Difference between revisions of "Subversion"

From SourceWiki
Jump to navigation Jump to search
 
(141 intermediate revisions by 2 users not shown)
Line 1: Line 1:
 
[[Category:Pragmatic Programming]]
 
[[Category:Pragmatic Programming]]
  
=Version control concepts=
+
=Introduction=
Subversion is a centralised version control system. Centralised version control means that a copy of your project is held in a central location called the '''repository''' and the subversion server logs all operations happening on the repository: everytime something is changed in the repository, the server logs the '''time and date''', the ''changes''', the '''author''' as well as a '''log message'''. The server can be configured to let some people do actions that other cannot. For instance, in the practical, the server allows anonymous read-only access but only a selected number of people can changes things.
 
  
All the operations described above (logging and authentication) happen on the server. However, the server is only accessible directly to system administrators. To interact with the server, a user makes use of a subversion '''client'''. Some of you might already know about some graphical subversion clients such as TortoiseSVN. This practical will show how the command line client can be used. The subversion client can be used to (1) ask information from and (2) send information to the server. The client can also be used to get informationabout your '''working copy''' which is the local copy of the project that resides on your filespace. You can use the client to ask questions such as:
+
In this workshop, we'll look at using a particular Version Control System (VCS) called Subversion (often abbreviated to SVN).  Before getting into the nitty-gritty of using SVN, we'll pause to consider the motivations for adopting version control and also the key concepts that are common to most available systems.
 +
 
 +
==Why is Version Control useful?==
 +
 
 +
OK, here's the sales pitch:
 +
 
 +
* It '''removes confusion''' about versions.  For example, you will no longer have to keep inventing names for different versions of essentially the same document  e.g. blah.old, blah.sav, blah.older, blah.newest2 (look familiar?).
 +
* It makes '''collaborative working''' easier.  Version control assists in coordination as it removes any confusion about versions, highlights conflicts, allows the use of independent working copies, records log messages and much more besides.
 +
* It makes '''distributing your code''' easier.  A version control repository can be visible to the world (often as a URL).  However, using some highly customisable access controls, you can arrange for some (perhaps anyone) to download your project while also specifying that only a select few may be trusted to upload.
 +
* It makes '''reproducing experiments''' easier.  The ability to reproduce an experiment is a ''key characteristic of science''.  However, all too often, in the digital age, people are unable to run the same version of a model that they ran six months ago.  With version control, you can always access any previous version of your model.
 +
* It aids '''disaster recovery'''.  You computer is fried?  No problem!  Just checkout your code to another and you're working productively again in minutes.
 +
 
 +
=Version Control Concepts=
 +
 
 +
A picture can be worth a thousand words, so let's try illustrating some of the key version control concepts, before wading into acres of text:
 +
 
 +
[[Image:Svn-cartoon.jpg|700px|thumbnail|center|Files stored in a repository; a checkout; modification and commit.  All versions recorded.]]
 +
[[Image:File-tree.jpg|400px|thumbnail|center|A checkout adds a copy of the files held in the repository to your local computer.]]
 +
 
 +
<!--
 +
usernames are countries:
 +
 
 +
'''greece, germany, switzerland, egypt, ireland, cuba, finland, portugal, england, spain, russia, norway, canada, france, italy, japan'''
 +
 
 +
files are capital cities:
 +
 
 +
'''athens, berlin, bern, cairo, dublin, havana, helsinki, lisbon, london, madrid, moscow, oslo, ottawa, paris, rome, tokyo'''
 +
-->
 +
 
 +
Subversion is a centralised version control system. Centralised version control means that a copy of your project is held in a central location called the '''repository''' and the subversion server logs all operations happening on the repository: every time something is changed in the repository, the server logs the '''time and date''', the '''changes''', the '''author''' as well as a '''log message'''. The server can be configured to give privavcy; allowing some people actions which are disallowed for others. For instance, in the practical, the server allows anonymous read-only access but only a selected number of people can changes things.
 +
 
 +
All the operations described above (logging and authentication) happen on the server. However, the server is only accessible directly to system administrators. To interact with the server, a user makes use of a subversion '''client'''. Some of you might already know about some graphical subversion clients such as TortoiseSVN (see the screen grab below). This practical will show how the command line client can be used. The subversion client can be used to (1) ask information from and (2) send information to the server. The client can also be used to get information about your '''working copy''' which is the local copy of the project that resides on your filespace. You can use the client to ask questions such as:
 
* which files have I modified since I last synchronised with the server?
 
* which files have I modified since I last synchronised with the server?
 
* when was that file last modified?
 
