R Installation and Administration

Sources, binaries and documentation for R can be obtained via CRAN, the “Comprehensive R Archive Network” whose current members are listed at https://CRAN.R-project.org/mirrors.html.

1.1 Getting and unpacking the sources

The simplest way is to download the most recent R-x.y.z.tar.gz file, and unpack it with

on systems that have a suitable¹ tar installed. On other systems you need to have the gzip program installed, when you can use

The pathname of the directory into which the sources are unpacked should not contain spaces, as most make programs (and specifically GNU make) do not expect spaces.

If you want the build to be usable by a group of users, set umask before unpacking so that the files will be readable by the target group (e.g., umask 022 to be usable by all users). Keep this setting of umask whilst building and installing.

If you use a fairly recent GNU version of tar and do this as a root account (which on Windows includes accounts with administrator privileges) you may see many warnings about changing ownership. In which case you can use

and perhaps also include the option --no-same-permissions. (These options can also be set in the TAR_OPTIONS environment variable: if more than one option is included they should be separated by spaces.)

1.2 Getting patched and development versions

A patched version of the current release, ‘r-patched’, and the current development version, ‘r-devel’, are available as daily tarballs and via access to the R Subversion repository. (For the two weeks prior to the release of a minor (4.x.0) version, ‘r-patched’ tarballs may refer to beta/release candidates of the upcoming release, the patched version of the current release being available via Subversion.)

The tarballs are available from https://stat.ethz.ch/R/daily/. Download R-patched.tar.gz or R-devel.tar.gz (or the .tar.bz2 versions) and unpack as described in the previous section. They are built in exactly the same way as distributions of R releases.

1.2.1 Using Subversion and rsync

Sources are also available via https://svn.R-project.org/R/, the R Subversion repository. If you have a Subversion client (see https://subversion.apache.org/), you can check out and update the current ‘r-devel’ from https://svn.r-project.org/R/trunk/ and the current ‘r-patched’ from ‘https://svn.r-project.org/R/branches/R-x-y-branch/’ (where x and y are the major and minor number of the current released version of R). E.g., use

to check out ‘r-devel’ into directory path (which will be created if necessary). The alpha, beta and RC versions of an upcoming x.y.0 release are available from ‘https://svn.r-project.org/R/branches/R-x-y-branch/’ in the four-week period prior to the release.

Note that ‘https:’ is required², and that the SSL certificate for the Subversion server of the R project should be recognized as from a trusted source.

Note that retrieving the sources by e.g. wget -r or svn export from that URL will not work (and will give a error early in the make process): the Subversion information is needed to build R.

The Subversion repository does not contain the current sources for the recommended packages, which can be obtained by rsync or downloaded from CRAN. To use rsync to install the appropriate sources for the recommended packages, run ./tools/rsync-recommended from the top-level directory of the R sources.

If downloading manually from CRAN, do ensure that you have the correct versions of the recommended packages: if the number in the file VERSION is ‘x.y.z’ you need to download the contents of ‘https://CRAN.R-project.org/src/contrib/dir’, where dir is ‘x.y.z/Recommended’ for r-devel or x.y-patched/Recommended for r-patched, respectively, to directory src/library/Recommended in the sources you have unpacked. After downloading manually you need to execute tools/link-recommended from the top level of the sources to make the requisite links in src/library/Recommended. A suitable incantation from the top level of the R sources using wget might be (for the correct value of dir)

2 Installing R under Unix-alikes

R will configure and build under most common Unix and Unix-alike platforms including ‘cpu-*-linux-gnu’ for the ‘alpha’, ‘arm64’, ‘hppa’, ‘ix86’, ‘m68k’, ‘mips’, ‘mipsel’#, ‘ppc64’, ‘s390x’, ‘sparc64’, and ‘x86_64’ CPUs, ‘x86_64-apple-darwin’ and ‘i386-sun-solaris’ as well as perhaps (it is tested less frequently on these platforms) ‘i386-*-freebsd’, ‘x86_64-*-freebsd’, ‘i386-*-netbsd’, ‘x86_64-*-openbsd’ and ‘powerpc-ibm-aix6*’

In addition, binary distributions are available for some common Linux distributions and for macOS. See the FAQ for current details. These are installed in platform-specific ways, so for the rest of this chapter we consider only building from the sources.

Cross-building is not possible: installing R builds a minimal version of R and then runs many R scripts to complete the build.

2.1 Simple compilation

First review the essential and useful tools and libraries in Essential and useful other programs under a Unix-alike, and install those you want or need. Ensure that either the environment variable TMPDIR is either unset (and /tmp exists and can be written in and scripts can be executed from) or points to the absolute path to a valid temporary directory (one from which execution of scripts is allowed) which does not contain spaces.³

Choose a directory to install the R tree (R is not just a binary, but has additional data sets, help files, font metrics etc). Let us call this place R_HOME. Untar the source code. This should create directories src, doc, and several more under a top-level directory: change to that top-level directory (At this point North American readers should consult Setting paper size.) Issue the following commands:

(See Using make if your make is not called ‘make’.) Users of Debian-based 64-bit systems⁴ may need

Failures are not necessarily problems as they might be caused by missing functionality, but you should look carefully at any reported discrepancies. (Some non-fatal errors are expected in locales that do not support Latin-1, in particular in true C locales and non-UTF-8 non-Western-European locales.) A failure in tests/ok-errors.R may indicate inadequate resource limits (see Running R).

see file tests/README and Testing a Unix-alike Installation for the possibilities of doing this in parallel. Note that these checks are only run completely if the recommended packages are installed.

If the configure and make commands execute successfully, a shell-script front-end called R will be created and copied to R_HOME/bin. You can link or copy this script to a place where users can invoke it, for example to /usr/local/bin/R. You could also copy the man page R.1 to a place where your man reader finds it, such as /usr/local/man/man1. If you want to install the complete R tree to, e.g., /usr/local/lib/R, see Installation. Note: you do not need to install R: you can run it from where it was built.

You do not necessarily have to build R in the top-level source directory (say, TOP_SRCDIR). To build in BUILDDIR, run

and so on, as described further below. This has the advantage of always keeping your source tree clean and is particularly recommended when you work with a version of R from Subversion. (You may need GNU make to allow this, and you will need no spaces in the path to the build directory. It is unlikely to work if the source directory has previously been used for a build.)

There are many settings which can be customized when building R and most are described in the file config.site in the top-level source directory. This can be edited, but for an installation using BUILDDIR it is better to put the changed settings in a newly-created file config.site in the build directory.

Now rehash if necessary, type R, and read the R manuals and the R FAQ (files FAQ or doc/manual/R-FAQ.html, or https://CRAN.R-project.org/doc/FAQ/R-FAQ.html which always has the version for the latest release of R).

Note: if you already have R installed, check that where you installed R replaces or comes earlier in your path than the previous installation. Some systems are set up to have /usr/bin (the standard place for a system installation) ahead of /usr/local/bin (the default place for installation of R) in their default path, and some do not have /usr/local/bin on the default path.

2.2 Help options

R by default provides help pages as plain text displayed in a pager, with the options (see the help for help) of displaying help as HTML or PDF.

By default HTML help pages are created when needed rather than being built at install time.

If you need to disable the server and want HTML help, there is the option to build HTML pages when packages are installed (including those installed with R). This is enabled by the configure option --enable-prebuilt-html. Whether R CMD INSTALL (and hence install.packages) pre-builds HTML pages is determined by looking at the R installation and is reported by R CMD INSTALL --help: it can be overridden by specifying one of the INSTALL options --html or --no-html.

The server is disabled by setting the environment variable R_DISABLE_HTTPD to a non-empty value, either before R is started or within the R session before HTML help (including help.start) is used. It is also possible that system security measures will prevent the server from being started, for example if the loopback interface has been disabled. See ?tools::startDynamicHelp for more details.

2.3 Making the manuals

You will not be able to build any of these unless you have texi2any version 5.1 or later installed, and for PDF you must have texi2dvi and texinfo.tex installed (which are part of the GNU texinfo distribution but are, especially texinfo.tex, often made part of the TeX package in re-distributions). The path to texi2any can be set by macro ‘TEXI2ANY’ in config.site. NB: texi2any requires perl.

The PDF versions can be viewed using any recent PDF viewer: they have hyperlinks that can be followed. The info files are suitable for reading online with Emacs or the standalone GNU info program. The PDF versions will be created using the paper size selected at configuration (default ISO a4): this can be overridden by setting R_PAPERSIZE on the make command line, or setting R_PAPERSIZE in the environment and using make -e. (If re-making the manuals for a different paper size, you should first delete the file doc/manual/version.texi. The usual value for North America would be ‘letter’.)

There are some issues with making the PDF reference manual, fullrefman.pdf or refman.pdf. The help files contain both ISO Latin1 characters (e.g. in text.Rd) and upright quotes, neither of which are contained in the standard LaTeX Computer Modern fonts. We have provided four alternatives:

The default can be overridden by setting the environment variable R_RD4PDF. (On Unix-alikes, this will be picked up at install time and stored in etc/Renviron, but can still be overridden when the manuals are built, using make -e.) The usual⁶ default value for R_RD4PDF is ‘times,inconsolata,hyper’: omit ‘hyper’ if you do not want hyperlinks (e.g. for printing the manual) or do not have LaTeX package hyperref, and omit ‘inconsolata’ if you do not have LaTeX package inconsolata installed.

Further options, e.g for hyperref, can be included in a file Rd.cfg somewhere on your LaTeX search path. For example, if you prefer to hyperlink the text and not the page number in the table of contents use

Ebook versions of most of the manuals in one or both of .epub and .mobi formats can be made by running in doc/manual one of

This requires ebook-convert from Calibre, or from most Linux distributions. If necessary the path to ebook-convert can be set as make macro EBOOK to by editing doc/manual/Makefile (which contains a commented value suitable for macOS) or using make -e.

2.4 Installation

To ensure that the installed tree is usable by the right group of users, set umask appropriately (perhaps to ‘022’) before unpacking the sources and throughout the build process.

(or, when building outside the source, TOP_SRCDIR/configure, etc) have been completed successfully, you can install the complete R tree to your system by typing

A parallel make can be used (but run make before

make
install

). Those using GNU make 4.0 or later may want to use make -j n -O to avoid interleaving of output.

where prefix is determined during configuration (typically /usr/local) and can be set by running configure with the option --prefix, as in

where the value should be an absolute path. This causes

make
install

to install the R script to /where/you/want/R/to/go/bin, and so on. The prefix of the installation directories can be seen in the status message that is displayed at the end of configure. The installation may need to be done by the owner of prefix, often a root account.

More precise control is available at configure time via options: see configure --help for details. (However, most of the ‘Fine tuning of the installation directories’ options are not used by R.)

Configure options --bindir and --mandir are supported and govern where a copy of the R script and the man page are installed.

The configure option --libdir controls where the main R files are installed: the default is ‘eprefix/LIBnn’, where eprefix is the prefix used for installing architecture-dependent files, defaults to prefix, and can be set via the configure option --exec-prefix.

Each of bindir, mandir and libdir can also be specified on the make install command line (at least for GNU make).

The configure or make variables rdocdir and rsharedir can be used to install the system-independent doc and share directories to somewhere other than libdir. The C header files can be installed to the value of rincludedir: note that as the headers are not installed into a subdirectory you probably want something like rincludedir=/usr/local/include/R-4.1.0.

If you want the R home to be something other than libdir/R, use rhome: for example

If you have made R as a shared/static library you can install it in your system’s library directory by

where prefix is optional, and libdir will give more precise control.⁷ However, you should not install to a directory mentioned in LDPATHS (e.g. /usr/local/lib64) if you intend to work with multiple versions of R, since that directory may be given precedence over the lib directory of other R installations.

will install stripped executables, and on platforms where this is supported, stripped libraries in directories lib and modules and in the standard packages.

Note that installing R into a directory whose path contains spaces is not supported, and some aspects (such as installing source packages) will not work.

Once again, it is optional to specify prefix, libdir or rhome (the PDF manuals are installed under the R home directory).

More precise control is possible. For info, the setting used is that of infodir (default prefix/info, set by configure option --infodir). The PDF files are installed into the R doc tree, set by the make variable rdocdir.

A staged installation is possible, that it is installing R into a temporary directory in order to move the installed tree to its final destination. In this case prefix (and so on) should reflect the final destination, and DESTDIR should be used: see https://www.gnu.org/prep/standards/html_node/DESTDIR.html.

You can optionally install the run-time tests that are part of make check-all by

2.5 Uninstallation

optionally specifying prefix etc in the same way as specified for installation.

This will also uninstall any installed manuals. There are specific targets to uninstall info and PDF manuals in file doc/manual/Makefile.

Target uninstall-tests will uninstall any installed tests, as well as removing the directory tests containing the test results.

2.6 Sub-architectures

Some platforms can support closely related builds of R which can share all but the executables and dynamic objects. Examples include builds under Linux and Solaris for different CPUs or 32- and 64-bit builds.

R supports the idea of architecture-specific builds, specified by adding ‘r_arch=name’ to the configure line. Here name can be anything non-empty, and is used to name subdirectories of lib, etc, include and the package libs subdirectories. Example names from other software are the use of sparcv9 on Sparc Solaris and 32 by gcc on ‘x86_64’ Linux.

If you have two or more such builds you can install them over each other (and for 32/64-bit builds on one architecture, one build can be done without ‘r_arch’). The space savings can be considerable: on ‘x86_64’ Linux a basic install (without debugging symbols) took 74Mb, and adding a 32-bit build added 6Mb. If you have installed multiple builds you can select which build to run by

R CMD INSTALL will detect if more than one build is installed and try to install packages with the appropriate library objects for each. This will not be done if the package has an executable configure script or a src/Makefile file. In such cases you can install for extra builds by

If you want to mix sub-architectures compiled on different platforms (for example ‘x86_64’ Linux and ‘i686’ Linux), it is wise to use explicit names for each, and you may also need to set libdir to ensure that they install into the same place.

When sub-architectures are used the version of Rscript in e.g. /usr/bin will be the last installed, but architecture-specific versions will be available in e.g. /usr/lib64/R/bin/exec${R_ARCH}. Normally all installed architectures will run on the platform so the architecture of Rscript itself does not matter. The executable Rscript will run the R script, and at that time the setting of the R_ARCH environment variable determines the architecture which is run.

Sub-architectures are also used on Windows, but by selecting executables within the appropriate bin directory, R_HOME/bin/i386 or R_HOME/bin/x64. For backwards compatibility there are executables R_HOME/bin/R.exe and R_HOME/bin/Rscript.exe: these will run an executable from one of the subdirectories, which one being taken first from the R_ARCH environment variable, then from the --arch command-line option⁸ and finally from the installation default (which is 32-bit for a combined 32/64 bit R installation).

2.6.1 Multilib

For some Linux distributions⁹, there is an alternative mechanism for mixing 32-bit and 64-bit libraries known as multilib. If the Linux distribution supports multilib, then parallel builds of R may be installed in the sub-directories lib (32-bit) and lib64 (64-bit). The build to be run may then be selected using the setarch command. For example, a 32-bit build may be run by

The setarch command is only operational if both 32-bit and 64-bit builds are installed. If there is only one installation of R, then this will always be run regardless of the architecture specified by the setarch command.

There can be problems with installing packages on the non-native architecture. It is a good idea to run e.g. setarch i686 R for sessions in which packages are to be installed, even if that is the only version of R installed (since this tells the package installation code the architecture needed).

There is a potential problem with packages using Java, as the post-install for a ‘i686’ RPM on ‘x86_64’ Linux reconfigures Java and will find the ‘x86_64’ Java. If you know where a 32-bit Java is installed you may be able to run (as root)

When this mechanism is used, the version of Rscript in e.g. /usr/bin will be the last installed, but an architecture-specific version will be available in e.g. /usr/lib64/R/bin. Normally all installed architectures will run on the platform so the architecture of Rscript does not matter.

2.7 Other Options

There are many other installation options, most of which are listed by configure --help. Almost all of those not listed elsewhere in this manual are either standard autoconf options not relevant to R or intended for specialist uses by the R developers.

One that may be useful when working on R itself is the option --disable-byte-compiled-packages, which ensures that the base and recommended packages are not byte-compiled. (Alternatively the (make or environment) variable R_NO_BASE_COMPILE can be set to a non-empty value for the duration of the build.)

Option --with-internal-tzcode makes use of R’s own code and copy of the IANA database for managing timezones. This will be preferred where there are issues with the system implementation, usually involving times after 2037 or before 1916. An alternative time-zone directory¹⁰ can be used, pointed to by environment variable TZDIR: this should contain files such as Europe/London. On all tested OSes the system timezone was deduced correctly, but if necessary it can be set as the value of environment variable TZ.

2.7.1 Debugging Symbols

By default, configure adds a flag (usually -g) to the compilation flags for C, Fortran and CXX sources. This will slow down compilation and increase object sizes of both R and packages, so it may be a good idea to change those flags (set ‘CFLAGS’ etc in config.site before configuring, or edit files Makeconf and etc/Makeconf between running configure and make).

Having debugging symbols available is useful both when running R under a debugger (e.g., R -d gdb) and when using sanitizers and valgrind, all things intended for experts.

Debugging symbols (and some others) can be ‘stripped’ on installation by using

How well this is supported depends on the platform: it works best on those using GNU binutils. On ‘x86_64’ Linux a typical reduction in overall size was from 92MB to 66MB. On macOS debugging symbols are not by default included in .dylib and .so files, so there is negligible difference.

