[R-pkg-devel] Working with R-devel

[William Dunlap]

On Linux you can trace all the getenv calls with ltrace.  It only
works on ELF files, not sh scripts:

% R CMD ltrace -e getenv `R RHOME`/bin/*/R
libR.so->getenv("R_HOME")
               = "/home/R/R-3.4.3/lib64/R"
libR.so->getenv("R_TRANSLATIONS")
               = nil
libR.so->getenv("R_PLATFORM")
               = "x86_64-pc-linux-gnu"
libR.so->getenv("R_PAPERSIZE")
              = "letter"
libR.so->getenv("R_PAPERSIZE")
              = "letter"
libR.so->getenv("R_PRINTCMD")
               = "lpr"
libR.so->getenv("R_RD4PDF")
               = "times,hyper"
...

Bill Dunlap
TIBCO Software
wdunlap tibco.com

On Wed, Jan 24, 2018 at 6:11 AM, Dirk Eddelbuettel <edd at debian.org> wrote:

>
> On 24 January 2018 at 07:03, luke-tierney at uiowa.edu wrote:
> | If you are picking up an R 3.4.x that is trying to load a package
> | built by R-devel you should also see a warning that the package was
> | built under R version 3.5.0
>
> Yes, I am of course very aware of the issue.
>
> But we had to separate out R-release and R-devel for years, and so I did. I
> had separate libPaths (or, at least, leading component) for R-devel because
> of that.
>
> | If you are not, then you could be picking up an older R-devel from
> | before the internals change. Probably more likely, the Rcpp being
> | loaded at that point was built by 3.4.x, which would suggest library
> | path settings are messed up at that point, but not clear why. [It
> | might be possible and useful to provide more information about the
> | broken pieces in the error message but I won't have time to look any
> | time soon.]
> |
> | At the moment when it matters I usually use
> |
> | env R_LIBS=<my R-devel library> R CMD check ...
> |
> | for checks under R_devel
>
> It was the R_LIBS_USER (which used to work) to R_LIBS change that got plus
> the fact that my R-devel was picking up my R-release check.Renviron. The
> hint
> from Martin got rid of that.
>
> Sometimes I wonder if we should have an extra-verbose mode that picks up
> EVERY single Sys.getenv() and alike to make it easier to trace this. But
> run-time cost of the check is probably unattractive.
>
> Dirk
>
> --
> http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org
>
> ______________________________________________
> R-package-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel
>

	[[alternative HTML version deleted]]