* when was that file last modified?
* who wrote the stupid bug at line 18 in file foo.c?
+
* who, inadvertently, created a bug at line 18 in file foo.c?
 
* what has changed in that file?
 
* what has changed in that file?
  
The repository that we are going to use for this practical is called subversion and is hosted in the "open" part of source server. You can access the repository directly by naviguating to [http://source.ggy.bris.ac.uk/subversion-open/subversion/trunk http://source.ggy.bris.ac.uk/subversion-open/subversion/trunk]. This is not a very intuitive interface and we also provide a nicer interface called [http://source.ggy.bris.ac.uk/websvn websvn] and select the "subversion" project.
+
<gallery widths=500px heights=350px perrow=2>
 +
File:Tsvn_switch1.png|TortoiseSVN for MS Windows
 +
File:Svn-cli.png|svn command line client for Linux
 +
</gallery>
 +
 
 +
=Acquiring a Repository=
 +
 
 +
For the purposes of this practical, you can get yourself a repository from one of the hosting sites that can be found out in the cloud.  Or you could use another repository that you have access to--perhaps hosted here in the University, some other portion of UK academia or elsewhere. I've prepared the examples using a repository accessed via the GitHub website (https://github.com). '''Note, however, that to obtain a free repository on GitHub, you must agree to it being readable by anyone.''' 
 +
 
 +
'''NB With that in mind, you may want to have a think about the right home for any of your intellectual property.'''
 +
 
 +
OK, let's assume that you are happy to work with a GitHub hosted repository--at least for your initial steps learning about version control and subversion. (A natty feature of GitHub repositories is that they can be used with both Subversion and Git VCS.)
  
If you don't have a subversion login, use [http://source.ggy.bris.ac.uk/websvn-open this link] instead, but then you will '''not''' be able to send anything to the server.
+
Registering and creating a repository is easy, just follow the instructions on the webpage:
  
Keep a websvn browser window open to check the state of the repository throughout the practical.
+
[[Image:Github.png|500px|thumbnail|center|The Github web interface]]
  
=svn: the subversion command line client=
+
Be sure to check the box: '''Initialize this repository with a README'''
The subversion command line client is called "svn". To execute a subversion command, simply type:
+
 
 +
In addition to the command line client that we will describe in the following sections, you can manage your repository, view and even edit your files through the GitHub website:
 +
 
 +
[[Image:Github-ggdagw-test.png|500px|thumbnail|center|A test repository]]
 +
 
 +
=svn: The Subversion Command Line Client=
 +
The subversion command line client is called '''svn'''. To execute a subversion command, simply type:
 
<pre>
 
<pre>
 
$ svn command arguments
 
$ svn command arguments
 
</pre>
 
</pre>
  
Somme command can also use options which are given withdashes:
+
Some commands can also use options which are given with dashes:
 
<pre>
 
<pre>
 
svn command arguments --option optionvalue
 
svn command arguments --option optionvalue
Line 31: Line 78:
 
</pre>
 
</pre>
  
=Getting the content for this practical=
+
=Checkout a Working Copy=
You have used the subversion client for all the practicals in the [[:category:Pragmatic Programming | Pragmatic Programming]] course. To create a '''working copy''' of the repository type (first make sure you don't have a local folder called "subversion"):
+
 
 +
Now that you have access to a repository, let's create a '''working copy''' of the files in the repository.  To do this we use the '''svn checkout''' command, or '''svn co''', for short:
 +
 
 
<pre>
 
<pre>
$ svn checkout http://source.ggy.bris.ac.uk/subversion-open/subversion/trunk subversion
+
gethin@gethin-desktop:~$ svn co https://github.com/ggdagw/test ./test
A    subversion/file1
+
A    test/branches
A    subversion/file2
+
A    test/trunk
A    subversion/folder1
+
A    test/trunk/README.md
A    subversion/folder1/somefile
+
Checked out revision 1.
A    subversion/folder2
 
A    subversion/folder2/somefile
 
 
</pre>  
 
</pre>  
  
What this command does is import the content of the URL into a new folder called subversion. The letter "A" simply means that these files have been added.
+
This command gets a copy of the content at the URL and places it in a new directory called test. The letter "A" simply means that these files have been added to your working copy.  You'll also notice two subdirectories called '''trunk''' and '''branches'''.  This pattern follows an convention.  Usually, subversion repositories are organised so that the main strand of development is in the ''trunk''.  Sometimes it is useful to store variants of the trunk version (more of that later) and the ''branches'' folder exists to accommodate those.  (This is purely convention as far as subversion is concerned, however, and "trunk" and "branches" are merely two folders under the URL.)
  