2.7.2 OpenMP Support

By default configure searches for suitable flags¹¹ for OpenMP support for the C, C++ (default standard) and Fortran compilers.

Only the C result is currently used for R itself, and only if MAIN_LD/DYLIB_LD were not specified. This can be overridden by specifying

Use for packages has similar restrictions (involving SHLIB_LD and similar: note that as Fortran code is by default linked by the C (or C++) compiler, both need to support OpenMP) and can be overridden by specifying some of

Setting these to an empty value will disable OpenMP for that compiler (and configuring with --disable-openmp will disable all detection¹² of OpenMP). The configure detection test is to compile and link a standalone OpenMP program, which is not the same as compiling a shared object and loading it into the C program of R’s executable. Note that overridden values are not tested.

2.7.3 C++ Support

C++ is not used by R itself, but support is provided for installing packages with C++ code via make macros defined in file etc/Makeconf (and with explanations in file config.site):

The macros CXX etc are those used by default for C++ code. configure will attempt to set the rest suitably, choosing for CXXSTD and CXX11STD a suitable flag such as -std=c++11 for C++11 support (which is required if C++ is to be supported at all). Similarly, configure will if possible choose for CXX14STD a flag¹³ such as -std=c++14 for C++14 support, -std=c++17 or -std=c++1z for support for C++17 and -std=c++20 or -std=c++2a for support for C++20.. The inferred values can be overridden in file config.site or on the configure command line: user-supplied values will be tested by compiling some C++11/14/17/20 code.

It may be that there is no suitable flag for C++14/17/20 support with the default compiler, in which case a different compiler could be selected for CXX14 or CXX17 or CXX20 with its corresponding flags.

The -std flag is supported by the GCC, clang++, Intel and Solaris compilers (the latter from version 12.4). Currently accepted values are (plus some synonyms)

‘Standards’ for g++ starting with ‘gnu’ enable ‘GNU extensions’: what those are is hard to track down.

For the use of C++11 and later in R packages see the ‘Writing R Extensions’ manual. Prior to R 3.6.0 the default C++ standard was that of the compiler used: currently it is C++11: this can be overridden by setting ‘CXXSTD’ when R is configured, for example to ‘-std=gnu++14’.

2.7.4 Link-Time Optimization

There is support for using link-time optimization (LTO) if the toolchain supports it: configure with flag --enable-lto. When LTO is enabled it is used for compiled code in add-on packages unless the flag --enable-lto=R is used¹⁴.

The main benefit seen to date from LTO has been detecting long-standing bugs in the ways packages pass arguments to compiled code and between compilation units. Benchmarking in 2020 with gcc/gfortran 10 showed gains of a few percent in increased performance and reduction in installed size for builds without debug symbols, but large size reductions for some packages¹⁵ with debug symbols. (Performance and size gains are said to be most often seen in complex C++ builds.)

Whether toolchains support LTO is often unclear: all of the C compiler, the Fortran compiler¹⁶ and linker have to support it, and support it by the same mechanism (so mixing compiler families may not work and a non-default linker may be needed). It has been supported by the GCC and LLVM projects for some years with diverging implementations.

LTO support was added in 2011 for GCC 4.5 on Linux but was little used before 2019: compiler support has steadily improved over those years and --enable-lto=R is nowadays used for some routine CRAN checking.

Unfortunately --enable-lto may be accepted but silently do nothing useful if some of the toolchain does not support LTO: this is less common than it once was.

Various macros can be set in file config.site to customize how LTO is used. If the Fortran compiler is not of the same family as the C/C++ compilers, set macro ‘LTO_FC’ (probably to empty). Macro ‘LTO_LD’ can be used to select an alternative linker should that be needed.

2.7.4.1 LTO with GCC

This has been tested on Linux with gcc/gfortran 8 and later: that needed setting (e.g. in config.site)

For non-system compilers or if those wrappers have not been installed one may need something like

amd NM may be needed to be set analogously. (If using an LTO-enabled build to check packages, set environment variable UserNM¹⁷ to ‘gcc-nm’.)

With GCC 5 and later it is possible to parallelize parts of the LTO linking process: set the make macro ‘LTO’ to something like ‘LTO=-flto=8’ (to use 8 threads), for example in file config.site.

Under some circumstances and for a few packages, the PIC flags have needed overriding on Linux with GCC 9 and later: e.g use in config.site:

We suggest only using these if the problem is encountered (it was not seen on CRAN with GCC 10 at the time of writing).

Note that R may need to be re-compiled after even a minor update to the compiler (e.g. from 10.1 to 10.2) but this may not be clear from confused compiler messages.

2.7.4.2 LTO with LLVM

LLVM supports another type of LTO called ‘Thin LTO’ as well as a similar implementation to GCC, sometimes called ‘Full LTO’. (See https://clang.llvm.org/docs/ThinLTO.html.) Currently the only LLVM compiler relevant to R is clang for which this can be selected by setting macro ‘LTO=-flto=thin’. LLVM has

(but macOS does not, and these are not needed there). Where the linker supports a parallel backend for Thin LTO this can be specified via the macro ‘LTO_LD’: see the URL above for per-linker settings and further linking optimizations.)

to use Thin LTO with 4 threads for C/C++ code, but skip LTO for Fortran code compiled with gfortran.

It is said to be particularly beneficial to use -O3 for clang in conjunction with LTO.

2.7.4.3 LTO for package checking

LTO effectively compiles all the source code in a package as a single compilation unit and so allows the compiler (with sufficient diagnostic flags such as -Wall) to check consistency between what are normally separate compilation units.

With gcc/gfortran 9.x and later¹⁸ LTO will flag inconsistencies in calls to Fortran subroutines/functions, both between Fortran source files and between Fortran and C/C++. gfortran 8.4, 9.2 and later can help understanding these by extracting C prototypes from Fortran source files with option -fc-prototypes-external, e.g. that (at the time of writing) Fortran LOGICAL corresponds to int_least32_t * in C.

On some systems it is possible to build the BLAS, LINPACK and LAPACK support as static libraries containing intermediate-level code that LTO will compile for all objects linked against these libraries, by configuring R with --enable-lto=check. This checks the consistency of calls to BLAS/LINPACK/LAPACK in any packages installed using the build. NB: as its name suggests, this option is intended only for checking installation of R and packages: it includes these library routines (those called directly and all they depend on) in each package. This is unlikely to work in conjunction with non-default options for BLAS and LAPACK, and ‘linking’ against these libraries may be very slow.

2.8 Testing an Installation

Full post-installation testing is possible only if the test files have been installed with

If this has been done, two testing routes are available. The first is to move to the home directory of the R installation (as given by R RHOME or from R as R.home()) and run

and other useful targets are test-BasePackages and test-Recommended to run tests of the standard and recommended packages (if installed) respectively.

This re-runs all the tests relevant to the installed R (including for example the code in the package vignettes), but not for example the ones checking the example code in the manuals nor making the standalone Rmath library. This can occasionally be useful when the operating environment has been changed, for example by OS updates or by substituting the BLAS (see Shared BLAS).

Parallel checking of packages may be possible: set the environment variable TEST_MC_CORES to the maximum number of processes to be run in parallel. This affects both checking the package examples (part of make check) and package sources (part of

make
check-devel

and make check-recommended). It does require a make command which supports the make -j n option: most do but on Solaris you need to select GNU make or dmake.

runs the basic tests and then all the tests on the standard and recommended packages. These tests can be run from anywhere: the basic tests write their results in the tests folder of the R home directory and run fewer tests than the first approach: in particular they do not test things which need Internet access—that can be tested by

These tests work best if diff (in Rtools*.exe for Windows users) is in the path.

It is possible to test the installed packages (but not their package-specific tests) by testInstalledPackages even if make install-tests was not run.

Note that the results may depend on the language set for times and messages: for maximal similarity to reference results you may want to try setting (before starting the R session)

3 Installing R under Windows

The bin/windows directory of a CRAN site contains binaries for a base distribution and a large number of add-on packages from CRAN to run on 32- or 64-bit Windows (Windows 7 and later are tested; XP is known to fail some tests) on ‘ix86’ and ‘x86_64’ CPUs.

Your file system must allow long file names (as is likely except perhaps for some network-mounted systems). If it does not also support conversion to short name equivalents (a.k.a. DOS 8.3 names), then R must be installed in a path that does not contain spaces.

Installation is via the installer -win.exe. Just double-click on the icon and follow the instructions. When installing on a 64-bit version of Windows the options will include 32- or 64-bit versions of R (and the default is to install both). You can uninstall R from the Control Panel.

You will be asked to choose a language for installation: that choice applies to both installation and un-installation but not to running R itself.

3.1 Building from source

It is possible to use other toolchains (including that used for R 3.6.3, ‘MSYS2’ and ‘MXE’ (M cross environment)) to build R, but this manual only documents that used for binary distributions of R 4.0.x.

3.1.1 The Windows toolset

The binary distribution of R is currently built with tools distributed in rtools40-x86_64.exe. The webpage https://cran.r-project.org/bin/windows/Rtools/ describes how to install the toolset and has links to example scripts to build and package R.

This toolset was modified (by Jeroen Ooms) from the ’MSYS2’ project, which uses compilers (GCC version 8.3.0) and runtime libraries from the ‘Mingw-w64’ project ¹⁹ and provides a package manager (‘pacman’) for components and additional libraries.

The toolsets used for 64-bit Windows from 2008–20 were based on Mingw-w64. The assistance of Yu Gong at a crucial step in porting R to Mingw-w64 is gratefully acknowledged, as well as help from Kai Tietz, the lead developer of the Mingw-w64 project.

3.1.2 LaTeX

Both building R and checking packages need a distribution of LaTeX installed, with the directory containing pdflatex on the path.

The ‘MiKTeX’ (https://miktex.org/) distribution of LaTeX is that used on CRAN. This can be set up to install extra packages ‘on the fly’ (without asking), which is the simplest way to use it. The ‘basic’ version of ‘MiKTeX’ will need to add some packages.²⁰ In any case ensure that the inconsolata package is installed—you can check with the ‘MiKTeX’ Package Manager.

It is also possible to use the TeX Live distribution from https://www.tug.org/texlive/. (The CRAN package tinytex can install and manage a subset of TeX Live.)

3.2 Checking the build

If a test fails, there will almost always be a .Rout.fail file in the directory being checked (often tests/Examples or tests): examine the file to help pinpoint the problem.

Parallel checking of package sources (part of make check-devel and make check-recommended) is possible: see the environment variable TEST_MC_CORES to the maximum number of processes to be run in parallel.

3.3 Testing an Installation

The toolset is not needed to run these tests, but more comprehensive analysis of errors will be given if diff is in the path (and errorsAreFatal = FALSE is then not needed below).

runs the basic tests and then all the tests on the standard and recommended packages. These tests can be run from anywhere: they write some of their results in the tests folder of the R home directory (as given by R.home()), and hence may need to be run under the account used to install R.

The results of example(md5sums) when testing tools will differ from the reference output as some files are installed with Windows’ CRLF line endings.

4 Installing R under macOS

[The rest of this paragraph is only relevant after release.] The front page of a CRAN site has a link ‘Download R for (Mac) OS X’. Click on that, then download the file R-4.1.0.pkg and install it. This runs on macOS 10.13 and later on ‘x86_64’ (Intel) CPUs²¹ (High Sierra, Mojave, Catalina, Big Sur, …).

Installers for R-patched and R-devel are usually available from https://mac.R-project.org. (Some of these packages are unsigned/not notarized: to install those Control/right/two-finger click, select ‘Open With’ and ‘Installer’.)

For some older versions of the OS you can in principle (it is little tested) install R from the sources (see macOS).

It is important that if you use a binary installer package that your OS is fully updated: look at ‘Updates’ from the ‘App Store’ (‘Software Update’ in ’System Preferences’ on Catalina) to be sure. (If using XQuartz, check that is current.)

To install, just double-click on the icon of the file you downloaded. At the ‘Installation Type’ stage, note the option to ‘Customize’. This currently shows four components: everyone will need the ‘R Framework’ component: the remaining components are optional. (The ‘Tcl/Tk’ component is needed to use package tcltk. The ‘Texinfo’ component is only needed by those installing source packages or R from its sources.)

This is an Apple Installer package. If you encounter any problem during the installation, please check the Installer log by clicking on the “Window” menu and item “Installer Log”. The full output (select “Show All Log”) is useful for tracking down problems. Note the the installer is clever enough to try to upgrade the last-installed version of the application where you installed it (which may not be where you want this time …).

Various parts of the build require XQuartz to be installed: see https://xquartz.macosforge.org/. These include the tcltk package and the X11 device: attempting to use these without XQuartz will remind you.²² Also for the cairographics-based devices (which are not often used on macOS) such as png(type = "cairo").

If you update your macOS version, you should re-install R (and perhaps XQuartz): the installer may tailor the installation to the current version of the OS.

4.1 Running R under macOS

There is a GUI console normally installed with the R icon in /Applications which you can run by double-clicking (e.g. from Launchpad or Finder). (If you cannot find it there it was possibly installed elsewhere so try searching for it in Spotlight.) This is usually referred to as R.APP to distinguish it from command-line R: its user manual is currently part of the macOS FAQ at https://cran.r-project.org/bin/macosx/RMacOSX-FAQ.html and can be viewed from R.APP’s ‘Help’ menu.

You can run command-line R and Rscript from a Terminal²³ so these can be typed as commands like any other Unix-alike: see the next chapter of this manual. There are some small differences which may surprise users of R on other platforms, notably the default location of the personal library directory (under ~/Library/R, e.g. ~/Library/R/4.0/library), and that warnings, messages and other output to stderr are highlighted in bold.

Those using the zsh shell (the default for new user accounts as from Catalina) might find the command R being masked by the zsh builtin r (which recalls commands). One can use a full path to R in an alias, or add disable r to ~/.zshrc.

It has been reported that running R.APP may fail if no preferences are stored, so if it fails when launched for the very first time, try it again (the first attempt will store some preferences).

Using the X11 device or the X11-based versions of View() and edit() for data frames and matrices (the latter are the default for command-line R but not R.APP) requires an X sub-system to be installed: see macOS. So do the tcltk package and some third-party packages.

4.2 Uninstalling under macOS

R for macOS consists of two parts: the GUI (R.APP) and the R framework. Un-installation is as simple as removing those folders (e.g. by dragging them onto the Trash aka Bin). The typical installation will install the GUI into the /Applications/R.app folder and the R framework into the /Library/Frameworks/R.framework folder. The links to R and Rscript in /usr/local/bin should also be removed.

The installation consists of up to four Apple packages:²⁴ org.r-project.R.fw.pkg, org.r-project.R.GUI.pkg, org.r-project.x86_64.tcltk.x11 and org.r-project.x86_64.texinfo. You can use pkgutil --forget if you want the Apple Installer to forget about the package without deleting its files (useful for the R framework when installing multiple R versions in parallel), or after you have deleted the files.

Uninstalling the Tcl/Tk or Texinfo components (which are installed under /usr/local) is not as simple. You can list the files they installed in a Terminal by

4.3 Multiple versions

The installer will remove any previous version²⁵ of the R framework which it finds installed. This can be avoided by using pkgutil --forget (see the previous section). However, note that different versions are installed under /Library/Frameworks/R.framework/Versions as 4.1, 4.0 and so on, so it is not possible to have different ‘4.x.y’ versions installed for the same ‘x’.

However, R.APP will always run the ‘current’ version, that is the last installed version. A small utility, Rswitch.app (available at https://mac.R-project.org/#other: it is 32-bit so not usable on Catalina), can be used to change the ‘current’ version. This is of limited use as R.APP is compiled against a particular version of R and will likely crash if switched to an earlier version. This may allow you to install a development version of R (de-selecting R.APP) and then switch back to the release version.

5 Running R

How to start R and what command-line options are available is discussed in Invoking R in An Introduction to R.

You should ensure that the shell has set adequate resource limits: R expects a stack size of at least 8MB and to be able to open at least 256 file descriptors. (Any modern OS should have default limits at least as large as these, but apparently NetBSD may not. Use the shell command ulimit (sh/bash) or limit (csh/tcsh) to check.) For some compilers²⁶ and packages a larger stack size has been needed: 20-25MB has sufficed to date.

R makes use of a number of environment variables, the default values of many of which are set in file R_HOME/etc/Renviron (there are none set by default on Windows and hence no such file). These are set at configure time, and you would not normally want to change them – a possible exception is R_PAPERSIZE (see Setting paper size). The paper size will be deduced from the ‘LC_PAPER’ locale category if it exists and R_PAPERSIZE is unset, and this will normally produce the right choice from ‘a4’ and ‘letter’ on modern Unix-alikes (but can always be overridden by setting R_PAPERSIZE).

Various environment variables can be set to determine where R creates its per-session temporary directory. The environment variables TMPDIR, TMP and TEMP are searched in turn and the first one which is set and points to a writable area is used. If none do, the final default is /tmp on Unix-alikes and the value of R_USER on Windows. The path should be an absolute path not containing spaces (and it is best to avoid non-alphanumeric characters such as +).

Some Unix-alike systems are set up to remove files and directories periodically from /tmp, for example by a cron job running tmpwatch. Set TMPDIR to another directory before starting long-running jobs on such a system.

Note that TMPDIR will be used to execute configure scripts when installing packages, so if /tmp has been mounted as ‘noexec’, TMPDIR needs to be set to a directory from which execution is allowed.

