Backing up darcs
----------------

While revision control ideally means that you can recreate any
point in the development history of your project, it is sometimes
comfortable to retain snapshots of the whole thing at regular
intervals.

The simplest way is probably to just back up the **\_darcs**
directory. To restore a working copy, just unpack the appropriate
**\_darcs** and do a **darcs revert**.

Makefile
~~~~~~~~

For simplicity of use, I've put the necessary commands in a
Makefile in my home directory:

::

    DARCS=`find . -name _darcs -type d | grep -v trash`
    DATE=`date +%Y-%m-%d`
    
    backup:
            tar jcf archive/darcs-$(DATE).tbz $(DARCS)
    
    sync:
            rsync --rsh=ssh -av archive host.elsewhere.org:archive

Now **make backup** will find all darcs repositories, except any
with the word *trash* in their name or path, and **make sync** will
push it all to a remote computer, so that not all is lost when (and
not if) the disk on my laptop breaks down.

Keeping unrecorded changes
~~~~~~~~~~~~~~~~~~~~~~~~~~

The observant reader has probably noticed that any unrecorded
changes will not be backed up. This is easily resolved by doing a
**darcs diff** with output to a file, and including this file in
the backup.

I'll leave the exact commands and Makefile modifications as an
exercise for the reader :-)

Questions
~~~~~~~~~

For a shared repository, or an automated backup process that
accesses a live repository, is it correct to assume that one should
make a temporary copy of the repository, using darcs, and then back
that up in order to ensure a good snapshot of its state?