[Bioc-devel] Replicating Windows build error for BioC package submission

Benjamin Story @tory@benj@m|n @end|ng |rom gm@||@com
Mon Jul 5 20:59:57 CEST 2021


Hi all,



I’m trying to debug why my package (mitoClone2) submission fails the R CMD
build step on Windows (Linux and Mac are working) specifically on the
Bioconductor test machine.



My first steps were to try to reproduce the output on my personal Windows
computer by installing the package from source on a fresh install of R
4.0.1. After installing Rtools40, I encountered different errors suggesting
certain libraries were missing or non-accessible. I then installed
rtools-packages in order to install the missing libraries, openssl and
curl. There was still an error on install so I manually added -lcrypt32 to
my PKG_LIBS variable in my Makevars.win (see
https://stackoverflow.com/questions/30126970/linking-issue-with-libcrypto-undefined-reference-to-imp-certfreecertificate).
The package now builds on my personal machine *without any errors* both
using R CMD build and install.packages(…,from='source').



To test further, I submitted the resulting tarball from R CMD build to
https://win-builder.r-project.org/ (the same -lcrypt32 fix was needed) and
now builds successfully there for R-devel (
https://win-builder.r-project.org/Lb1w7hJiUfJl/00check.log - link expires
in less than 72 hours). R-release installs but hangs at some point during
vignette rebuilding but the C++ functions are compiled successfully (
https://win-builder.r-project.org/QbI7yqY0oDNe/).



I’m not really a developer and am thus unsure as to where to go from here
for further troubleshooting. Any assistance or suggestions would be greatly
appreciated. You can see my current Makevars.win here:
https://github.com/benstory/mitoClone2/blob/main/src/Makevars.win Parts of
the code were duplicated for testing but they don't affect the build step.


Thanks,

Ben

	[[alternative HTML version deleted]]



More information about the Bioc-devel mailing list