6 Add-on packages

It is helpful to use the correct terminology. A package is loaded from a library by the function library(). Thus a library is a directory containing installed packages; the main library is R_HOME/library, but others can be used, for example by setting the environment variable R_LIBS or using the R function .libPaths(). To avoid any confusion you will often see a library directory referred to as a ‘library tree’.

6.1 Default packages

(plus, of course, base) and this can be changed by setting the option in startup code (e.g. in ~/.Rprofile). It is initially set to the value of the environment variable R_DEFAULT_PACKAGES if set (as a comma-separated list). Setting R_DEFAULT_PACKAGES=NULL ensures that only package base is loaded.

Changing the set of default packages is normally used to reduce the set for speed when scripting: in particular not using methods will reduce the start-up time by a factor of up to two. But it can also be used to customize R, e.g. for class use. Rscript also checks the environment variable R_SCRIPT_DEFAULT_PACKAGES; if set, this takes precedence over R_DEFAULT_PACKAGES.

6.2 Managing libraries

R packages are installed into libraries, which are directories in the file system containing a subdirectory for each package installed there.

R comes with a single library, R_HOME/library which is the value of the R object ‘.Library’ containing the standard and recommended²⁷ packages. Both sites and users can create others and make use of them (or not) in an R session. At the lowest level ‘.libPaths()’ can be used to add paths to the collection of libraries or to report the current collection.

R will automatically make use of a site-specific library R_HOME/site-library if this exists (it does not in a vanilla R installation). This location can be overridden by setting²⁸ ‘.Library.site’ in R_HOME/etc/Rprofile.site, or (not recommended) by setting the environment variable R_LIBS_SITE. Like ‘.Library’, the site libraries are always included by ‘.libPaths()’.

Users can have one or more libraries, normally specified by the environment variable R_LIBS_USER. This has a default value (to see it, use ‘Sys.getenv("R_LIBS_USER")’ within an R session), but that is only used if the corresponding directory actually exists (which by default it will not).

Both R_LIBS_USER and R_LIBS_SITE can specify multiple library paths, separated by colons (semicolons on Windows).

6.3 Installing packages

Packages may be distributed in source form or compiled binary form. Installing source packages which contain C/C++/Fortran code requires that compilers and related tools be installed. Binary packages are platform-specific and generally need no special tools to install, but see the documentation for your platform for details.

Note that you may need to specify implicitly or explicitly the library to which the package is to be installed. This is only an issue if you have more than one library, of course.

Ensure that the environment variable TMPDIR is either unset (and /tmp exists and can be written in and executed from) or is the absolute path to a valid temporary directory, not containing spaces.

For most users it suffices to call ‘install.packages(pkgname)’ or its GUI equivalent if the intention is to install a CRAN package and internet access is available.²⁹ On most systems ‘install.packages()’ will allow packages to be selected from a list box (typically with thousands of items).

The part ‘-l /path/to/library’ can be omitted, in which case the first library of a normal R session is used (that shown by .libPaths()[1]).

There are a number of options available: use R CMD INSTALL --help to see the current list.

Alternatively, packages can be downloaded and installed from within R. First choose your nearest CRAN mirror using chooseCRANmirror(). Then download and install packages pkg1 and pkg2 by

The essential dependencies of the specified packages will also be fetched. Unless the library is specified (argument lib) the first library in the library search path is used: if this is not writable, R will ask the user (in an interactive session) if the default personal library should be created, and if allowed to will install the packages there.

If you want to fetch a package and all those it depends on (in any way) that are not already installed, use e.g.

install.packages can install a source package from a local .tar.gz file (or a URL to such a file) by setting argument repos to NULL: this will be selected automatically if the name given is a single .tar.gz file.

install.packages can look in several repositories, specified as a character vector by the argument repos: these can include a CRAN mirror, Bioconductor, R-forge, rforge.net, local archives, local files, …). Function setRepositories() can select amongst those repositories that the R installation is aware of.

Something which sometimes puzzles users is that install.packages() may report that a package which they believe should be available is not found. Some possible reasons:

Naive users sometimes forget that as well as installing a package, they have to use library to make its functionality available.

6.3.1 Windows

What install.packages does by default is different on Unix-alikes (except macOS) and Windows. On Unix-alikes it consults the list of available source packages on CRAN (or other repository/ies), downloads the latest version of the package sources, and installs them (via R CMD INSTALL). On Windows it looks (by default) first at the list of binary versions of packages available for your version of R and downloads the latest versions (if any). If no binary version is available or the source version is newer, it will install the source versions of packages without compiled C/C++/Fortran code, and offer to do so for those with, if make is available (and this can be tuned by option "install.packages.compile.from.source").

On Windows install.packages can also install a binary package from a local zip file (or the URL of such a file) by setting argument repos to NULL. Rgui.exe has a menu Packages with a GUI interface to install.packages, update.packages and library.

Windows binary packages for R are distributed as a single binary containing either or both architectures (32- and 64-bit).

R CMD INSTALL works in Windows to install source packages. No additional tools are needed if the package does not contain compiled code, and install.packages(type="source") will work for such packages. Those with compiled code need the tools (see The Windows toolset) to be on the path: this can be set from within an R session by something like

Occasional permission problems after unpacking source packages have been seen on some systems: these have been circumvented by setting the environment variable R_INSTALL_TAR to ‘tar.exe’.

If you have only a source package that is known to work with current R and just want a binary Windows build of it, you could make use of the building service offered at https://win-builder.r-project.org/.

For almost all packages R CMD INSTALL will attempt to install both 32- and 64-bit builds of a package if run from a 32/64-bit install of R. It will report success if the installation of the architecture of the running R succeeded, whether or not the other architecture was successfully installed. The exceptions are packages with a non-empty configure.win script or which make use of src/Makefile.win. If configure.win does something appropriate to both architectures use³⁰ option --force-biarch: otherwise

R CMD INSTALL
--merge-multiarch

can be applied to a source tarball to merge separate 32- and 64-bit installs. (This can only be applied to a tarball, and will only succeed if both installs succeed.)

If you have a package without compiled code and no Windows-specific help, you can zip up an installation on another OS and install from that zip file on Windows. However, such a package can be installed from the sources on Windows without any additional tools.

6.3.2 macOS

On macOS install.packages works as it does on other Unix-alike systems, but there is an additional type mac.binary (available for the CRAN distribution but not when compiling R from source) which can be passed to install.packages in order to download and install binary packages from a suitable repository. These binary package files for macOS have the extension ‘.tgz’. The R.APP GUI provides menus for installation of either binary or source packages, from CRAN or local files.

On R builds using binary packages, the default is type both: this looks first at the list of binary packages available for your version of R and installs the latest versions (if any). If no binary version is available or the source version is newer, it will install the source versions of packages without compiled C/C++/Fortran code and offer to do so for those with, if make is available.

Note that most binary packages which include compiled code are tied to a particular series (e.g. R 4.0.x or 4.1.x) of R.

Installing source packages which do not contain compiled code should work with no additional tools. For others you will need the ‘Command Line Tools’ for Xcode and compilers which match those used to build R: see macOS.

Package rJava and those which depend on it need a Java runtime installed and several packages need X11 installed, including those using Tk. See macOS and Java (macOS).

Tcl/Tk extension BWidget used to be distributed with R but no longer is; Tktable has been distributed with most versions of R (but not 4.0.0).

The default compilers specified are shown in file /Library/Frameworks/R.framework/Resources/etc/Makeconf. At the time of writing those settings assumed that the C, Fortran and C++ compilers were on the path, using gfortran 8.2.0 (see macOS). The settings can be changed, either by editing that file or in a file such as ~/.R/Makevars (see the next section). Entries which may need to be changed include ‘CC’, ‘CXX’, ‘FC’, ‘FLIBS’ and the corresponding flags, and perhaps ‘CXXCPP’, ‘DYLIB_LD’, ‘MAIN_LD’, ‘SHLIB_CXXLD’ and ‘SHLIB_LD’, as well as the ‘CXX11’, ‘CXX14’, ‘CXX17’ and ‘CXX20’ variants

So for example on Mojave or later you could select a specific non-Apple build of clang for both C and C++ with extensive checking by having in ~/.R/Makevars

(with lines broken here for legibility). If that clang build supports OpenMP, you can add

to compile OpenMP-using packages. It will also be necessary to arrange for the libomp.dylib library to be found at both install time and run time, for example by copying/linking it somewhere that is searched such as /usr/local/lib.

Apple includes many Open Source libraries in macOS but increasingly without the corresponding headers (not even in Xcode nor the Command Line Tools): they are often rather old versions. If installing packages from source using them it is usually easiest to install a statically-linked up-to-date copy of the Open Source package from its sources or from https://mac.R-project.org/libs-4/. But sometimes it is desirable/necessary to use Apple’s dynamically linked library, in which case appropriate headers could be extracted from the sources³¹ available via https://opensource.apple.com – this has been used for OpenSSL and iodbc.

Those using Xcode/Command Line Tools 12 or later (as released for macOS 11 ‘Big Sur’) probably want to arrange that the flag

is part of CFLAGS. Apple has changed the default to make implicit declarations a compilation error (something seen in no other compiler) and undesrstandably package authors have been unaware that this might be done — issues are most often seen in configure scripts.

6.3.3 Customizing package compilation

The R system and package-specific compilation flags can be overridden or added to by setting the appropriate Make variables in the personal file HOME/.R/Makevars-R_PLATFORM (but HOME/.R/Makevars.win or HOME/.R/Makevars.win64 on Windows), or if that does not exist, HOME/.R/Makevars, where ‘R_PLATFORM’ is the platform for which R was built, as available in the platform component of the R variable R.version. The full path to an alternative personal file³² can be specified via the environment variable R_MAKEVARS_USER.

Package developers are encouraged to use this mechanism to enable a reasonable amount of diagnostic messaging (“warnings”) when compiling, such as e.g. -Wall -pedantic for tools from GCC, the GNU Compiler Collection, or for clang.

Note that this mechanism can also be used when it necessary to change the optimization level whilst installing a particular package. For example

Another use is to override the settings in a binary installation of R. For example, to use a different Fortran compiler on macOS

There is also provision for a site-wide Makevars.site file under R_HOME/etc (in a sub-architecture-specific directory if appropriate). This is read immediately after Makeconf, and the path to an alternative file can be specified by environment variable R_MAKEVARS_SITE.

Note that these mechanisms do not work with packages which fail to pass settings down to sub-makes, perhaps reading etc/Makeconf in makefiles in subdirectories. Fortunately such packages are unusual.

6.3.4 Multiple sub-architectures

When installing packages from their sources, there are some extra considerations on installations which use sub-architectures. These are commonly used on Windows but can in principle be used on other platforms.

When a source package is installed by a build of R which supports multiple sub-architectures, the normal installation process installs the packages for all sub-architectures. The exceptions are

In those cases only the current architecture is installed. Further sub-architectures can be installed by

using the path to R or R --arch to select the additional sub-architecture. There is also

R CMD INSTALL
--merge-multiarch

to build and merge the two architectures, starting with a source tarball.

6.3.5 Byte-compilation

Byte-compilation can be controlled on a per-package basis by the ‘ByteCompile’ field in the DESCRIPTION file.

6.3.6 External software

Some R packages contain compiled code which links to external software libraries. Unless the external library is statically linked (which is done as much as possible for binary packages on Windows and macOS), the libraries have to be found when the package is loaded and not just when it is installed. How this should be done depends on the OS (and in some cases the version).

For Unix-alikes except macOS the primary mechanism is the ld.so cache controlled by ldconfig: external dynamic libraries recorded in that cache will be found. Standard library locations will be covered by the cache, and well-designed software will add its locations (as for example openmpi does on Fedora). The secondary mechanism is to consult the environment variable LD_LIBRARY_PATH. The R script controls that variable, and sets it to the concatenation of R_LD_LIBRARY_PATH, R_JAVA_LD_LIBRARY_PATH and the environment value of LD_LIBRARY_PATH. The first two have defaults which are normally set when R is installed (but can be overridden in the environment) so LD_LIBRARY_PATH is the best choice for a user to set.

On macOS the primary mechanism is to embed the absolute path to dependent dynamic libraries into an object when it is compiled. Few R packages arrange to do so, but it can be edited³³ via install_name_tool — that only deals with direct dependencies and those would also need to be compiled to include the absolute paths of their dependencies. If the choice of absolute path is to be deferred to load time, how they are resolved is described in man dyld: the role of LD_LIBRARY_PATH is replaced on macOS by DYLD_LIBRARY_PATH and DYLD_FALLBACK_LIBRARY_PATH. Running R CMD otool -L on the package shared object will show where (if anywhere) its dependencies are resolved. DYLD_FALLBACK_LIBRARY_PATH is preferred (and it is that which is manipulated by the R script), but as from 10.11 (‘El Capitan’) the default behaviour had been changed for security reasons to discard these environment variables when invoking a shell script (and R is a shell script). That makes the only portable option to set R_LD_LIBRARY_PATH in the environment, something like

The precise rules for where Windows looks for DLLs are complex and depend on the version of Windows. But for present purposes the main solution is to put the directories containing the DLLs the package links to (and any those DLLs link to) on the PATH. 64-bit versions of Windows will ignore 32-bit DLLs from 64-bit R and vice versa.

The danger with any of the methods which involve setting environment variables is of inadvertently masking a system library. This is less for DYLD_FALLBACK_LIBRARY_PATH and for appending to PATH on Windows (as it should already contain the system library paths).

6.4 Updating packages

The command update.packages() is the simplest way to ensure that all the packages on your system are up to date. It downloads the list of available packages and their current versions, compares it with those installed and offers to fetch and install any that have later versions on the repositories.

An alternative interface to keeping packages up-to-date is provided by the command packageStatus(), which returns an object with information on all installed packages and packages available at multiple repositories. The print and summary methods give an overview of installed and available packages, the upgrade method offers to fetch and install the latest versions of outdated packages.

One sometimes-useful additional piece of information that packageStatus() returns is the status of a package, as "ok", "upgrade" or "unavailable" (in the currently selected repositories). For example

6.5 Removing packages

Packages can be removed in a number of ways. From a command prompt they can be removed by

6.6 Setting up a package repository

Utilities such as install.packages can be pointed at any CRAN-style repository, and R users may want to set up their own. The ‘base’ of a repository is a URL such as https://www.stats.ox.ac.uk/pub/RWin/: this must be an URL scheme that download.packages supports (which also includes ‘https://’, ‘ftp://’ and ‘file://’). Under that base URL there should be directory trees for one or more of the following types of package distributions:

Each terminal directory must also contain a PACKAGES file. This can be a concatenation of the DESCRIPTION files of the packages separated by blank lines, but only a few of the fields are needed. The simplest way to set up such a file is to use function write_PACKAGES in the tools package, and its help explains which fields are needed. Optionally there can also be PACKAGES.rds and PACKAGES.gz files, downloaded in preference to PACKAGES. (If you have a mis-configured server that does not report correctly non-existent files you may need these files.)

To add your repository to the list offered by setRepositories(), see the help file for that function.

Incomplete repositories are better specified via a contriburl argument than via being set as a repository.

A repository can contain subdirectories, when the descriptions in the PACKAGES file of packages in subdirectories must include a line of the form

6.7 Checking installed source packages

It can be convenient to run R CMD check on an installed package, particularly on a platform which uses sub-architectures. The outline of how to do this is, with the source package in directory pkg (or a tarball filename):

Where sub-architectures are in use the R CMD check line can be repeated with additional architectures by

where --extra-arch selects only those checks which depend on the installed code and not those which analyse the sources. (If multiple sub-architectures fail only because they need different settings, e.g. environment variables, --no-multiarch may need to be added to the INSTALL lines.) On Unix-alikes the architecture to run is selected by --arch: this can also be used on Windows with R_HOME/bin/R.exe, but it is more usual to select the path to the Rcmd.exe of the desired architecture.

So on Windows to install, check and package for distribution a source package from a tarball which has been tested on another platform one might use

where one might want to run the second and third lines in a different shell with different settings for environment variables and the path (to find external software, notably for Gtk+).

R CMD INSTALL can do a i386 install and then add the x64 DLL from a single command by

7 Internationalization and Localization

Internationalization refers to the process of enabling support for many human languages, and localization to adapting to a specific country and language.

Current builds of R support all the character sets that the underlying OS can handle. These are interpreted according to the current locale, a sufficiently complicated topic to merit a separate section. Note though that R has no built-in support for right-to-left languages and bidirectional output, relying on the OS services. For example, how character vectors in UTF-8 containing both English digits and Hebrew characters are printed is OS-dependent (and perhaps locale-dependent).

The other aspect of the internationalization is support for the translation of messages. This is enabled in almost all builds of R.

7.1 Locales

A locale is a description of the local environment of the user, including the preferred language, the encoding of characters, the currency used and its conventions, and so on. Aspects of the locale are accessed by the R functions Sys.getlocale and Sys.localeconv.

The system of naming locales is OS-specific. There is quite wide agreement on schemes, but not on the details of their implementation. A locale needs to specify

R is principally concerned with the first (for translations) and third. Note that the charset may be deducible from the language, as some OSes offer only one charset per language.

7.1.1 Locales under Unix-alikes

Modern Linux uses the XPG³⁴ locale specifications which have the form ‘en_GB’, ‘en_GB.UTF-8’, ‘aa_ER.UTF-8@saaho’, ‘de_AT.iso885915@euro’, the components being in the order listed above. (See man locale and locale -a for more details.) Similar schemes are used by most Unix-alikes: some (including some distributions of Linux) use ‘.utf8’ rather than ‘.UTF-8’.

