Dar Documentation

How to compile dar and libdar

There are two ways of compiling dar. The easy way and the less easy way:

The easy way:
This is when you grab the latest source package (file in the format dar-X.Y.Z.tar.gz where X.Y.Z is the dar version) and follow the requirements just below. The easy way also applies to interim releases when they exist, which are also fully supported and production grade software.
The less easy way:
This is when you grab the source code from a GIT repositories. You will then have to follow the preliminary steps, then continue following the easy way.

Requirements

To compile dar from a source package you need at least the following:

  1. a C++ compiler supporting C++14 syntax (for gcc that means version 6.1. minimum, clang is also supported). For version older than 2.7.0 only C++11 is required (at least gcc version 4.9), while version older than dar-2.5.0 C++11 support is not required but if using gcc you need at least version 3.4.x. A standard C++ library is also required, compilation has been tested with libc++ and stdlibc++.
  2. a linker like "ld" the GNU Linker
  3. the make program (tested with gnu make)

In option you may also have installed the following tools and libraries:

The consequences of building dar without these optional tools are the following:

Dependencies in distro packages

For simplicity, here follows the package names that provide the previously mentionned libdar dependencies. If you have the equivalent names for other distro, feel free to contact the dar's manager maintainer for this table to be updated.

Distro Debian/Devuan/Ubuntu
gzip compression zlib1g-dev
bzip2 compression libbz2-dev
lzo compression liblzo2-dev
xz compression liblzma-dev
zstd compression libzstd-dev
lz4 compression liblz4-dev
symmetric strong encryption libgcrypt-dev
asymmetric strong encryption libgpgme-dev
documentation buiding doxygen
graphs in documentation graphviz
upx compressed binaries upx
man page in html groff
ext2/3/4 specific support libext2fs-dev
multi-threading libthreadar has to be installed manually
binary delta librsync-dev
embedded sftp and ftp support libcurl-dev
python binding pybind11-dev
argon key derivation fonction libargon2-dev

Compilation Process

Once you have the minimum requirements, Dar has to be compiled from source code in the following way:

./configure [eventually with some options] make make install-strip
Important:

due to a bug in the autoconf/libtool softwares used to build the configure script you must not have spaces in the name of the path where are extracted dar' sources. You can install dar binary anywhere you want, the problem does not concern dar itself but the ./configure script used to build dar: To work properly it must not be ran from a path which has a space in it.

Important too:

By default the configure script set optimization to -O2, depending on the compiler this may lead to problems in the resulting binary (or even in the compilation process), before reporting a bug try first to compile with less optimization:

CXXFLAGS=-O export CXXFLAGS make clean distclean ./configure [options...] make make install-strip

The configure script may receive several options (listed here), in particular the --prefix option let you install dar and libdar in another place than the default /usr/local. For example to have dar installed under /usr use the following: ./configure --prefix=/usr. You will be able to uninstall dar/libdar by calling make uninstall this implies keeping the source package directory around and using the same option given to ./configure as what has been used at installation time.

If you prefer building a package without installing dar, which is espetially suitable for package maintainers, the DESTDIR variable may be set at installation time to install dar in another root directory. This makes the creation of dar binary packages very easy. Here is an example:

./configure --prefix=/usr [eventually with some options] make make DESTDIR=/some/where install-strip

As result of the previous example, dar will be installed in /some/where/usr/{bin | lib | ...} directories, but built as if it was installed in /usr, thus it will look for /etc/darrc and not /some/where/etc/darrc. You can thus build a package from files under /some/where by mean of a pakage management tool, then install/remove this package as if it was provided by your distro maintainer.

Options for the configure script

Available options for the configure script

Optimization option:
--enable-mode

--enable-mode=32 or --enable-mode=infinint

if set,replace 64 bits integers used by default by 32 bits integers or "infinint" integers (limitless integers). Before release 2.6.0 the default integer used was infinint and this option was added for speed optimization at the cost of some limitations (See the limitations for more).

Since release 2.6.0, the default is 64 bits integers (limitations stay the same) instead of infinint. But if you hit the 64 bits integer limitations you can still use infinint which overcome this at the cost of slower performance and increase memory requirement.

