#18691: Trac #17572 breaks the installation of several R packages.
-------------------------------------+-------------------------------------
Reporter: charpent | Owner:
Type: defect | Status: needs_review
Priority: major | Milestone: sage-6.8
Component: packages: | Resolution:
standard | Merged in:
Keywords: r-project | Reviewers:
Authors: charpent | Work issues:
Report Upstream: N/A | Commit:
Branch: | 7288cdd4e8d2a04f4163dfac1249898d01e50492
u/charpent/trac__17572_breaks_the_installation_of_several_r_packages_|
Stopgaps:
Dependencies: |
-------------------------------------+-------------------------------------
Changes (by charpent):
* status: new => needs_review
* commit: => 7288cdd4e8d2a04f4163dfac1249898d01e50492
Comment:
The proposed patch is a "damage control" patch limiting the effect of
#17572 to Darwin OS.
Rationale : the problem (mis-)fixed by #17572 seems to be limited to
machines running clang. It seems, from reading pages pointed to in the
discussion of this patch, that this happens on Mac OS X due to slightly
suboptimal apple development software.
As already seen, this "solution" breaks the installation of several R
packages, which explicitly check for systemwide {{{Makevars}}} or user-
specific {{{Makevars.user}}} set by way of R_HOME and R_PROFILE
environment variables.
Our current solution is to unset these variables and export the variables
{{{R_MAKEVARS_SITE}}} and {{{R_MAKEVARS_SITE}}}, pointing to convenient
Sage-specific versions of similarly named files. As far as I can tell, we
do **not** have such files, which breaks the build of some R packages.
I do not have a Mac, and I do not use clang. Therefore, I'm ill-qualified
to propose a solution. A **stopgap** could be to create such **empty**
files at R installation, and keep the patch as it curently is.
However, I do not understand **why** clang breaks when the relevant
variables are not defined, and why gcc breaks when they are defined as
pointing to nonexistent files.
So we have a few courses of action :
1. keep #17572 as it is now, and add the creation of empty Makevars.site
and Makevars.user somewhere at the end of the R installation script. Not
that easy,..
2. Limit #17572 to Darwin machines (= current damage control patch) and
find some apple-specific way to fix the problem of non-installing R
packages.
3. Revert #17572 entierely and find some other way to induce clang that it
can in fact run without these environment variables.
In any case, fixing this one needs the expertise of someone having access
to a clang-plagued Mac OS X machine. I can't test a solution of this
combination.
I put this "damage control" patch as {{{needs_review}}}. Feel free to pull
it back to {{{needs_work}}} if you have some proposal to implement
solutions 1 or 3. Or something else entirely...
----
New commits:
||[http://git.sagemath.org/sage.git/commit/?id=7288cdd4e8d2a04f4163dfac1249898d01e50492
7288cdd]||{{{Damage control : limit #17572 to Darwin OS.}}}||
--
Ticket URL: <http://trac.sagemath.org/ticket/18691#comment:3>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-trac.
For more options, visit https://groups.google.com/d/optout.