Note that whereas UTF-8 locales are nowadays almost universally used, locales such as ‘en_GB’ use 8-bit encodings for backwards compatibility.

7.1.2 Locales under Windows

Windows also uses locales, but specified in a rather less concise way. Most users will encounter locales only via drop-down menus, but more information and lists can be found by searching for ‘Windows language country strings’).

Some care is needed with Windows’ locale names. For example, chinese is Traditional Chinese and not Simplified Chinese as used in most of the Chinese-speaking world.

7.1.3 Locales under macOS

macOS supports locales in its own particular way, but the R GUI tries to make this easier for users. See https://developer.apple.com/library/archive/documentation/MacOSX/Conceptual/BPInternational/ for how users can set their locales. As with Windows, end users will generally only see lists of languages/territories. Users of R in a terminal may need to set the locale to something like ‘en_GB.UTF-8’ if it defaults to ‘C’ (as it sometimes does when logging in remotely and for batch jobs: note whether Terminal sets the LANG environment variable is an (advanced) preference, but does so by default).

Internally macOS uses a form similar to Linux: the main difference from other Unix-alikes is that where a character set is not specified it is assumed to be UTF-8.

7.2 Localization of messages

The preferred language for messages is by default taken from the locale. This can be overridden first by the setting of the environment variable LANGUAGE and then³⁵ by the environment variables LC_ALL, LC_MESSAGES and LANG. (The last three are normally used to set the locale and so should not be needed, but the first is only used to select the language for messages.) The code tries hard to map locales to languages, but on some systems (notably Windows) the locale names needed for the environment variable LC_ALL do not all correspond to XPG language names and so LANGUAGE may need to be set. (One example is ‘LC_ALL=es’ on Windows which sets the locale to Estonian and the language to Spanish.)

It is usually possible to change the language once R is running via (not Windows)

Sys.setlocale("LC_MESSAGES",
"new_locale")

, or by setting an environment variable such as LANGUAGE, provided³⁶ the language you are changing to can be output in the current character set. But this is OS-specific, and has been known to stop working on an OS upgrade.

Messages are divided into domains, and translations may be available for some or all messages in a domain. R makes use of the following domains.

Dividing up the messages in this way allows R to be extensible: as packages are loaded, their message translation catalogues can be loaded too.

R-level and C-level domains are subtly different, for example in the way strings are canonicalized before being passed for translation.

Translations are looked for by domain according to the currently specified language, as specifically as possible, so for example an Austrian (‘de_AT’) translation catalogue will be used in preference to a generic German one (‘de’) for an Austrian user. However, if a specific translation catalogue exists but does not contain a translation, the less specific catalogues are consulted. For example, R has catalogues for ‘en_GB’ that translate the Americanisms (e.g., ‘gray’) in the standard messages into English.³⁷ Two other examples: there are catalogues for ‘es’, which is Spanish as written in Spain and these will by default also be used in Spanish-speaking Latin American countries, and also for ‘pt_BR’, which are used for Brazilian locales but not for locales specifying Portugal.

Translations in the right language but the wrong charset are made use of by on-the-fly re-encoding. The LANGUAGE variable (only) can be a colon-separated list, for example ‘se:de’, giving a set of languages in decreasing order of preference. One special value is ‘en@quot’, which can be used in a UTF-8 locale to have American error messages with pairs of single quotes translated to Unicode directional quotes.

If no suitable translation catalogue is found or a particular message is not translated in any suitable catalogue, ‘English’³⁸ is used.

8 Choosing between 32- and 64-bit builds

Almost all current CPUs have both 32- and 64-bit sets of instructions. Most OSes running on such CPUs offer the choice of building a 32-bit or a 64-bit version of R (and details are given below under specific OSes). For most a 32-bit version is the default, but for some (e.g., ‘x86_64’ Linux and macOS ≥ 10.6) 64-bit is.

All current versions of R use 32-bit integers (this is enforced in the build) and ISO/IEC 60559³⁹ double-precision reals, and so compute to the same precision⁴⁰ and with the same limits on the sizes of numerical quantities. The principal difference is in the size of the pointers.

So, for speed you may want to use a 32-bit build (especially on a laptop), but to handle large datasets (and perhaps large files) a 64-bit build. You can often build both and install them in the same place: See Sub-architectures. (This is done for the Windows binary distributions.)

Even on 64-bit builds of R there are limits on the size of R objects (see help("Memory-limits")), some of which stem from the use of 32-bit integers (especially in Fortran code). For example, each dimension of an array is limited to 2^{31} - 1.

9 The standalone Rmath library

The routines supporting the distribution and special⁴¹ functions in R and a few others are declared in C header file Rmath.h. These can be compiled into a standalone library for linking to other applications. (Note that they are not a separate library when R is built, and the standalone version differs in several ways.)

The makefiles and other sources needed are in directory src/nmath/standalone, so the following instructions assume that is the current working directory (in the build directory tree on a Unix-alike if that is separate from the sources).

Rmath.h contains ‘R_VERSION_STRING’, which is a character string containing the current R version, for example "4.0.0".

There is full access to R’s handling of NaN, Inf and -Inf via special versions of the macros and functions

There is no support for R’s notion of missing values, in particular not for NA_INTEGER nor the distinction between NA and NaN for doubles.

A little care is needed to use the random-number routines. You will need to supply the uniform random number generator

or use the one supplied (and with a shared library or DLL you may have to use the one supplied, which is the Marsaglia-multicarry with an entry point

The facilities to change the normal random number generator are available through the constant N01_kind. This takes values from the enumeration type

9.1 Unix-alikes

If R has not already been made in the directory tree, configure must be run as described in the main build instructions.

will make standalone libraries libRmath.a and libRmath.so (libRmath.dylib on macOS): ‘make static’ and ‘make shared’ will create just one of them.

and link against ‘-lRmath’ (and ‘-lm’ if needed on your OS). The example file test.c does nothing useful, but is provided to test the process (via make test). Note that you will probably not be able to run it unless you add the directory containing libRmath.so to the LD_LIBRARY_PATH environment variable (libRmath.dylib, DYLD_FALLBACK_LIBRARY_PATH on macOS).

will (un)install the header Rmath.h and shared and static libraries (if built). Both prefix= and DESTDIR are supported, together with more precise control as described for the main build.

On some systems ‘make install-strip’ will install a stripped shared library.

9.2 Windows

You need to set up⁴² almost all the tools to make R and then run (in a Unix-like shell)

This creates a static library libRmath.a and a DLL Rmath.dll. If you want an import library libRmath.dll.a (you don’t need one), use

and link against ‘-lRmath’. This will use the first found of libRmath.dll.a, libRmath.a and Rmath.dll in that order, so the result depends on which files are present. You should be able to force static or dynamic linking via

or by linking to explicit files (as in the ‘test’ target in Makefile.win: this makes two executables, test.exe which is dynamically linked, and test-static.exe, which is statically linked).

It is possible to link to Rmath.dll using other compilers, either directly or via an import library: if you make a MinGW-w64 import library as above, you will create a file Rmath.def which can be used (possibly after editing) to create an import library for other systems such as Visual C++.

to ensure that the constants like NA_REAL are linked correctly. (Auto-import will probably work with MinGW-w64, but it is better to be sure. This is likely to also work with VC++, Borland and similar compilers.)

Appendix A Essential and useful other programs under a Unix-alike

This appendix gives details of programs you will need to build R on Unix-like platforms, or which will be used by R if found by configure.

Remember that some package management systems (such as RPM and Debian/Ubuntu’s) make a distinction between the user version of a package and the development version. The latter usually has the same name but with the extension ‘-devel’ or ‘-dev’: you need both versions installed.

A.1 Essential programs and libraries

You need a means of compiling C and Fortran 90 (see Using Fortran). Your C compiler should be ISO/IEC 60059⁴³, POSIX 1003.1 and C99-compliant.⁴⁴ R tries to choose suitable flags⁴⁵ for the C compilers it knows about, but you may have to set CC or CFLAGS suitably. For versions of gcc prior to 5.1 with glibc-based Linux this means including -std=gnu99⁴⁶. (Note that options essential to run the compiler even for linking, such as those to set the architecture, should be specified as part of CC rather than in CFLAGS.)

Unless you do not want to view graphs on-screen (or use macOS) you need ‘X11’ installed, including its headers and client libraries. For recent Fedora/RedHat distributions it means (at least) RPMs ‘libX11’, ‘libX11-devel’, ‘libXt’ and ‘libXt-devel’. On Debian/Ubuntu we recommend the meta-package ‘xorg-dev’. If you really do not want these you will need to explicitly configure R without X11, using --with-x=no.

The command-line editing (and command completion) depends on the GNU readline library (including its headers): version 6.0 or later is needed for all the features to be enabled. Otherwise you will need to configure with --with-readline=no (or equivalent).

A suitably comprehensive iconv function is essential. The R usage requires iconv to be able to translate between "latin1" and "UTF-8", to recognize "" (as the current encoding) and "ASCII", and to translate to and from the Unicode wide-character formats "UCS-[24][BL]E" — this is true by default for glibc⁴⁷ but not of most commercial Unixes. However, you can make use of GNU libiconv (as used on macOS: see https://www.gnu.org/software/libiconv/).

The OS needs to have enough support⁴⁸ for wide-character types: this is checked at configuration. Some C99 functions⁴⁹ are required and checked for at configuration. A small number of POSIX functions⁵⁰ are essential, and others⁵¹ will be used if available.

Installations of zlib (version 1.2.5 or later), libbz2 (version 1.0.6 or later: called bzip2-libs/bzip2-devel or libbz2-1.0/libbz2-dev by some Linux distributions) and liblzma⁵² version 5.0.3 or later are required.

Either PCRE1 (version 8.32 or later, formerly known as just PCRE) or PCRE2 is required: PCRE2 is preferred and using PCRE1 requires configure option --with-pcre1. Only the 8-bit library and headers are needed if these are packaged separately. JIT support (optional) is desirable for the best performance. For PCRE2 >= 10.30 (which is desirable as matching has been re-written not to use recursion and the Unicode tables were updated to version 10)

suffices. If building PCRE1 for use with R a suitable configure command might be

The --enable-jit flag is supported for most common CPUs. (See also the comments for Solaris.)

Some packages require the ‘Unicode properties’ which are optional for PCRE1: support for this and JIT can be checked at run-time by calling pcre_config().

Library libcurl (version 7.28.0 or later⁵³) is required. Information on libcurl is found from the curl-config script: if that is missing or needs to be overridden⁵⁴ there are macros to do so described in file config.site.

A tar program is needed to unpack the sources and packages (including the recommended packages). A version⁵⁵ that can automagically detect compressed archives is preferred for use with untar(): the configure script looks for gtar and gnutar before tar – use environment variable TAR to override this. (On NetBSD/OpenBSD systems set this to bsdtar if that is installed.)

There need to be suitable versions of the tools grep and sed: the problems are usually with old AT&T and BSD variants. configure will try to find suitable versions (including looking in /usr/xpg4/bin which is used on some commercial Unixes).

You will not be able to build most of the manuals unless you have texi2any version 5.1 or later installed (which requires perl), and if not most of the HTML manuals will be linked to a version on CRAN. To make PDF versions of the manuals you will also need file texinfo.tex installed (which is part of the GNU texinfo distribution but is often made part of the TeX package in re-distributions) as well as texi2dvi.⁵⁶ Further, the versions of texi2dvi and texinfo.tex need to be compatible: we have seen problems with older TeX distributions.

If you want to build from the R Subversion repository then texi2any is highly recommended as it is used to create files which are in the tarball but not stored in the Subversion repository.

The PDF documentation (including doc/NEWS.pdf) and building vignettes needs pdftex and pdflatex. We require LaTeX version 2005/12/01 or later (for UTF-8 support). Building PDF package manuals (including the R reference manual) and vignettes is sensitive to the version of the LaTeX package hyperref and we recommend that the TeX distribution used is kept up-to-date. A number of standard LaTeX packages are required (including url and some of the font packages such as times, helvetic, ec and cm-super) and others such as hyperref and inconsolata are desirable (and without them you may need to change R’s defaults: see Making the manuals). Note that package hyperref (currently) requires packages kvoptions, ltxcmds and refcount. For distributions based on TeX Live the simplest approach may be to install collections collection-latex, collection-fontsrecommended, collection-latexrecommended, collection-fontsextra and collection-latexextra (assuming they are not installed by default): Fedora uses names like texlive-collection-fontsextra and Debian/Ubuntu like texlive-fonts-extra.

The essential programs should be in your PATH at the time configure is run: this will capture the full paths.

Those distributing binary versions of R may need to be aware of the licences of the external libraries it is linked to (including ‘useful’ libraries from the next section). The liblzma library is in the public domain and X11, libbzip2, libcurl and zlib have MIT-style licences. PCRE and PCRE2 have a BSD-style licence which requires distribution of the licence (included in R’s COPYRIGHTS file) in binary distributions. GNU readline is licensed under GPL (which version(s) of GPL depends on the readline version).

A.2 Useful libraries and programs

The ability to use translated messages makes use of gettext and most likely needs GNU gettext: you do need this to work with new translations, but otherwise the version contained in the R sources will be used if no suitable external gettext is found.

The ‘modern’ version of the X11(), jpeg(), png() and tiff() graphics devices uses the cairo and (optionally) Pango libraries. Cairo version 1.2.0 or later is required. Pango needs to be at least version 1.10, and 1.12 is the earliest version we have tested. (For Fedora users we believe the pango-devel RPM and its dependencies suffice.) R checks for pkg-config, and uses that to check first that the ‘pangocairo’ package is installed (and if not, ‘cairo’) and if additional flags are needed for the ‘cairo-xlib’ package, then if suitable code can be compiled. These tests will fail if pkg-config is not installed⁵⁷, and are likely to fail if cairo was built statically (unusual). Most systems with Gtk+ 2.8 or later installed will have suitable libraries.

For the best font experience with these devices you need suitable fonts installed: Linux users will want the urw-fonts package. On platforms which have it available, the msttcorefonts package⁵⁸ provides TrueType versions of Monotype fonts such as Arial and Times New Roman. Another useful set of fonts is the ‘liberation’ TrueType fonts available at https://fedorahosted.org/liberation-fonts/,⁵⁹ which cover the Latin, Greek and Cyrillic alphabets plus a fair range of signs. These share metrics with Arial, Times New Roman and Courier New, and contain fonts rather similar to the first two (https://en.wikipedia.org/wiki/Liberation_fonts). Then there is the ‘Free UCS Outline Fonts’ project (https://www.gnu.org/software/freefont/) which are OpenType/TrueType fonts based on the URW fonts but with extended Unicode coverage. See the R help on X11 on selecting such fonts.

The bitmapped graphics devices jpeg(), png() and tiff() need the appropriate headers and libraries installed: jpeg (version 6b or later, or libjpeg-turbo) or libpng (version 1.2.7 or later) and zlib or libtiff (versions 4.0.[5-10] and 4.1.0 have been tested) respectively. pkg-config is used if available and so needs the appropriate .pc file (which requires libtiff version 4.x and is not available on all platforms for jpeg before version 9c). They also need support for either X11 or cairo (see above). Should support for these devices not be required or broken system libraries need to be avoided there are configure options --without-libpng, --without-jpeglib and --without-libtiff. The TIFF library has many optional features such as jpeg, libz, lzma, jbig and jpeg12, none of which is required for the tiff() devices but may need to be present to link the library (usually only an issue for static linking).

Option --with-system-tre is also available: it needs a recent version of TRE. (The latest sources are in the git repository at https://github.com/laurikari/tre/, but at the time of writing (2016) the resulting build did not pass its checks.).

An implementation of XDR is required, and the R sources contain one which is likely to suffice (although a system version may have higher performance). XDR is part of RPC and historically has been part of libc on a Unix-alike. (In principle man xdr_string should tell you which library is needed, but it often does not: on Solaris and others it is provided by libnsl.) However some builds⁶⁰ of glibc omit or hide it with the intention that the TI-RPC library be used, in which case libtirpc (and its development version) should be installed, and its headers⁶¹ need to be on the C include path or under /usr/include/tirpc.

Use of the X11 clipboard selection requires the Xmu headers and libraries. These are normally part of an X11 installation (e.g. the Debian meta-package ‘xorg-dev’), but some distributions have split this into smaller parts, so for example recent versions of Fedora require the ‘libXmu’ and ‘libXmu-devel’ RPMs.

Some systems (notably macOS and at least some FreeBSD systems) have inadequate support for collation in multibyte locales. It is possible to replace the OS’s collation support by that from ICU (International Components for Unicode, http://site.icu-project.org/), and this provides much more precise control over collation on all systems. ICU is available as sources and as binary distributions for (at least) most Linux distributions, Solaris, FreeBSD, macOS and AIX, usually as libicu or icu4c. It will be used by default where available: should a very old or broken version of ICU be found this can be suppressed by --without-ICU.

The bitmap and dev2bitmap devices and function embedFonts() use ghostscript (https://www.ghostscript.com/). This should either be in your path when the command is run, or its full path specified by the environment variable R_GSCMD at that time.