Deactivation options:
--disable-largefile Whatever your system is, dar will not be able to handle file of size larger than 4GB
--disable-ea-support Whatever your system is, dar will not be able to save or restore Extended Attributes (see the Notes paragraphs I and V)
--disable-nodump-flag Whatever your system is, dar will not be able to take care of the nodump-flag (thanks to the --nodump option)
--disable-linux-statx Even if your system provides the statx() system call, dar will ignore it and will not save birthtime of files as Linux FSA. At restoration time you will not be bothered by the warning telling you that birthime is not possible to be restored under Linux, if you still want to restore other Linux FSA.
--disable-dar-static dar_static binary (statically linked version of dar) will not be built
--disable-special-alloc dar uses a special allocation scheme by default (gather the many small allocations in big fewer ones), this improves dar's execution speed
--disable-upx If upx is found in the PATH, binary are upx compressed at installation step. This can be disabled by this option, when upx is available and you don't want compressed binaries.
--disable-gnugetopt on non GNU systems (Solaris, etc.) configure looks for libgnugetopt to have the long options support thanks to the gnu getopt_long() call, this can be disabled.
--disable-thread-safe libdar may need POSIX mutex to be thread safe. If you don't want libdar relaying on POSIX mutex even if they are available, use this option. The resulting library may not be thread safe. But it will always be thread safe if you use --disable-special-alloc, and it will never be thread safe if --enable-test-memory is used.
--disable-libdl-linking Ignore any libdl library and avoid linking with it
--disable-libz-linking Disable linking to libz, thus -zgzip:* option (gzip compression) will not be available
--disable-libbz2-linking Disable linking to libbz2, thus -zbzip2:* option (libbz2 compression) will not be available
--disable-liblzo2-linking Disable linking to liblzo2, thus -zlzo:* option (lzo compression) will not be available
--disable-libxz-linking Disable linking to liblzma5 this -zxz:* option (xz compression) will not be available
--disable-libgcrypt-linking Disable linking with libgcrypt library. Strong encryption will not be available neither a hashing of generated slices.
--disable-gpgme-linking Disable linking with gpgme library. Asymetric strong encryption algorithms will not be available
--disable-build-html Do not build API documentation reference with Doxygen (when it is available)
--disable-furtive-read Do not try to detect whether the system does support furtive read mode. This will lead furtive read mode to stay disabled in any case.
--disable-fast-dir Disable optimization for large directories, doing so has a little positive impact on memory requirement but a huge drawback on execution time
--disable-execinfo Disable reporting stack information on self diagnostic bugs even
--disable-threadar Avoid linking with libthreadar even if available, libdar will not create threads
--disable-birthtime Disable the HFS+ Filesystem Specific Attribute support
--disable-librsync-linking Disable linking with librsync, thus delta binary will not be available
--disable-libcurl-linking Disable linking with libcurl, thus remote repository support using ftp or sftp will not be available
--enable-limit-time-accuracy={s|us|ns} Limit the timestamp precision of files to seconds, microseconds (lowercase U not the μ greek letter for μs) and nanoseconds respectively, by default dar uses the maximum time precision supported by the operating system.
Troubleshooting option:
--enable-os-bits If set, dar uses the given argument (32 or 64) to determine which integer type to use. This much match your CPU register size. By default dar uses the system <stdint.h> file to determine the correct integer type to use
Debugging options:
--enable-examples If set, example programs based on infinint will also be built
--enable-debug If set, use debug compilation option, and if possible statically link binaries
--enable-pedantic If set, transmits the -pedantic option to the compiler
--enable-profiling Enable executable profiling
--enable-debug-memory If set, logs all memory allocations and releases to /tmp/dar_debug_mem_allocation.txt. The resulting executable is expected to be very slow

GIT

Presentation

To manage its source code versions DAR uses GIT (it used CVS up to Q1 2012 and even the older RCS in 2002). Since September 2nd 2017, the GIT repository has been cloned to GitHub while the original repository at Sourceforge still stays updated. Both should contain the exact same code, that's up to you to choose the repository you prefer. For sanity also, starting September 3rd 2017, git tags including those for release candidates will be signed with GPG.

Dar's repository Organization

GIT (more than CVS) eases the use of branches. In dar repository, there are thus a lot of them: the first and main one is called "master". It contains current development and most probably unstable code. There are other permanent branches that hold stable code. They are all named by "branch_A.B.x" where A and B are the numbers corresponding to a released versions family. For example, "branch_2.6.x" holds the stable code for releases 2.6.0, 2.6.1, 2.6.2 and so on. It also holds pending fixes for the next release on that branch you might be interested in.

The global organisation of the repository is thus the following:

  (HEAD of "master" branch)   new feature 101   |   ^   |   new feature 100   |   ^   |   new feature 99   |   +--->-- fix 901 ->- fix 902 (release 2.4.1) ->- fix 903 ->- fix 904 (release 2.4.2) ->- fix 905 (HEAD of branch_2.4.x)   |   new feature 98   |   ^   |   new feature 97   |   +--->-- fix 801 ->- fix 802 (release 2.3.1) (also HEAD of branch_2.3.x as no pending fix is waiting for release)   |   ...   |   ^   |   initial version

Usage

To get dar source code from GIT you have first to clone the repository hosted at sourceforge:

git clone https://github.com/Edrusb/DAR.git cd DAR

You will probably not want to use current development code so you have to change from the branch master to the branch "branch_A.B.x" of you choice:

git checkout branch_2.6.x

That's all. You now have the most recent stable code (for branch_2.6.x in this example). To see what changes have been brought since the last release, use the following command:

git log

If you plan to keep the repository you've cloned, updating the change is as easy as using (no need to clone the repository from scratch again):

git pull

There is also a web interface to git at github

Having the sources ready for compilation

Please read the fiile named USING_SOURCE_FROM_GIT located at the root of the directory tree you retrieved through GIT, it contains up to date information about the required tools and how to generate the configuration file. Then you can proceed to source compilation as done with regular source package

Related Softwares