The content that you saw in websvn in now in your own file space. You will also notice hidden directories called ".svn". It is '''very''' important that you do not touch these directories.
+
The content that you saw through your browser is now in your own file space. You may also notice hidden directories called ".svn". '''It is very important that you do not touch these directories.'''
  
=Modifying the working copy=
 
The working copy is your to work in so go ahead and modify some things. To simplify the practical, only modify the file named after you to start with (in the wiki, file2 was modified). To see what files you have modified, you ask the client for the status of your working copy:
 
 
<pre>
 
<pre>
$ svn status
+
gethin@gethin-desktop:~$ cd test/
M      file2
+
gethin@gethin-desktop:~/test$ ls -al
 +
total 28
 +
drwxr-xr-x  5 gethin gethin  4096 2013-07-26 12:29 .
 +
drwxr-xr-x 117 gethin gethin 12288 2013-07-26 12:29 ..
 +
drwxr-xr-x  3 gethin gethin  4096 2013-07-26 12:29 branches
 +
drwxr-xr-x  6 gethin gethin  4096 2013-07-26 12:29 .svn
 +
drwxr-xr-x  3 gethin gethin  4096 2013-07-26 12:29 trunk
 
</pre>
 
</pre>
  
The status shows the letter "M" for file2, indicating it has been modified.
+
=Modifying your Working Copy=
  
Note that this status only shows the things that have changed in '''your''' working copy. Not the changes made by others of on the server.
+
Right oh.  The working copy is yours to work with so let's go ahead and modify the README.md file.
 
 
You can also add a new file. Name the new file after your name and add a suffix. "newfile" is used in the wiki:
 
  
 
<pre>
 
<pre>
$ touch newfile
+
gethin@gethin-desktop:~/test$ cd trunk
$ svn status
+
gethin@gethin-desktop:~/test/trunk$ ls
?      newfile
+
README.md
 +
gethin@gethin-desktop:~/test/trunk$ emacs -nw README.md
 
</pre>
 
</pre>
  
The interogation mark shows that the subversion client does not know what to do with the new file. By default, svn will ignore new files. To indicate that a new file should be versionned, use the add command:
+
(See, e.g. https://www.acrc.bris.ac.uk/acrc/pdf/emacs.pdf, if you'd like to use the emacs text editor, but are new to it.)
 +
 
 +
To see what files you have modified, you ask the client for the status of your working copy:
 
<pre>
 
<pre>
$ svn add newfile
+
gethin@gethin-desktop:~/test/trunk$ svn status
A        newfile
+
M       README.md
$ svn status
 
M     file2
 
A      newfile
 
 
</pre>
 
</pre>
  
Note that the letter "A" is used showing an addition.
+
The status shows the letter "M" for README.md, indicating it has been modified.
 +
 
 +
Note that this status only shows the things that have changed in '''your''' working copy.  It does not show any changes made by others, either in the repository or in their own working copies.
 +
 
 +
You can also add a new file.  Let's add a file called '''foo.txt''':
  
The same things happen for deletions. Let's try to delete file1 for instance:
 
 
<pre>
 
<pre>
$ rm -f file1
+
gethin@gethin-desktop:~/test/trunk$ touch foo.txt
$ svn status
+
gethin@gethin-desktop:~/test/trunk$ svn status
!      file1
+
?      foo.txt
M      file2
+
M       README.md
A      newfile
 
$ svn delete file1
 
D        file1
 
$ svn status
 
D      file1
 
M     file2
 
 
</pre>
 
</pre>
  
The letter "D" is used for deletions.
+
The question mark shows that the subversion client knows nothing about the new file (i.e. it is not currently under the auspices to version control). By default, svn will ignore new files. To indicate that a new file should be versioned, use the '''add''' command:
 
 
Subversion allows you to revert changes when you have made an error. Let's assume that file1 was deleted by error, you can get it back with:
 
  
 
<pre>
 
<pre>
$ svn revert file1
+
gethin@gethin-desktop:~/test/trunk$ svn add foo.txt
Reverted 'file1'
+
A        foo.txt
$ svn status
+
gethin@gethin-desktop:~/test/trunk$ svn status
M     file2
+
A      foo.txt
A      newfile
+
M       README.md
 
</pre>
 
</pre>
  
The next step is to send all these changes to the server. To avoid conflicts at this stage, make sure that you have only modified files containing your name (although this wiki page uses file2 and newfile). If you have modified anything else, revert the relevant changes.
+
The letter "A" is used to indicate an addition.
  