At the time of writing a full installation on Fedora Linux used the following packages and their development versions, and this may provide a useful checklist for other systems:

A.2.1 Tcl/Tk

The tcltk package needs Tcl/Tk ≥ 8.4 installed: the sources are available at https://www.tcl.tk/. To specify the locations of the Tcl/Tk files you may need the configuration options

or use the configure variables TCLTK_LIBS and TCLTK_CPPFLAGS to specify the flags needed for linking against the Tcl and Tk libraries and for finding the tcl.h and tk.h headers, respectively. If you have both 32- and 64-bit versions of Tcl/Tk installed, specifying the paths to the correct config files may be necessary to avoid confusion between them.

Versions of Tcl/Tk up to 8.5.19 and 8.6.9 have been tested (including most versions of 8.4.x, but not recently).

Note that the tk.h header includes⁶² X11 headers, so you will need X11 and its development files installed.

A.2.2 Java support

The build process looks for Java support on the host system, and if it finds it sets some settings which are useful for Java-using packages (such as rJava and JavaGD: these require a full JDK). This check can be suppressed by configure option --disable-java. Configure variable JAVA_HOME can be set to point to a specific JRE/JDK, on the configure command line or in the environment.

Principal amongst these settings are some paths to the Java libraries and JVM, which are stored in environment variable R_JAVA_LD_LIBRARY_PATH in file R_HOME/etc/ldpaths (or a sub-architecture-specific version). A typical setting for ‘x86_64’ Linux is

Unfortunately this depends on the exact version of the JRE/JDK installed, and so may need updating if the Java installation is updated. This can be done by running R CMD javareconf which updates settings in both R_HOME/etc/Makeconf and R_HOME/etc/ldpaths. See R CMD javareconf --help for details: note that this needs to be done by the account owning the R installation.

Another way of overriding those settings is to set the environment variable R_JAVA_LD_LIBRARY_PATH (before R is started, hence not in ~/.Renviron), which suffices to run already-installed Java-using packages. For example

It may be possible to avoid this by specifying an invariant link as the path when configuring. For example, on that system any of

worked (since the ‘auto’ setting of /etc/alternatives chose Java 8 aka 1.8.0).

‘Non-server’ Oracle distributions of Java as from version 11 are of a full JDK. However, Linux distributions can be confusing: for example Fedora 32 had

where the -devel RPMs are needed to complete the JDK. (At the time of writing java-openjdk was Java 14.) Debian/Ubuntu use ‘-jre’ and ‘-jdk’, e.g.

A.2.3 Other compiled languages

Some add-on packages need a C++ compiler. This is specified by the configure variables CXX, CXXFLAGS and similar. configure will normally find a suitable compiler. It is possible to specify an alternative C++14 compiler by the configure variables CXX14, CXX14STD, CXX14FLAGS and similar (see C++ Support). Again, configure will normally find a suitable value for CXX14STD if the compiler given by CXX is capable of compiling C++14 code, but it is possible that a completely different compiler will be needed. (Similar macros are provided for C++17 and C++20.)

For source files with extension .f90 or .f95 containing free-form Fortran, the compiler defined by the macro FC is used by R CMD INSTALL. Note that it is detected by the name of the command without a test that it can actually compile Fortran 90 code. Set the configure variable FC to override this if necessary: variables FCFLAGS and FCLIBS_XTRA might also need to be set.

A.3 Linear algebra

A.3.1 BLAS

The linear algebra routines in R can make use of enhanced BLAS (Basic Linear Algebra Subprograms, https://www.netlib.org/blas/faq.html) routines. However, these have to be explicitly requested at configure time: R provides an internal BLAS which is well-tested and will be adequate for most uses of R.

You can specify a particular BLAS library via a value for the configuration option --with-blas and not to use an external BLAS library by --without-blas (the default). If --with-blas is given with no =, its value is taken from the environment variable BLAS_LIBS, set for example in config.site. If neither the option nor the environment variable supply a value, a search is made for a suitable⁶³ BLAS. If the value is not obviously a linker command (starting with a dash or giving the path to a library), it is prefixed by ‘-l’, so

is an instruction to link against ‘-lfoo’ to find an external BLAS (which needs to be found both at link time and run time).

The configure code checks that the external BLAS is complete (it must include all double precision and double complex routines, as well as LSAME), and appears to be usable. However, an external BLAS has to be usable from a shared object (so must contain position-independent code), and that is not checked.

Some enhanced BLASes are compiler-system-specific (sunperf on Solaris⁶⁴, libessl on IBM, Accelerate on macOS). The correct incantation for these is often found via --with-blas with no value on the appropriate platforms.

Some of the external BLASes are multi-threaded. One issue is that R profiling (which uses the SIGPROF signal) may cause problems, and you may want to disable profiling if you use a multi-threaded BLAS. Note that using a multi-threaded BLAS can result in taking more CPU time and even more elapsed time (occasionally dramatically so) than using a similar single-threaded BLAS. On a machine running other tasks, there can be contention for CPU caches that reduces the effectiveness of the optimization of cache use by a BLAS implementation: some people warn that this is especially problematic for hyperthreaded CPUs.

Note that under Unix (but not under Windows) if R is compiled against a non-default BLAS and --enable-BLAS-shlib is not used (it is the default on all platforms except AIX), then all BLAS-using packages must also be. So if R is re-built to use an enhanced BLAS then packages such as quantreg will need to be re-installed; they may be under other circumstances.

R relies on ISO/IEC 60559 compliance of an external BLAS. This can be broken if for example the code assumes that terms with a zero factor are always zero and do not need to be computed—whereas x*0 can be NaN. This is checked in the test suite.

External BLAS implementations often make less use of extended-precision floating-point registers (where available) and will almost certainly re-order computations. This can result in less accuracy than using a reference BLAS, and may result in different solutions, e.g. different signs in SVD and eigendecompositions.

Debian/Ubuntu systems provide a system-specific way to switch the BLAS in use. Build R with -with-blas to select the OS version of the reference BLAS, and then use update-alternatives to switch between the available BLAS libraries. See https://wiki.debian.org/DebianScience/LinearAlgebraLibraries.

The URIs for several of these BLAS have been subject to frequent gratuitous changes, so you will need to search for their current locations.

BLAS (and LAPACK) routines may be used inside threaded code, for example in OpenMP sections in packages such as mgcv. The reference implementations are thread-safe but external ones may not be (even single-threaded ones): this can lead to hard-to-track-down incorrect results or segfaults.

NOTE: BLAS libraries built with gfortran 9 (and later and versions 8.4, 7.5 and later in those series) require calls from C/C++ to handle ‘hidden’ character lengths — R itself does so but many packages do not and some segfault. (This applies also to external LAPACK libraries.)

A.3.1.1 ATLAS

ATLAS (http://math-atlas.sourceforge.net/) is a “tuned” BLAS that runs on a wide range of Unix-alike platforms. Unfortunately it is built by default as a static library that on some platforms may not be able to be used with shared objects such as are used in R packages. Be careful when using pre-built versions of ATLAS static libraries (they seem to work on ‘ix86’ platforms, but not always on ‘x86_64’ ones).

ATLAS contains replacements for a small number of LAPACK routines, but can be built to merge these with LAPACK sources to include a full LAPACK library.

Recent versions of ATLAS can be built as a single shared library, either libsatlas or libtatlas (serial or threaded respectively): these may even contain a full LAPACK. Such builds can be used by one of

Distributed ATLAS libraries cannot be tuned to your machine and so are a compromise: for example Fedora tunes ‘x86_64’ for CPUs with SSE3 extensions, and separate ‘atlas-sse2’ and ‘atlas-sse3’ ‘i686’ RPMs are available.⁶⁵

Note that building R on Linux against distributed shared libraries may need ‘-devel’ or ‘-dev’ packages installed.

Consult its installation guide⁶⁶ for how to build ATLAS as a shared library or as a static library with position-independent code (on platforms where that matters).

According to the ATLAS FAQ⁶⁷ the maximum number of threads used by multi-threaded ATLAS is set at compile time. Also, the author advises against using multi-threaded ATLAS on hyperthreaded CPUs without restricting affinities at compile-time to one virtual core per physical CPU. (For the Fedora libraries the compile-time flag specifies 4 threads.)

A.3.1.2 OpenBLAS

Dr Kazushige Goto wrote a tuned BLAS for several processors and OSes, which was frozen in mid-2010. OpenBLAS (https://www.openblas.net/) is a descendant project with support for some later CPUs.

See see Shared BLAS for an alternative (and in many ways preferable) way to use them.

Some platforms provide multiple builds of OpenBLAS: for example Fedora 32 has RPMs⁶⁸

respectively, each of which can be used as a shared BLAS. For the second and third the number of threads is controlled by OPENBLAS_NUM_THREADS and OMP_NUM_THREADS (as usual for OpenMP) respectively. There is also a Fedora RPM ‘openblas-Rblas’ to replace libRblas.so in their distribution of R.

Note that building R on Linux against distributed libraries may need ‘-devel’ or ‘-dev’ packages installed.

For ‘ix86’ and ‘x86_64’ most distributed libraries contain several alternatives for different CPU microarchitectures with the choice being made at run time.

A.3.1.3 Intel MKL

For Intel processors (and perhaps others) and some distributions of Linux, there is Intel’s Math Kernel Library. You are strongly encouraged to read the MKL User’s Guide, which is installed with the library, before attempting to link to MKL. This includes a ‘link line advisor’ which will suggest appropriate incantations: its use is recommended. Or see https://software.intel.com/content/www/us/en/develop/articles/intel-mkl-link-line-advisor.html.

There are also versions of MKL for macOS⁶⁹ and Windows, but when these have been tried they did not work with the default compilers used for R on those platforms.

The MKL interface has changed several times but has been stable in recent years: the following examples have been used with versions 10.3 to 2020.2, for GCC compilers on ‘x86_64’.

The option --with-lapack is used since MKL contains a tuned copy of LAPACK (often older than the current version) as well as BLAS (see LAPACK), although this can be omitted.

R can also be linked against a single shared library, libmkl_rt.so, for both BLAS and LAPACK, but the correct OpenMP and MKL interface layer then has to be selected via environment variables. With 64-bit builds and the GCC compilers, we used

On Debian/Ubuntu, MKL is provided by package intel-mkl-full and one can set libmkl_rt.so as the system-wide implementation of both BLAS and LAPACK during installation of the package, so that also R installed from Debian/Ubuntu package r-base would use it. It is, however, still essential to set MKL_INTERFACE_LAYER and MKL_THREADING_LAYER before running R, otherwise MKL computations will produce incorrect results. R does not have to be rebuilt to use MKL, but configure include tests which may discover some errors such as a failure to set the correct OpenMP and MKL interface layer.

The default number of threads will be chosen by the OpenMP software, but can be controlled by setting OMP_NUM_THREADS or MKL_NUM_THREADS, and in recent versions seems to default to a sensible value for sole use of the machine. (Parallel MKL has not always passed make check-all, but did with MKL 2019.4 and later.)

MKL includes a partial implementation of FFTW3, which causes trouble for applications that require some of the FFTW3 functionality unsupported in MKL. Please see the MKL manuals for description of these limitations and for instructions on how to create a custom version of MKL which excludes the FFTW3 wrappers.

A.3.1.4 Shared BLAS

The BLAS library will be used for many of the add-on packages as well as for R itself. This means that it is better to use a shared/dynamic BLAS library, as most of a static library will be compiled into the R executable and each BLAS-using package.

R offers the option of compiling the BLAS into a dynamic library libRblas stored in R_HOME/lib and linking both R itself and all the add-on packages against that library.

This is the default on all platforms except AIX unless an external BLAS is specified and found: for the latter it can be used by specifying the option --enable-BLAS-shlib, and it can always be disabled via --disable-BLAS-shlib.

Another option to change the BLAS in use is to symlink a single dynamic BLAS library to R_HOME/lib/libRblas.so. For example, just

on ‘x86_64’ Fedora will change the BLAS used to multithreaded OpenBLAS. A similar link works for most versions of the OpenBLAS (provided the appropriate lib directory is in the run-time library path or ld.so cache). It can also be used for a single-library ATLAS, so on ‘x86_64’ Fedora

can be used with its distributed ATLAS libraries. (If you have the ‘-devel’ RPMS installed you can omit the .0/.3.)

Note that rebuilding or symlinking libRblas.so may not suffice if the intention is to use a modified LAPACK contained in an external BLAS: the latter could even cause conflicts. However, on Fedora where the OpenBLAS distribution contains a copy of LAPACK, it is the latter which is used.

A.3.2 LAPACK

Provision is made for using an external LAPACK library, principally to cope with BLAS libraries which contain a copy of LAPACK (such as sunperf on Solaris, Accelerate on macOS and ATLAS and MKL on ‘ix86’/‘x86_64’ Linux). At least LAPACK version 3.2 is required. This can only be done if --with-blas has been used.

However, the likely performance gains are thought to be small (and may be negative). The default is not to search for a suitable LAPACK library, and this is definitely not recommended. You can specify a specific LAPACK library or a search for a generic library by the configuration option --with-lapack. The default for --with-lapack is to check the BLAS library and then look for an external library ‘-llapack’. Sites searching for the fastest possible linear algebra may want to build a LAPACK library using the ATLAS-optimized subset of LAPACK: this is simplest with a dynamic ATLAS library which contains a full LAPACK, when --with-lapack suffices.

A value for --with-lapack can be set via the environment variable LAPACK_LIBS, but this will only be used if --with-lapack is specified (as the default value is no) and the BLAS library does not contain LAPACK.

If you do use --with-lapack, be aware of potential problems with bugs in the LAPACK sources (or in the posted corrections to those sources). In particular, bugs in DGEEV and DGESDD have resulted in error messages such as

. Other potential problems are incomplete versions of the libraries, seen several times in Linux distributions over the years.

Please do bear in mind that using --with-lapack is ‘definitely not recommended’: it is provided only because it is necessary on some platforms and because some users want to experiment with claimed performance improvements. Reporting problems where it is used unnecessarily will simply irritate the R helpers.

Note too the comments about ISO/IEC 60559 compliance in the section of external BLAS: these apply equally to an external LAPACK, and for example the Intel MKL documentation has said

We rely on limited support in LAPACK for matrices with 2^{31} or more elements: it is possible that an external LAPACK will not have that support.

A.3.3 Caveats

As with all libraries, you need to ensure that they and R were compiled with compatible compilers and flags. For example, this has meant that on Sun Sparc using the Oracle compilers the flag -dalign is needed if sunperf is to be used.

On some systems it has been necessary that an external BLAS/LAPACK was built with the same Fortran compiler used to build R.

LAPACK 3.9.0 has a bug in which the DCOMBSSQ subroutine has a bug that may cause NA to be interpreted as zero. This is fixed in the R 3.6.3 and later sources, but if you use an external LAPACK, you may need to fix it there.

Appendix B Configuration on a Unix-alike

B.1 Configuration options

will give a list. Probably the most important ones not covered elsewhere are (defaults in brackets)

You will want to use --disable-R-profiling if you are building a profiled executable of R (e.g. with ‘-pg)’. Support for R profiling requires OS support for POSIX threads (aka ‘pthreads’), which are available on all mainstream Unix-alike platforms.

Flag --enable-R-shlib causes the make process to build R as a dynamic (shared) library, typically called libR.so, and link the main R executable R.bin against that library. This can only be done if all the code (including system libraries) can be compiled into a dynamic library, and there may be a performance⁷⁰ penalty. So you probably only want this if you will be using an application which embeds R. Note that C code in packages installed on an R system linked with --enable-R-shlib is linked against the dynamic library and so such packages cannot be used from an R system built in the default way. Also, because packages are linked against R they are on some OSes also linked against the dynamic libraries R itself is linked against, and this can lead to symbol conflicts.

For maximally effective use of valgrind, R should be compiled with valgrind instrumentation. The configure option is --with-valgrind-instrumentation=level, where level is 0, 1 or 2. (Level 0 is the default and does not add anything.) The system headers for valgrind can be requested by option --with-system-valgrind-headers: they will be used if present (on Linux they may be in a separate package such as valgrind-devel). Note though that there is no guarantee that the code in R will be compatible with very old⁷¹ or future valgrind headers.

If you need to re-configure R with different options you may need to run make clean or even make distclean before doing so.

The configure script has other generic options added by autoconf and which are not supported for R: in particular building for one architecture on a different host is not possible.

B.2 Internationalization support

Translation of messages is supported via GNU gettext unless disabled by the configure option --disable-nls. The configure report will show NLS as one of the ‘Additional capabilities’ if support has been compiled in, and running in an English locale (but not the C locale) will include

B.3 Configuration variables

If you need or want to set certain configure variables to something other than their default, you can do that by either editing the file config.site (which documents many of the variables you might want to set: others can be seen in file etc/Renviron.in) or on the command line as

If you are building in a directory different from the sources, there can be copies of config.site in the source and the build directories, and both will be read (in that order). In addition, if there is a file ~/.R/config, it is read between the config.site files in the source and the build directories.

There is also a general autoconf mechanism for config.site files, which are read before any of those mentioned in the previous paragraph. This looks first at a file specified by the environment variable CONFIG_SITE, and if not is set at files such as /usr/local/share/config.site and /usr/local/etc/config.site in the area (exemplified by /usr/local) where R would be installed.

These variables are precious, implying that they do not have to be exported to the environment, are kept in the cache even if not specified on the command line, checked for consistency between two configure runs (provided that caching is used), and are kept during automatic reconfiguration as if having been passed as command line arguments, even if no cache is used.

See the variable output section of configure --help for a list of all these variables.

