*.R
would also reproduce the error.

Or, in R:
library(RUnit)
source("mypkg/inst/unitTests/test_my_test_file.R")
test_name_of_my_test_function()

In this particular case, it seems asBam is writing a temp file differently
under R CMD check than interactively. This is my current hunch since I've
hit issues on my system with file.rename (which asBam uses) when the "from"
and "to" arg  refer to files on different disks. This could be going on
here if R CMD check sets ?TMPDIR?, ?TMP? or ?TEMP? differently than in an
interactive session.

Are you using something other than tempfile() to create the temp
files? If so, do you get different results if you use tempfile()?

Any thoughts on how to debug failing test cases in general or this issue in
particular?

Let us know if the above helps.

Dan

Thanks much for the assistance.

-Cory

        [[alternative HTML version deleted]]

_______________________________________________
Bioc-devel at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/bioc-devel