=Sending changes to the repository=
+
=Recording Changes in the Repository=
  
Sending changes to the repository is called a '''commit'''. The syntax to commit your changes is:
+
Sending changes to the repository is called a '''commit'''. Here's the command I used to send our two local modifications:
  
 
<pre>
 
<pre>
$ svn commit [--message "Log message."] [--username subversionlogin]
+
gethin@gethin-desktop:~/test/trunk$ svn commit --message "Added text to README.md and added the empty file foo.txt"
Sending        file2
+
Sending        trunk/README.md
Adding        newfile
+
Adding        trunk/foo.txt
 
Transmitting file data ..
 
Transmitting file data ..
 
Committed revision 2.
 
Committed revision 2.
 
</pre>
 
</pre>
  
You might have noticed the revision number. You should all get different revision numbers are your commits will be done one after the other. Because you have all worked on separate files, there is no risk of conflicts and subversion will happily combine all your changes together. Have a look in websvn (refresh the page) to see the state of the repository.
+
where the '''--message''', or '''-m''' for short, allows us to write a log message inline.
  
Sometimes, you want a long message to go with a commit. To do this, simply execute the commit without the --message option. A text editor will then popup to be used to write the commit and by saving-exiting, the commit will be done. Note that svn use the editor stored in the EDITOR environment variable and often defaults to vi if this variable is undefined. So if you are an emacs fan, set the variable first:
+
Notice the revision number.  These numbers encode the state of the whole repository at a given juncture and are the passport to retrieving earlier versions of your project.  As you commit future changes to your repository, your revision numbers will steadily increase.   
 +
 
 +
Sometimes, you want a long message to go with a commit. To do this, simply execute the commit without the --message option. A text editor will then pop-up to be used to write the message and by saving and exiting, the commit will be done. Note that svn uses the editor indicated by the '''EDITOR''' environment variable.  The editor often defaults to vi if this variable is undefined. If you are an emacs fan, set the variable first:
  
 
<pre>
 
<pre>
Line 125: Line 171:
 
</pre>
 
</pre>
  
At this stage in the practical you and others have sent changes to the server. To bring these changes back to your working copy, you need to update it.
+
(Note that you can use ''':q!''' to get out of vi, if you started it by accident.  You could also set EDITOR=nano or gedit etc.  You can also use the SVN_EDITOR environment variable.)
 +
 
 +
=Revert: Your "Get-Out-of-Jail Card"=
 +
 
 +
Just as we can add files, we can '''delete''', for example:
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test/trunk$ svn delete foo.txt
 +
D        foo.txt
 +
gethin@gethin-desktop:~/test/trunk$ ls
 +
README.md
 +
</pre>
 +
 
 +
The letter "D" indicates deletions and we see from typing 'ls' that the file has gone.
 +
 
 +
Subversion allows you to '''revert''' changes when you have made an error.  Let's assume that 'foo.txt' was deleted in error.  Fear not, you can get it back with:
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test/trunk$ svn revert foo.txt
 +
Reverted 'foo.txt'
 +
gethin@gethin-desktop:~/test/trunk$ ls
 +
foo.txt  README.md
 +
gethin@gethin-desktop:~/test/trunk$ svn status
 +
gethin@gethin-desktop:~/test2/trunk$
 +
</pre>
 +
 
 +
and foo.txt is back!  A silent return from svn status, svn stat for short, indicates that there are no pending modifications in your working copy.  Put another way, it exactly matches the repository version of the project when you made the checkout.
  
=Updating your working copy=
+
=Updating your Working Copy=
 +
 
 +
You can '''update''' your working copy to synchronise it with the latest version (known as the HEAD) held in the repository.  The general form of the update command is:
  
To update your working copy, simply run:
 
 
<pre>
 
<pre>
$ svn commit [--message "Log message."] [--username subversionlogin]
+
$ svn update
 
... <- list of files that have been added/modified
 
... <- list of files that have been added/modified
At revision 8.
+
At revision X.
 +
</pre>
 +
 
 +
If I update my working copy now:
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test/trunk$ svn update
 +
At revision 2.
 