If you find you need to alter configure variables, it is worth noting that some settings may be cached in the file config.cache, and it is a good idea to remove that file (if it exists) before re-configuring. Note that caching is turned off by default: use the command line option --config-cache (or -C) to enable caching.

B.3.1 Setting paper size

One common variable to change is R_PAPERSIZE, which defaults to ‘a4’, not ‘letter’. (Valid values are ‘a4’, ‘letter’, ‘legal’ and ‘executive’.)

This is used both when configuring R to set the default, and when running R to override the default. It is also used to set the paper size when making PDF manuals.

The configure default will most often be ‘a4’ if R_PAPERSIZE is unset. (If the (Debian Linux) program paperconf is found or the environment variable PAPERSIZE is set, these are used to produce the default.)

B.3.2 Setting the browsers

Another precious variable is R_BROWSER, the default HTML browser, which should take a value of an executable in the user’s path or specify a full path.

B.3.3 Compilation flags

If you have libraries and header files, e.g., for GNU readline, in non-system directories, use the variables LDFLAGS (for libraries, using ‘-L’ flags to be passed to the linker) and CPPFLAGS (for header files, using ‘-I’ flags to be passed to the C/C++ preprocessors), respectively, to specify these locations. These default to ‘-L/usr/local/lib’ (LDFLAGS, ‘-L/usr/local/lib64’ on most 64-bit Linux OSes) and ‘-I/usr/local/include’ (CPPFLAGS, but note that on most systems /usr/local/include is regarded as a system include directory and so instances in that macro will be skipped) to catch the most common cases. If libraries are still not found, then maybe your compiler/linker does not support re-ordering of -L and -l flags. In this case, use a different compiler (or a front-end shell script which does the re-ordering).

These flags can also be used to build a faster-running version of R. On most platforms using gcc, having ‘-O3’ in CFLAGS and FFLAGS produces worthwhile performance gains with gcc and gfortran, but may result in a less reliable build (both segfaults and incorrect numeric computations have been seen). On systems using the GNU linker (especially those using R as a shared library), it is likely that including ‘-Wl,-O1’ in LDFLAGS is worthwhile, and ‘'-Bdirect,--hash-style=both,-Wl,-O1'’ is recommended at https://lwn.net/Articles/192624/. Tuning compilation to a specific CPU family (e.g. ‘-mtune=native’ for gcc) can give worthwhile performance gains, especially on older architectures such as ‘ix86’.

B.3.4 Making manuals

The default settings for making the manuals are controlled by R_RD4PDF and R_PAPERSIZE.

B.4 Setting the shell

By default the shell scripts such as R will be ‘#!/bin/sh’ scripts (or using the SHELL chosen by configure). This is almost always satisfactory, but on a few systems /bin/sh is not a Bourne shell or clone, and the shell to be used can be changed by setting the configure variable R_SHELL to a suitable value (a full path to a shell, e.g. /usr/local/bin/bash).

B.5 Using make

To compile R, you will most likely find it easiest to use GNU make, although the Sun make works on Solaris.

To build in a separate directory you need a make that supports the VPATH variable, for example GNU make and Sun make.

If you want to use a make by another name, for example if your GNU make is called ‘gmake’, you need to set the variable MAKE at configure time, for example

B.6 Using Fortran

To compile R, you need a Fortran 90 compiler. The current default is to search for gfortran, g95, xlf95 f95, fort, ifort, ifc, efc, pgfortran, pgf95 lf95, ftn, nagfor, xlf90, f90, pgf90, pghpf, epcf90. (Note that these are searched for by name, without checking the standard of Fortran they support.) The command and flags used should support fixed-form Fortran with extension .f: in the unusual case that a specific flag is needed for free-form Fortran with extension .f90 or .f95, this can be specified as part of FCFLAGS.

The search mechanism can be changed using the configure variable FC which specifies the command that runs the Fortran compiler. If your Fortran compiler is in a non-standard location, you should set the environment variable PATH accordingly before running configure, or use the configure variable FC to specify its full path.

If your Fortran libraries are in slightly peculiar places, you should also look at LD_LIBRARY_PATH (or your system’s equivalent) to make sure that all libraries are on this path.

Note that only Fortran compilers which convert identifiers to lower case are supported.

You must set whatever compilation flags (if any) are needed to ensure that Fortran integer is equivalent to a C int pointer and Fortran double precision is equivalent to a C double pointer. This is checked during the configuration process.

Some of the Fortran code makes use of DOUBLE COMPLEX and COMPLEX*16 variables. This is checked for at configure time, as well as its equivalence to the Rcomplex C structure defined in R_ext/Complex.h.

gfortran 10 by default gives a compilation error for the previously widespread practice of passing a Fortran array element where an array is expected, or a scalar instead of a length-one array. See https://gcc.gnu.org/gcc-10/porting_to.html.

B.7 Compile and load flags

A wide range of flags can be set in the file config.site or as configure variables on the command line. We have already mentioned

Library paths specified as -L/lib/path in LDFLAGS are collected together and prepended to LD_LIBRARY_PATH (or your system’s equivalent), so there should be no need for -R or -rpath flags.

Variables such as CPICFLAGS are determined where possible by configure. Some systems allows two types of PIC flags, for example ‘-fpic’ and ‘-fPIC’, and if they differ the first allows only a limited number of symbols in a shared object. Since R as a shared library has about 6200 symbols, if in doubt use the larger version.

Other variables often set by configure include ‘MAIN_LDFLAGS’, ‘SAFE_FFLAGS’, ‘SHLIB_LDFLAGS’ and ‘SHLIB_CXXLDFLAGS’: see file config.site in the sources for more documentation on these and others.

To compile a profiling version of R, one might for example want to use ‘MAIN_CFLAGS=-pg’, ‘MAIN_FFLAGS=-pg’, ‘MAIN_LDFLAGS=-pg’ on platforms where ‘-pg’ cannot be used with position-independent code.

Beware: it may be necessary to set CFLAGS and FFLAGS in ways compatible with the libraries to be used: one possible issue is the alignment of doubles, another is the way structures are passed.

On some platforms configure will select additional flags for CFLAGS, CPPFLAGS and LIBS in R_XTRA_CFLAGS (and so on). These are for options which are always required, for example to force IEC 60559 compliance.

B.8 Maintainer mode

There are several files that are part of the R sources but can be re-generated from their own sources by configuring with option --enable-maintainer-mode and then running make in the build directory. This requires other tools to be installed, discussed in the rest of this section.

File configure is created from configure.ac and the files under m4 by autoconf and aclocal (part of the automake package). There is a formal version requirement on autoconf of 2.69 or later, but it is unlikely that anything other than the most recent versions⁷² have been thoroughly tested.

Grammar files *.y are converted to C sources by an implementation of yacc, usually bison -y: these are found in src/main and src/library/tools/src. It is known that earlier versions of bison generate code which reads (and in some cases writes) outside array bounds: bison 2.6.1 was found to be satisfactory.

The ultimate sources for package compiler are in its noweb directory. To re-create the sources from src/library/compiler/noweb/compiler.nw, the command notangle is required. Some Linux distributions include this command in package noweb. It can also be installed from the sources at https://www.cs.tufts.edu/~nr/noweb/⁷³. The package sources are only re-created even in maintainer mode if src/library/compiler/noweb/compiler.nw has been updated.

Appendix C Platform notes

This section provides some notes on building R on different Unix-alike platforms. These notes are based on tests run on one or two systems in each case with particular sets of compilers and support libraries. Success in building R depends on the proper installation and functioning of support software; your results may differ if you have other versions of compilers and support libraries.

Older versions of this manual contain notes on platforms such as HP-UX, IRIX, Alpha/OSF1 (for R < 2.10.0, and support has since been removed for all of these) and AIX (for R < = 3.5.x) for which we have had no recent reports.

C macros to select particular platforms can be tricky to track down (there is a fair amount of misinformation on the Web). The Wiki (currently) at https://sourceforge.net/p/predef/wiki/Home/ can be helpful. The R sources have used (often in included software under src/extra)

C.1 X11 issues

The ‘X11()’ graphics device is the one started automatically on Unix-alikes (except most macOS builds) when plotting. As its name implies, it displays on a (local or remote) X server, and relies on the services provided by the X server.

The ‘modern’ version of the ‘X11()’ device is based on ‘cairo’ graphics and (in most implementations) uses ‘fontconfig’ to pick and render fonts. This is done on the server, and although there can be selection issues, they are more amenable than the issues with ‘X11()’ discussed in the rest of this section.

When X11 was designed, most displays were around 75dpi, whereas today they are of the order of 100dpi or more. If you find that X11() is reporting⁷⁴ missing font sizes, especially larger ones, it is likely that you are not using scalable fonts and have not installed the 100dpi versions of the X11 fonts. The names and details differ by system, but will likely have something like Fedora’s

and you need to ensure that the ‘-100dpi’ versions are installed and on the X11 font path (check via xset -q). The ‘X11()’ device does try to set a pointsize and not a pixel size: laptop users may find the default setting of 12 too large (although very frequently laptop screens are set to a fictitious dpi to appear like a scaled-down desktop screen).

More complicated problems can occur in non-Western-European locales, so if you are using one, the first thing to check is that things work in the C locale. The likely issues are a failure to find any fonts or glyphs being rendered incorrectly (often as a pair of ASCII characters). X11 works by being asked for a font specification and coming up with its idea of a close match. For text (as distinct from the symbols used by plotmath), the specification is the first element of the option "X11fonts" which defaults to

If you are using a single-byte encoding, for example ISO 8859-2 in Eastern Europe or KOI8-R in Russian, use xlsfonts to find an appropriate family of fonts in your encoding (the last field in the listing). If you find none, it is likely that you need to install further font packages, such as ‘xorg-x11-fonts-ISO8859-2-75dpi’ and ‘xorg-x11-fonts-cyrillic’ shown in the listing above.

Multi-byte encodings (most commonly UTF-8) are even more complicated. There are few fonts in ‘iso10646-1’, the Unicode encoding, and they only contain a subset of the available glyphs (and are often fixed-width designed for use in terminals). In such locales fontsets are used, made up of fonts encoded in other encodings. If the locale you are using has an entry in the ‘XLC_LOCALE’ directory (typically /usr/share/X11/locale), it is likely that all you need to do is to pick a suitable font specification that has fonts in the encodings specified there. If not, you may have to get hold of a suitable locale entry for X11. This may mean that, for example, Japanese text can be displayed when running in ‘ja_JP.UTF-8’ but not when running in ‘en_GB.UTF-8’ on the same machine (although on some systems many UTF-8 X11 locales are aliased to ‘en_US.UTF-8’ which covers several character sets, e.g. ISO 8859-1 (Western European), JISX0208 (Kanji), KSC5601 (Korean), GB2312 (Chinese Han) and JISX0201 (Kana)).

On some systems scalable fonts are available covering a wide range of glyphs. One source is TrueType/OpenType fonts, and these can provide high coverage. Another is Type 1 fonts: the URW set of Type 1 fonts provides standard typefaces such as Helvetica with a larger coverage of Unicode glyphs than the standard X11 bitmaps, including Cyrillic. These are generally not part of the default install, and the X server may need to be configured to use them. They might be under the X11 fonts directory or elsewhere, for example,

C.2 Linux

Linux is the main development platform for R, so compilation from the sources is normally straightforward with the most common compilers and libraries.⁷⁵

Recall that some package management systems (such as RPM and deb) make a distinction between the user version of a package and the developer version. The latter usually has the same name but with the extension ‘-devel’ or ‘-dev’: you need both versions installed. So please check the configure output to see if the expected features are detected: if for example ‘readline’ is missing add the developer package. (On most systems you will also need ‘ncurses’ and its developer package, although these should be dependencies of the ‘readline’ package(s).) You should expect to see in the configure summary

When R has been installed from a binary distribution there are sometimes problems with missing components such as the Fortran compiler. Searching the ‘R-help’ archives will normally reveal what is needed.

It seems that ‘ix86’ Linux accepts non-PIC code in shared libraries, but this is not necessarily so on other platforms, in particular on 64-bit CPUs such as ‘x86_64’. So care can be needed with BLAS libraries and when building R as a shared library to ensure that position-independent code is used in any static libraries (such as the Tcl/Tk libraries, libpng, libjpeg and zlib) which might be linked against. Fortunately these are normally built as shared libraries with the exception of the ATLAS BLAS libraries.

The default optimization settings chosen for CFLAGS etc are conservative. It is likely that using -mtune will result in significant performance improvements on recent CPUs: one possibility is to add -mtune=native for the best possible performance on the machine on which R is being installed. It is also possible to increase the optimization levels to -O3: however for many versions of the compilers this has caused problems in at least one CRAN package.

is appropriate since most (but not all) software installs its 64-bit libraries in /usr/local/lib64. To build a 32-bit version of R on ‘x86_64’ with Fedora 32 we used

Note the use of ‘LIBnn’: ‘x86_64’ Fedora installs its 64-bit software in /usr/lib64 and 32-bit software in /usr/lib. Linking will skip over inappropriate binaries, but for example the 32-bit Tcl/Tk configure scripts are in /usr/lib. It may also be necessary to set the pkg-config path, e.g. by

The 32-bit system libcurl did not work with the system CA certificates: this is worked around in R’s test suite.

64-bit versions on Linux are built with support for files > 2Gb, and 32-bit versions will be if possible unless --disable-largefile is specified.

Note that 32-bit glibc before version 2.29 uses a 32-bit time_t type, so to pass all the date-time checks needs R built with flag --with-internal-tzcode.

Users of ‘ix86’ CPUs with SSE2 support⁷⁶ may prefer to use the C/C++/Fortran flags

to force floating-point to use the same instructions as ‘x86_64’ builds and hence not make use of 80-bit ‘extended precision’ intermediate results. (NB: this affects more than floating-point operations. For some OSes and versions of gcc it might be necessary to add -mstackrealign.)

To build a 64-bit version of R on ‘ppc64’ (also known as ‘powerpc64’) with gcc 4.1.1, Ei-ji Nakama used

the additional flags being needed to resolve problems linking against libnmath.a and when linking R as a shared library.

The setting of the macro ‘SAFE_FFLAGS’ may need some help. It should not need additional flags on platforms other than ‘68000’ (not likely to be encountered) and ‘ix86’. For the latter, if the Fortran compiler is GNU (gfortran or possibly g77) the flags

are added: earlier versions of R added -ffloat-store and this might still be needed if a ‘ix86’ CPU is encountered without SSE2 support. Note that it is a replacement for ‘FFLAGS’, so should include all the flags in that macro (except perhaps the optimization level).

Additional compilation flags can be specified for added safety/security checks. For example Fedora adds

to all the C, C++ and Fortran compiler flags (even though _GLIBCXX_ASSERTIONS is only for C++ in current GCC and glibc and none of these are documented for gfortran). Use of _GLIBCXX_ASSERTIONS will link abort and printf into almost all C++ code, and

R CMD check
--as-cran

will warn.

C.2.1 Clang

R has been built with Linux ‘ix86’ and ‘x86_64’ C and C++ compilers (https://clang.llvm.org) based on the Clang front-ends, invoked by CC=clang CXX=clang++, together with gfortran. These take very similar options to the corresponding GCC compilers.

This has to be used in conjunction with a Fortran compiler: the configure code will remove -lgcc from FLIBS, which is needed for some versions of gfortran.

The current out-of-the-box default for clang++ is to use the C++ runtime from the installed g++. Using the runtime from the libc++ project (Fedora RPM libcxx-devel) via -stdlib=libc++ has also been tested.

There is a project called flang (https://github.com/flang-compiler/flang)) to develop a Fortran compiler similar to clang but based on the Portland Group’s front end. This needs something like

Note that flang accepts all the flags which clang does (the driver is a modified version of clang, and flang is a symbolic link to clang), but does not implement all of them for Fortran compilation: it also accepts most PGI-style flags such as -mp for OpenMP. It currently produces few diagnostics even with -Wall -pedantic.

flang’s Fortran runtime is compiled against OpenMP and it seems this conflicts with using OpenMP in R. So it may be necessary to disable the latter by configuring R using --disable-openmp.

It is not clear what architectures flang intends to support: our experiments were done on ‘x86_64’. At the time of writing binary ‘releases’ were available for that platform (called by them ‘x86’) and ‘ppc64le’.

C.2.2 Intel compilers

Intel compilers have been used under ‘ix86’ and ‘x86_64’ Linux. Brian Ripley used version 9.0 of the compilers for ‘x86_64’ on Fedora Core 5 with

It may be necessary to use CC="icc -std=c99" or

CC="icc
-c99"

for C99-compliance. The flag -wd188 suppresses a large number of warnings about the enumeration type ‘Rboolean’. Because the Intel C compiler sets ‘__GNUC__’ without complete emulation of gcc, we suggest adding CPPFLAGS=-no-gcc.

To maintain correct IEC 60559 arithmetic you most likely need add flags to CFLAGS, FFLAGS and CXXFLAGS such as -mp (shown above) or -fp-model precise -fp-model source, depending on the compiler version.

Others have reported success with versions 10.x and 11.x. Bjørn-Helge Mevik reported success with version 2015.3 of the compilers, using (for a SandyBridge CPU on Centos 6.x)

It is possible that 32-builds need to force the use of SSE2 instructions in SAFE_FFLAGS, e.g. by

C.3 macOS

The instructions here are for Intel 64-bit (‘x86_64’) builds on 10.13–10.15 (High Sierra, Mojave and Catalina) and 11.0 (Big Sur). R can in principle be built for 10.6–10.12, although this is little tested and it may be necessary to install later versions of software such as libcurl and an older version⁷⁸ of gfortran.

