An update: The RStudio bug that causes problems here is triggered by setting "Automatically activate project-local Python environments" in the Python global option. Unsetting that checkbox stops /usr/local/bin from being prepended.

I've decided not to try to work around this in rgl's configure script.

Duncan Murdoch


On 2024-06-15 2:13 p.m., Duncan Murdoch wrote:
RStudio is causing problems for rgl builds.  In particular, on my M3
system, I have installed the arm64 binaries and tools in
/opt/R/arm64/bin as recommended, and builds work fine if I run them in a
shell with that directory first in the PATH.

However, when I try to do builds in RStudio, they fail.  I think this is
because it prepends /usr/local/bin to my PATH before configuring, and
the rgl configure script finds the x86_64 versions of some files before
the arm64 ones, and they don't work.  So then rgl thinks I don't have
X11 libs available, and builds as though I don't.

This isn't the right place to ask how to fix RStudio.  I've already done
that in one of their forums, and I already know how to build rgl outside
of it.  However, I do get complaints from users who try to build rgl and
have the builds fail (e.g. https://github.com/dmurdoch/rgl/issues/423),
and I'd like to make the configure script resilient against this sort of
problem.  (And this might not be RStudio specific:  there are lots of
ways to end up with a PATH that doesn't have things in the right order.)

So here are my questions:

   - What is the correct way in a package configure script (or
configure.ac, since I use autoconf) to determine that the build is
happening on an arm64 Mac?

   - Is it correct in a configure script to force the arm64 directory to
be first?  If so, how do I do that?

   - How do I determine if the user really did install the arm64 files to
/opt/R/arm64/bin?

   - Should I really be doing this, or is my system broken since my
/usr/local/bin directory contains things it shouldn't, or should I just
tell people to put /opt/R/arm64/bin first in their PATH?

Duncan Murdoch

_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-mac

Reply via email to