I haven't tested on Windows yet, but I will try to find some time to do so.
The first thing would be to check if this .Renviron file is read at all, e.g. if you define a variable `MYPATH='aaa'` can you see it from inside RStudio, or is only `PATH` overwritten?
Other environment variables can be set normally. It appears that only
PATH is overwritten, as described in the GitHub issue I linked to earlier.
I would also check that RStudio is loading the right version of R (e.g. sometimes after installing R4.3.2 RStudio is still loading R4.3.1). That's particularly meaningful if you are using a .Renviron in R_HOME/etc, not if it's in your `~`.
I am using RStudio Version 2023.09.1+494 (2023.09.1+494) on Apple silicon macOS 13.5.2, which loads R v4.3.2. Looks ok to me.
I would also check for a project-specific file. If you create a new project, does RStudio still not take into account your .Renviron?
There is no project, nor a project-specific file. This happens with no project open, and the wd set to "~" (by default).
I can tell that the
.Rprofile are read - other effects are persistent, including your suggestion of a custom
MYPATH environment variable. Also (as described in my comment on the GitHub issue), If I add this line to the end of
it shows the expected modified PATH in the console output at the start of a session:
But once the session starts, the same command (
Sys.getenv("PATH")) shows a completely different output, without any of the changes applied by
.Rprofile, as far as the
PATH environment variable is concerned. So it looks very much like RStudio is overwriting
PATH with something else, after loading and running
.Rprofile. I know that GUI applications inherit their environment variables (such as
PATH) differently than shells in macOS, but the same thing does not happen in the base R GUI, and I'm wondering what R Studio is doing differently (and why?).