C.3.1 Prerequisites

To build R itself from the sources with the compilers in the Command Line Tools (or Xcode) and gfortran from the installer mentioned above, use a file config.site containing

To install packages using compiled code one needs the Command Line Tools and appropriate compilers, e.g. Fortran and the C/C++ compilers from those tools. Some packages have further requirements such as pkg-config.

A subversion client, svn version 1.10, used to be part of macOS⁸². A more recent client can be obtained from https://mac.r-project.org/tools/, for example by

For some distributions of Command Line Tools version 12, -Wno-implicit-function-declaration needs to be included in CFLAGS to configure R 4.0.2 and earlier as well as some external software.

C.3.1.1 Note for Catalina and Big Sur users

The default security settings for Catalina and later can make it difficult to install Apple packages built after 2019-06-01 which have not been ‘notarized’⁸³ by Apple. And not just packages, as this has been seen for executables contained in tarballs/zipfiles (for example, for pandoc).

There are workarounds, including in some cases not installing the latest version. Usually one can use ‘Open With’ (Control/right/two-finger-click in Finder), then select ‘Installer’ and ‘Open’ if you get a further warning message.

If you run into problems with ‘quarantine’ for tarballs downloaded in a browser, consider using curl -OL to download (as illustrated above) or xattr -c to remove extended attributes.

C.3.2 Other C/C++ compilers

Other distributions of clang are available from https://releases.llvm.org/. In particular, these include support for OpenMP which Apple builds of clang do not, and may contain more bugfixes and new features (such as better diagnostics) than that in Command Line Tools/Xcode.

Suppose one of these distributions is installed under /usr/local/clang. Use a file config.site containing

The care to specify library paths is to ensure that the OpenMP runtime library, here /usr/local/clang/lib/libomp.dylib, is found when needed. If this works, you should see the line

in the configure output. Also, ‘R_LD_LIBRARY_PATH’ needs to be set to find the latest version of the C++ run-time libraries rather than the system ones.

(Long lines split for the manual, only.) Should the location of the SDK change (or Xcode provides the SDK rather than the Command Line Tools), it can be found by running xcrun -show-sdk-path.

The approach in the previous paragraph also works for Mojave, but for versions 10.x of the Command Line Tools an alternative is to install the headers to /usr/include: from a Terminal run

(This will need to be re-run if the OS decides to update the Command Line Tools.)

It is usually possible to build R with GCC (built from the sources, from a gfortran distribution, from Homebrew, …). When last tested⁸⁴ it was not possible to use gcc to build the quartz() device, so configure --without-aqua may be required.

C.3.3 Other libraries

to provide potentially higher-performance versions of the BLAS and LAPACK routines.⁸⁶ This also includes a full LAPACK which can be used via --with-lapack: however, the version of LAPACK it contains has often been seriously old (and is not used unless --with-lapack is specified). (For the Command Line Tools 12 and later, you need to ensure that -Wno-implicit-function-declaration is part of CFLAGS.)

Threading in Accelerate is controlled by ‘Grand Central Dispatch’ and is said not to need user control. Test nls.R in package stats usually fails with the Accelerate BLAS.

Looking at the top of /Library/Frameworks/R.framework/Resources/etc/Makeconf will show the compilers and configuration options used for the CRAN binary package for R: at the time of writing the non-default options

Configure option --with-internal-tzcode is the default on macOS, as the system implementation of time zones does not work correctly for times before 1902 or after 2037 (despite using a 64-bit time_t).

The TeX implementation used by the developers is MacTeX⁸⁷ (https://www.tug.org/mactex/): the full installation is about 7GB, but a smaller version (‘Basic TeX’) is available at https://www.tug.org/mactex/morepackages.html to which you will need to add some packages to build R, e.g. for the 2020 version we needed to add⁸⁸ cm-super, helvetic, inconsolata and texinfo which brought this to about 380MB. ‘TeX Live Utility’ (available via the MacTeX front page) provides a graphical means to manage TeX packages. This is documented to require High Sierra or later: for earlier versions see the instructions on the MacTeX front page.

Checking packages thoroughly requires ghostscript (part of the full MacTeX distribution or separately from https://www.tug.org/mactex/morepackages.html) and qpdf (from https://mac.R-project.org/libs-4/, a version of which is in the bin directory of a binary installation of R, usually /Library/Frameworks/R.framework/Resources/bin/qpdf).

One macOS quirk is that the default path has /usr/local/bin after /usr/bin, contrary to common practice on Unix-alikes. This means that if you install tools from the sources they will by default be installed under /usr/local and not supersede the system versions.

C.3.4 Tcl/Tk headers and libraries

If you plan to use the tcltk package for R, you need to install a distribution of Tcl/Tk. There are two alternatives. If you use R.APP you will want to use X11-based Tcl/Tk (as used on other Unix-alikes), which is installed under /usr/local/lib as part of the CRAN binary for R.⁸⁹ This may need configure options

There is also a native (‘Aqua’) version of Tcl/Tk which produces widgets in the native macOS style: this will not work with R.APP because of conflicts over the macOS menu, but for those only using command-line R this provides a much more intuitive interface to Tk for experienced Mac users. Earlier versions of macOS came with an Aqua Tcl/Tk distribution but these were often not at all recent versions of Tcl/Tk (8.5.9 in High Sierra, which is not even the latest patched version in that series). (Catalina included a partial distribution which lacks the files needed to compile against Tcl/Tk.) It is better to install Tcl/Tk 8.6.x from the sources or a binary distribution from https://www.activestate.com/activetcl/downloads. For the latter, configure R with

C.3.5 Java

macOS does not comes with an installed Java runtime (JRE) and a macOS upgrade may remove one if already installed: it is intended to be installed at first use. Check if a JRE is installed by running java -version in a Terminal window: if Java is not installed⁹⁰ this should prompt you to install it.⁹¹ You can also install directly a recent Java from Oracle). Builds of OpenJDK may also be available, e.g. from AdoptOpenJDK, Azul or https://jdk.java.net/. We recommend you install a version with long-term support, e.g. 8 or 11 (but not 9, 10, 12–16 which have/had a 6-month lifetime).

Binary distributions of R are built against a specific version (e.g. 1.8.0_241 or 11.0.6) of Java so sudo R CMD javareconf will likely be needed to be run before using Java-using packages.

To see what compatible versions of Java are currently installed, run /usr/libexec/java_home -V -a x86_64. If needed, set the environment variable JAVA_HOME to choose between these, both when R is built from the sources and when R CMD javareconf is run.

Configuring and building R both looks for a JRE and for support for compiling JNI programs (used by packages rJava and JavaGD); the latter requires a JDK (Java SDK) and not just a JRE⁹².

The build process tries to fathom out what JRE/JDK to use, but it may need some help, e.g. by setting JAVA_HOME. A Oracle JDK can be specified explicitly by something like

Note that it is necessary to set the environment variable NOAWT to 1 to install many of the Java-using packages.

C.3.6 Frameworks

The CRAN build of R is installed as a framework, which is selected by the option

(This is intended to be used with an Apple toolchain: others may not support frameworks correctly but those from https://releases.llvm.org do.)

It is only needed if you want to build R for use with the R.APP console, and implies --enable-R-shlib to build R as a dynamic library. This option configures R to be built and installed as a framework called R.framework. The default installation path for R.framework is /Library/Frameworks but this can be changed at configure time by specifying the flag --enable-R-framework[=DIR] (or --prefix) or at install time via

Note that installation as a framework is non-standard (especially to a non-standard location) and Unix utilities may not support it (e.g. the pkg-config file libR.pc will be put somewhere unknown to pkg-config).

C.3.7 Building R.app

Note that building the R.APP GUI console is a separate project, using Xcode. Before compiling R.APP make sure the current version of R is installed in /Library/Frameworks/R.framework and working at the command-line (this can be a binary install).

and built by loading the R.xcodeproj project (select the R target and a suitable configuration), or from the command-line by e.g.

R.APP does not need to be installed in any specific way. Building R.APP results in the R.APP bundle which appears as one R icon. This application bundle can be run anywhere and it is customary to place it in the /Applications folder.

C.3.8 Building binary packages

CRAN macOS binary packages are distributed as tarballs with suffix .tgz to distinguish them from source tarballs. One can tar an existing installed package, or use

R CMD
INSTALL --build

The CRAN binary packages are built with the Apple compiler on High Sierra, which avoids the first two and any issues with C++ libraries.

C.3.9 ‘arm64’

In November 2020 Apple launched Macs using ‘Apple Silicon’: this architecture has unfortunately many names, including ‘aarch64’ (in the FOSS world) ‘arm64’ and ‘arm64e’ (by Apple): the machines are also known as ‘M1’, Apple’s name for their (first-generation) CPU.

It is possible to build and run R and the vast majority of packages for ‘x86_64’ using ‘Rosetta’ emulation. To do so, install all the prerequisites and other software described above, but include the flag -target x86_64-apple-macos10.13 for clang and clang++, for example with a config.site file containing

(This may give warnings, in particular for the gfortran 8.2 distribution that libgfortran was built for 10.14 — but these seem harmless.)

C.3.9.1 Native builds

No Fortran compiler has been released for this architecture at the time of writing although there are ports under development.⁹³ So this is section is completely experimental.

At the time of writing no distribution of X11⁹⁴ was available, so R was configured with --without-x. The prerequisites. jpeg, libpng, pcre2,⁹⁵ readline, tiff, xz and Tcl/Tk⁹⁶ can easily be built from their sources (although the libedit emulation of readline will also work). You will also need pkgconfig, zlib-system-stub from https://mac.R-project.org/libs-4/ for pkg-config and support files.

As with Intel builds, it is possible to use the Accelerate library via the configuration option

C.4 Solaris

R has been built successfully on Solaris 10 using the (zero cost) Oracle Developer Studio⁹⁷ compilers: there has also been success with gcc/gfortran. (Recent Sun machines are AMD Opterons or Intel Xeons (‘amd64’) rather than ‘x86’, but 32-bit ‘x86’ executables are the default.) How these compilers identify themselves is slightly confusing: commands CC -V with Developer Studio 12.5 and 12.6 report as versions 5.14 and 5.15. We will only consider Developer Studio versions 12.5 (May 2016) and 12.6 (July 2017).

The Solaris versions of several of the tools needed to build R (e.g. make, ar and ld) are in /usr/ccs/bin, so if using those tools ensure this is in your path. A version of the preferred GNU tar is (if installed) in /usr/sfw/bin. It may be necessary to avoid the tools in /usr/ucb: POSIX-compliant versions of some tools can be found in /usr/xpg4/bin and /usr/xpg6/bin.

A large selection of Open Source software can be installed from https://www.opencsw.org, by default installed under /opt/csw. Solaris 10 ships with bzlib version 1.0.6 (sufficient) but zlib version 1.2.3 (too old): OpenCSW has 1.2.8. (Note from 2019: updating of OpenCSW has slowed or stopped.)

At least when compiling with Oracle compilers, Solaris uses far more stack space than other platforms. This makes it desirable to build PCRE1 (and PCRE2 < 10.30) with the option --disable-stack-for-recursion: the OpenCSW distribution was at the time of writing.

The Oracle compilers are unusual in not including /usr/local/include in the default include search path: R’s default CPPFLAGS=-I/usr/local/include remedies this. If you rely on OpenCSW software you may need CPPFLAGS=-I/opt/csw/include (or both).

You will need GNU libiconv and readline: the Solaris version of iconv is not sufficiently powerful.

The native make suffices to build R but a number of packages require GNU make (some without declaring it as ‘SystemRequirements’ in the DESCRIPTION file).

The support for the C99 long double type on Sparc hardware uses quad-precision arithmetic, and this is usually slow because it is done by software emulation. On such systems the configure option --disable-long-double can be used for faster but less accurate computations.

The Solaris time-zone conversion services seem to be unreliable pre-1916 in Europe (when daylight-savings time was first introduced): most often reporting in the non-existent DST variant. Using configure option --with-internal-tzcode is recommended, and required if you find time-zone abbreviations being given odd values (as has been seen on 64-bit builds without it).

When using the Oracle compilers do not specify -fast, as this disables IEEE arithmetic and make check will fail.

A little juggling of paths was needed to ensure GNU libiconv (in /usr/local) was used rather than the Solaris iconv:

The Oracle compilers do not by default conform to the C99 standard (appendix F 8.9) on the return values of functions such as log: use -xlibmieee to ensure this.

A peculiarity of some versions of the Fortran compiler has been that when asked to link a shared object they did not link against all the Fortran 9x runtime libraries, hence

Using -xlibmil in CFLAGS or FFLAGS allows more system mathematical functions to be inlined.

but the use of -nofstore can be less numerically stable, and some packages have in the past failed to compile at optimization level 5.

The Oracle compilers provide several implementations of the C++ standards which select both the set of headers and a C++ runtime library. One of those is selected by the -library flag, which as it is needed for both compiling and linking is best specified as part of the compiler or standard. Current R expects a C++11 compiler, for which the choice given above is the only possibility. Although version 12.5 accepted the flag -std=c++14, it did not pass configure’s conformance tests: version 12.6 does.

The performance library sunperf is available for use with the Oracle compilers. If selected as a BLAS, it must also be selected as LAPACK via

This has often given test failures in the past, in several different places.⁹⁸

Parsing very complex R expressions needs a lot of stack space when the Oracle compilers are used: several packages require the stack increased to at least 20MB.

Some people have reported that the Solaris libintl needs to be avoided, for example by using --disable-nls or --with-included-gettext or using libintl from OpenCSW. (On the other hand, there have been many successful installs which automatically detected libintl from OpenCSW or selected the included gettext.)

on the configure command line or in file config.site; however, there have been many successful installs without this.

C.4.1 64-bit builds

On both ‘x86’ and ‘Sparc’ platforms the compilers default to 32-bit code.

For a 64-bit target add -m64 to the compiler macros and use something like LDFLAGS=-L/usr/local/lib/amd64 or LDFLAGS=-L/usr/local/lib/sparcv9 as appropriate (and other 64-bit library directories if used, e.g. -L/opt/csw/lib/amd64). It will also be necessary to point pkg-config at the 64-bit directories, e.g. by something like

C.4.2 Using gcc

If using gcc, ensure that the compiler was compiled for the version of Solaris in use. (This can be ascertained from

gcc
-v

.) gcc makes modified versions of some header files, and several reports of problems were due to using gcc compiled on one version of Solaris on a later version. Note that this can even apply to OS patches: some 2016 patches to Solaris 10 changed its C header files in way incompatible⁹⁹ with the modified versions included with OpenCSW’s binary distribution.

The notes here are for gcc set up to use the Solaris linker: it can also be set up to use GNU ld, but that has not been tested. The tests were for compilers from the OpenCSW repository: Solaris systems often come with much older compilers installed under /usr/sfw/bin. One of -m32 or -m64 will be the default and could be omitted, but it is not easy to find out which. (For OpenCSW it is -m32.)

Note that paths such as /opt/csw/lib, /usr/local/lib/amd64 and /opt/csw/lib/amd64 may need to be in the LD_LIBRARY_PATH during configuration.

C.5 FreeBSD

There have been few recent reports on FreeBSD: there is a ‘port’ at https://www.freebsd.org/ports/math.html. Recent versions of FreeBSD use Clang and the libc++ C++ headers and runtime, but the ‘port’ is configured to use GCC.

Use of ICU for collation and the configure option --with-internal-tzcode are desirable workarounds.

C.6 OpenBSD

Ingo Feinerer installed R version 3.2.2 on OpenBSD 5.8 arch ‘amd64’ (their name for ‘x86_64’). Details of the build (and patches applied) are at https://cvsweb.openbsd.org/cgi-bin/cvsweb/ports/math/R/. (Downgrading the zlib requirement to 1.2.3 is against the advice of the R developers.)

C.7 Cygwin

The 32-bit version never worked well enough to pass R’s

make
check

, and residual support from earlier experiments was removed in R 3.3.0.

C.8 New platforms

There are a number of sources of problems when installing R on a new hardware/OS platform. These include

Floating Point Arithmetic: R requires arithmetic compliant with IEC 60559, also known as IEEE 754. This mandates the use of plus and minus infinity and NaN (not a number) as well as specific details of rounding. Although almost all current FPUs can support this, selecting such support can be a pain. The problem is that there is no agreement on how to set the signalling behaviour; Sun/Sparc, SGI/IRIX and ‘ix86’ Linux require no special action, FreeBSD requires a call to (the macro) fpsetmask(0) and OSF1 required that computation be done with a -ieee_with_inexact flag etc. With Intel compilers on 32-bit and 64-bit Intel machines, one has to explicitly disable flush-to-zero and denormals-are-zero modes. Some ARM processors including A12Z and M1 (Apple Silicon) by default use runfast mode, which includes flush-to-zero and default-nan and hence has to be disabled. With default-nan mode, the NaN payload used for representation of numeric NA values is lost even on simple operations with finite values. On a new platform you must find out the magic recipe and add some code to make it work. This can often be done via the file config.site which resides in the top level directory.

Beware of using high levels of optimization, at least initially. On many compilers these reduce the degree of compliance to the IEEE model. For example, using -fast on the Oracle compilers has caused R’s NaN to be set incorrectly, and gcc’s -ffast-math and clang’s -Ofast have given incorrect results.

Shared Objects: There seems to be very little agreement across platforms on what needs to be done to build shared objects. there are many different combinations of flags for the compilers and loaders. GNU libtool cannot be used (yet), as it currently does not fully support Fortran: one would need a shell wrapper for this). The technique we use is to first interrogate the X window system about what it does (using xmkmf), and then override this in situations where we know better (for tools from the GNU Compiler Collection and/or platforms we know about). This typically works, but you may have to manually override the results. Scanning the manual entries for cc and ld usually reveals the correct incantation. Once you know the recipe you can modify the file config.site (following the instructions therein) so that the build will use these options.

It seems that gcc 3.4.x and later on ‘ix86’ Linux defeat attempts by the LAPACK code to avoid computations entirely in extended-precision registers, so file src/modules/lapack/dlamc.f may need to be compiled without optimization or with additional flags. Set the configure variable SAFE_FFLAGS to the flags to be used for this file.

If you do manage to get R running on a new platform please let us know about it so we can modify the configuration procedures to include that platform.

If you are having trouble getting R to work on your platform please feel free to use the ‘R-devel’ mailing list to ask questions. We have had a fair amount of practice at porting R to new platforms ...

Function and variable index

Concept index

Environment variable index

Jump to:	C I M R U

Jump to:	C I M R U

Jump to:	B F I L M O P R S U V

Jump to:	B F I L M O P R S U V

Jump to:	B C D J L P R T

Jump to:	B C D J L P R T

Footnotes

(1)

e.g. GNU tar version 1.15 or later, or that from the ‘libarchive’ (as used on macOS) or ‘Heirloom Toolchest’ distributions.

(2)

for some Subversion clients ‘http:’ may appear to work, but requires continual redirection.

(3)

Most aspects will work with paths containing spaces, but external software used by R may not.

(4)

which use lib rather than lib64 for their primary 64-bit library directories: attempts are made to detect such systems.

(5)

Instructions on how to install the latest version are at https://www.ctan.org/tex-archive/fonts/inconsolata/.

(6)

on a Unix-alike, ‘inconsolata’ is omitted if not found by configure.

(7)

This will be needed if more than one sub-architecture is to be installed.

(8)

with possible values ‘i386’, ‘x64’, ‘32’ and ‘64’.

(9)

mainly on RedHat and Fedora, whose layout is described here.

(10)

How to prepare such a directory is described in file src/extra/tzone/Notes in the R sources.

(11)

for example, -fopenmp, -xopenmp or -qopenmp. This includes for clang and the Intel and Oracle compilers.

(12)

This does not necessarily disable use of OpenMP – the configure code allows for platforms where OpenMP is used without a flag. For the flang compiler in late 2017, the Fortran runtime always used OpenMP.

(13)

This was a valid option for g++ 5 and later and 2016 versions of the Intel and Solaris compilers. For earlier versions of g++ one could try -std=c++1y.

(14)

Then recommended packages installed as part of the R installation do use LTO, but not packages installed later.

(15)

A complete CRAN installation reduced from 50 to 35GB.

(16)

although there is the possibility to exclude Fortran but that misses some of the benefits.

(17)

not NM as we found make overriding that.

(18)

probably also 8.4 and later.

(19)

Capitalization of this project’s name is inconsistent even on its own webpages.

(20)

There are reports of segfaults when ‘MiKTeX’ installs additional packages when making NEWS.pdf: re-running make seems to solve this.

(21)

macOS 11 ‘Big Sur’ has been tested on ‘arm64’ (aka ‘Apple Silicon’ aka ‘M1’) using ‘Rosetta’ emulation: there you need to install the Intel build of XQuartz.

(22)

At least if otool is installed: it is part of the macOS Command Line Tools.

(23)

The installer puts links to R and Rscript in /usr/local/bin. If these are missing (as they may be under ‘Big Sur’, you can run directly the copies in /Library/Frameworks/R.framework/Resources/.

(24)

At the time of writing: use pkgutil --pkgs | grep org.r-project to check..

(25)

More precisely, of the Apple package of the same name: this means that installing a package for 3.6.x does not remove an installation for 4.0 or later.

(26)

The Oracle compilers on Solaris (where the issue is parsing very complex R expressions) and GCC 9 on Linux.

(27)

unless they were excluded in the build.

(28)

its binding is locked once the startup files have been read, so users cannot easily change it.

(29)

If a proxy needs to be set, see ?download.file.

(30)

for a small number of CRAN packages where this is known to be safe and is needed by the autobuilder this is the default. Look at the source of tools:::.install_packages for the list. It can also be specified in the package’s DESCRIPTION file.

(31)

Note that capitalization and versioning may differ from the Open Source project.

(32)

using a path containing spaces is likely to cause problems

(33)

They need to have been created using -headerpad_max_install_names, which is the default for an R package.

(34)

‘X/Open Portability Guide’, which has had several versions.

(35)

On some systems setting LC_ALL or LC_MESSAGES to ‘C’ disables LANGUAGE.

(36)

If you try changing from French to Russian except in a UTF-8 locale, you will most likely find messages change to English.

(37)

the language written in England: some people living in the USA appropriate this name for their language.

(38)

with Americanisms.

(39)

also known as IEEE 754

(40)

at least when storing quantities: the on-FPU precision is allowed to vary

(41)

e.g. Bessel, beta and gamma functions

(42)

including copying MkRules.dist to MkRule.local and selecting the architecture.

(43)

also known as IEEE 754

(44)

Note that C11 compilers need not be C99-compliant: R requires support for double complex and variable-length arrays which are optional in C11 but are mandatory in C99. C18 (also known as C17) is a ‘bugfix release’ of C11, clarifying the standard.

(45)

Examples are -std=gnu99, -std=c99 and -c99.

(46)

-std=c99 excludes POSIX functionality, but config.h will turn on all GNU extensions to include the POSIX functionality for R itself: this does not apply to badly-written packages. The default mode for GCC 5.1 and later is -std=gnu11, which currently includes the optional features R needs.

(47)

However, it is possible to break the default behaviour of glibc by re-specifying the gconv modules to be loaded.

(48)

specifically, the C99 functionality of headers wchar.h and wctype.h, types wctans_t and mbstate_t and functions mbrtowc, mbstowcs, wcrtomb, wcscoll, wcstombs, wctrans, wctype, and iswctype.

(49)

including expm1, hypot, log1p, nearbyint and va_copy.

(50)

including opendir, readdir, closedir, popen, stat, glob, access, getcwd and chdir system calls, select on a Unix-alike, and either putenv or setenv.

(51)

such as realpath, symlink.

(52)

most often distributed as part of xz: possible names in Linux distributions include xz-devel/xz-libs and liblzma-dev.

(53)

but not a major version greater than 7 should there ever be one: the major version has been 7 since 2000.

(54)

for example to specify static linking with a build which has both shared and static libraries.

(55)

Such as GNU tar 1.15 or later, bsdtar (from https://github.com/libarchive/libarchive/, used as tar by FreeBSD and macOS 10.6 and later) or tar from the Heirloom Toolchest (http://heirloom.sourceforge.net/tools.html), although the latter does not support xz compression.

(56)

texi2dvi is normally a shell script. Some versions (including those from texinfo 5.2 and 6.0-6.6) need to be run under bash rather than a Bourne shell, especially on Solaris. Some of the issues which have been observed with broken versions of texi2dvi can be circumvented by setting the environment variable R_TEXI2DVICMD to the value emulation.

(57)

If necessary the path to pkg-config can be specified by setting PKG_CONFIG in config.site, on the configure command line or in the environment. There is a compatible re-implementation of pkg-config called pkgconf which can be used in the unlikely event that is installed but not linked to pkg-config.

(58)

also known as ttf-mscorefonts-installer in the Debian/Ubuntu world: see also https://en.wikipedia.org/wiki/Core_fonts_for_the_Web.

(59)

ttf-liberation in Debian/Ubuntu.

(60)

This is the default as from glibc 2.26 and has been confirmed for Fedora >= 28, which does not mention this on its man pages.

(61)

R uses rpc/xdr.h but that includes netconfig.h from the top tirpc directory.

(62)

This is true even for the ‘Aqua’ version of Tk on macOS, but distributions of that include a copy of the X11 files needed.

(63)

The search includes OpenBLAS, ATLAS and a generic libblas, plus some platform-specific choices (see below).

(64)

Using the Oracle Developer Studio cc and f95 compilers

(65)

There were others for earlier versions of ATLAS, and are for non-Intel architectures. The only way to see exactly which CPUs the distributed libraries have been tuned for is to read the atlas.spec file: at the time of writing ‘HAMMER64SSE3’ and ‘Corei264AVX’ for ‘x86_64’ Fedora.

• Obtaining R:
• Installing R under Unix-alikes:
• Installing R under Windows:
• Installing R under macOS:
• Running R:
• Add-on packages:
• Internationalization:
• Choosing between 32- and 64-bit builds:
• The standalone Rmath library:
• Essential and useful other programs under a Unix-alike:
• Configuration on a Unix-alike:
• Platform notes:
• Function and variable index:
• Concept index:
• Environment variable index:

• Simple compilation:
• Help options:
• Making the manuals:
• Installation:
• Uninstallation:
• Sub-architectures:
• Other Options:
• Testing a Unix-alike Installation:

• Debugging Symbols:
• OpenMP Support:
• C++ Support:
• Link-Time Optimization:

• Default packages:
• Managing libraries:
• Installing packages:
• Updating packages:
• Removing packages:
• Setting up a package repository:
• Checking installed source packages:

• Windows packages:
• macOS packages:
• Customizing package compilation:
• Multiple sub-architectures:
• Byte-compilation:
• External software:

• Getting and unpacking the sources:
• Getting patched and development versions:

• LTO with GCC:
• LTO with LLVM:
• LTO for package checking:

• Building from source:
• Checking the build:
• Testing a Windows Installation:

• Running R under macOS:
• Uninstalling under macOS:
• Multiple versions:

• Locales under Unix-alikes:
• Locales under Windows:
• Locales under macOS:

• Essential programs and libraries:
• Useful libraries and programs:
• Linear algebra:

• Configuration options:
• Internationalization support:
• Configuration variables:
• Setting the shell:
• Using make:
• Using Fortran:
• Compile and load flags:
• Maintainer mode:

• Setting paper size:
• Setting the browsers:
• Compilation flags:
• Making manuals:

• X11 issues:
• Linux:
• macOS:
• Solaris:
• FreeBSD:
• OpenBSD:
• Cygwin:
• New platforms:

• Prerequisites:
• Other C/C++ compilers:
• Other libraries:
• Tcl/Tk headers and libraries:
• Java (macOS):
• Frameworks:
• Building R.app:
• Building binary packages:
• arm64:

	Index Entry	Section

C
	`configure`:	Simple compilation
	`configure`:	Simple compilation
	`configure`:	Installation
	`configure`:	Installation
	`configure`:	Configuration variables
	`configure`:	Using make

I
	`install.packages`:	Installing packages

M
	`make`:	Using make

R
	`remove.packages`:	Removing packages
	`R_HOME`:	Simple compilation

U
	`update.packages`:	Updating packages

	Index Entry	Section

B
	BLAS library:	Linear algebra
	BLAS library:	Other libraries
	BLAS library:	Solaris

F
	Fortran:	Using Fortran
	FreeBSD:	FreeBSD

I
	Installation:	Installation
	Installing under Unix-alikes:	Installing R under Unix-alikes
	Installing under Windows:	Installing R under Windows
	Internationalization:	Internationalization

L
	LAPACK library:	LAPACK
	LAPACK library:	Other libraries
	LAPACK library:	Solaris
	Libraries:	Add-on packages
	Libraries, managing:	Managing libraries
	Libraries, site:	Managing libraries
	Libraries, user:	Managing libraries
	Linux:	Installing R under Unix-alikes
	Linux:	Linux
	Locale:	Internationalization
	Locale:	Locales
	Localization:	Internationalization

M
	macOS:	Installing R under Unix-alikes
	macOS:	Installing R under macOS
	macOS:	macOS
	Manuals:	Making the manuals
	Manuals, installing:	Installation

O
	Obtaining R:	Obtaining R
	OpenBSD:	OpenBSD

P
	Packages:	Add-on packages
	Packages, default:	Default packages
	Packages, installing:	Installing packages
	Packages, removing:	Removing packages
	Packages, updating:	Updating packages

R
	Repositories:	Setting up a package repository

S
	Site libraries:	Managing libraries
	Solaris:	Solaris
	Sources for R:	Getting and unpacking the sources
	Subversion:	Using Subversion and rsync
	Subversion:	Essential programs and libraries

U
	User libraries:	Managing libraries

V
	Vignettes:	Essential programs and libraries

	Index Entry	Section

B
	`BLAS_LIBS`:	BLAS

C
	`CONFIG_SITE`:	Configuration variables

D
	`DESTDIR`:	Installation
	`DESTDIR`:	Unix-alike standalone

J
	`JAVA_HOME`:	Java support

L
	`LANG`:	Localization of messages
	`LANGUAGE`:	Localization of messages
	`LANGUAGE`:	Localization of messages
	`LAPACK_LIBS`:	LAPACK
	`LC_ALL`:	Localization of messages
	`LC_COLLATE`:	Testing a Unix-alike Installation
	`LC_MESSAGES`:	Localization of messages
	`LD_LIBRARY_PATH`:	Unix-alike standalone
	`LD_LIBRARY_PATH`:	Using Fortran
	`LD_LIBRARY_PATH`:	Compile and load flags
	`LD_LIBRARY_PATH`:	Using gcc

P
	`PAPERSIZE`:	Setting paper size
	`PATH`:	Essential programs and libraries
	`PATH`:	Using Fortran

R
	`R_ARCH`:	Sub-architectures
	`R_ARCH`:	Sub-architectures
	`R_BROWSER`:	Setting the browsers
	`R_DEFAULT_PACKAGES`:	Default packages
	`R_DISABLE_HTTPD`:	Help options
	`R_GSCMD`:	Useful libraries and programs
	`R_INSTALL_TAR`:	Windows packages
	`R_JAVA_LD_LIBRARY_PATH`:	Java support
	`R_JAVA_LD_LIBRARY_PATH`:	Java support
	`R_LIBS`:	Add-on packages
	`R_LIBS_SITE`:	Managing libraries
	`R_LIBS_USER`:	Managing libraries
	`R_PAPERSIZE`:	Making the manuals
	`R_PAPERSIZE`:	Running R
	`R_PAPERSIZE`:	Setting paper size
	`R_PAPERSIZE`:	Making manuals
	`R_PDFVIEWER`:	Setting the browsers
	`R_RD4PDF`:	Making the manuals
	`R_RD4PDF`:	Making manuals
	`R_SCRIPT_DEFAULT_PACKAGES`:	Default packages
	`R_USER`:	Running R

T
	`TAR`:	Essential programs and libraries
	`TAR_OPTIONS`:	Getting and unpacking the sources
	`TEMP`:	Running R
	`TMP`:	Running R
	`TMPDIR`:	Simple compilation
	`TMPDIR`:	Running R
	`TMPDIR`:	Running R
	`TMPDIR`:	Installing packages

R Installation and Administration

Table of Contents

R Installation and Administration

1 Obtaining R

1.1 Getting and unpacking the sources

1.2 Getting patched and development versions

1.2.1 Using Subversion and rsync

2 Installing R under Unix-alikes

2.1 Simple compilation

2.2 Help options

2.3 Making the manuals

2.4 Installation

2.5 Uninstallation

2.6 Sub-architectures

2.6.1 Multilib

2.7 Other Options

2.7.1 Debugging Symbols

2.7.2 OpenMP Support

2.7.3 C++ Support

2.7.4 Link-Time Optimization

2.7.4.1 LTO with GCC

2.7.4.2 LTO with LLVM

2.7.4.3 LTO for package checking

2.8 Testing an Installation

3 Installing R under Windows

3.1 Building from source

3.1.1 The Windows toolset

3.1.2 LaTeX

3.2 Checking the build

3.3 Testing an Installation

4 Installing R under macOS

4.1 Running R under macOS

4.2 Uninstalling under macOS

4.3 Multiple versions

5 Running R

6 Add-on packages

6.1 Default packages

6.2 Managing libraries

6.3 Installing packages

6.3.1 Windows

6.3.2 macOS

6.3.3 Customizing package compilation

6.3.4 Multiple sub-architectures

6.3.5 Byte-compilation

6.3.6 External software

6.4 Updating packages

6.5 Removing packages

6.6 Setting up a package repository

6.7 Checking installed source packages

7 Internationalization and Localization

7.1 Locales

7.1.1 Locales under Unix-alikes

7.1.2 Locales under Windows

7.1.3 Locales under macOS

7.2 Localization of messages

8 Choosing between 32- and 64-bit builds

9 The standalone Rmath library

9.1 Unix-alikes

9.2 Windows

Appendix A Essential and useful other programs under a Unix-alike

A.1 Essential programs and libraries

A.2 Useful libraries and programs

A.2.1 Tcl/Tk

A.2.2 Java support

A.2.3 Other compiled languages

A.3 Linear algebra

A.3.1 BLAS

A.3.1.1 ATLAS

A.3.1.2 OpenBLAS

A.3.1.3 Intel MKL

A.3.1.4 Shared BLAS

A.3.2 LAPACK

A.3.3 Caveats

Appendix B Configuration on a Unix-alike

B.1 Configuration options

B.2 Internationalization support

B.3 Configuration variables

B.3.1 Setting paper size

B.3.2 Setting the browsers

B.3.3 Compilation flags

C.3.9 ‘`arm64`’