</pre>
 
</pre>
  
Again, because you have all worked on separate files, there is no risk of conflicts and subversion should fetch all the changes without problem.
+
we see an empty list of files--i.e. there is nothing to update and my working copy perfectly matches the HEAD of the repository.
Your working copy contains now all the bits and pieces chenged by other people. A lot neater that emailing batches files etc, no?
+
 
 +
That needn't be the case, however.  Let's imagine that you and a collaborator in Japan have access to the repository.  You obviously work independently and, for good measure, in different time zones.  Your collaborator may have committed some changes to the repository since you were last in front of a computer. That being the case, an update will bring all of her changes to your working copy.
 +
 
 +
A similar situation can arise if you are simultaneously operating two checkouts.  Perhaps one at work and another on your home computer.  If you had done some work at home yesterday evening and committed the fruits of your labours, and update will bring your work copy in line.
 +
 
 +
You can even update your working copy if you have some local modifications pending. In that situation, SVN will attempt to merge your changes with those from the southern hemisphere.  If you both have edited the same line in a file, a '''conflict''' is flagged.  More on that possibility later.
  
status, commit and update will probably be your most widely used commands:
+
With all the foregoing in mind, '''status''', '''commit''' and '''update''' will probably be your most widely used commands:
* update regularly to bring other people's work
+
# '''update''' regularly to bring other people's work
* status to make sure all is well
+
# '''status''' to make sure all is well
* commit
+
# '''commit''' frequently so that you can always recover a version you care about
  
However the subversion client can also be used to looks at changes in the repository and this can boost productivity.
+
=Investigating Changes=
  
=Investigating changes=
+
This section highlights some commands can be used to boost productivity.
This section shows some subversion commands. Some of these commands can also be done directly from the websvn interface.
 
  
 
==log==
 
==log==
 
To get a log of what happened in the repository, use the log command. To see the files that have been modified as well as the log messages, use the --verbose option:
 
To get a log of what happened in the repository, use the log command. To see the files that have been modified as well as the log messages, use the --verbose option:
 
<pre>
 
<pre>
$ svn log --verbose
+
gethin@gethin-desktop:~/test/trunk$ svn log --verbose
 
------------------------------------------------------------------------
 
------------------------------------------------------------------------
r2 | jprenaud | 2008-05-14 15:18:44 +0100 (Wed, 14 May 2008) | 1 line
+
r2 | ggdagw | 2013-07-26 14:58:08 +0100 (Fri, 26 Jul 2013) | 2 lines
 
Changed paths:
 
Changed paths:
   M /trunk/file2
+
   M /trunk/README.md
   A /trunk/newfile
+
   A /trunk/foo.txt
 +
 
 +
Added text to README.md and added the empty file foo.txt
  
First commit
 
 
------------------------------------------------------------------------
 
------------------------------------------------------------------------
r1 | jprenaud | 2008-05-14 13:42:43 +0100 (Wed, 14 May 2008) | 2 lines
+
r1 | ggdagw | 2013-07-26 12:28:32 +0100 (Fri, 26 Jul 2013) | 2 lines
 
Changed paths:
 
Changed paths:
 +
  A /branches
 
   A /trunk
 
   A /trunk
   A /trunk/file1
+
   A /trunk/README.md
  A /trunk/file2
 
  A /trunk/folder1
 
  A /trunk/folder1/somefile
 
  A /trunk/folder2
 
  A /trunk/folder2/somefile
 
  
Initial import into the repository.
+
Initial commit
 +
 
 +
------------------------------------------------------------------------
 
</pre>
 
</pre>
  
Line 182: Line 265:
  
 
==diff==
 
==diff==
After you have modified something, it's sometimes nice to see what you've actually done. This is done with the diff command. For instance add some text to file1 and use diff to see what you have done.
+
After you have modified something, it can be handy to highlight what you've done. You can do this using the '''diff''' command.
 +
 
 +
For instance add some text to 'README.md' and use '''diff''' to see what you have done.
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test/trunk$ svn stat
 +
M      README.md
 +
gethin@gethin-desktop:~/test/trunk$ svn diff README.md
 +
Index: README.md
 +
===================================================================
 +
--- README.md (revision 2)
 +
+++ README.md (working copy)
 +
@@ -7,3 +7,16 @@
 +
------------
 +
 +
This file is formatted in [Markdown](https://github.com/adam-p/markdown-here/wiki/Markdown-Cheatsheet) and will be automatically rendered on your GitHub webpage.
 +
+
 +
+Here is an itemised list:
 +
+* bread
 +
+* butter
 +
+* marmalade
 +
+
 +
+A Table:
 +
+
 +
+| Name    | Colour        | Price        |
 +
+| ------- |:-------------:|--------------:|
 +
+| Thomas  | centered      | right-aligned |
 +
+| Gordon  | blue          |        £3.56 |
 +
+| Henry  |        green | £2.81        |
 +
</pre>
 +
 
 +
You can also use diff to highlight differences between two versions of some file, as stored in the repository:
 +
 
 
<pre>
 
<pre>
$ echo "Added this line." >> file1
+
$ svn diff -r73:74 foo.txt
$ svn diff file1
+
...
Index: file1
+
</pre>
==blame==
 
  
 +
==blame (praise)==
 +
Sometimes, you want to know who wrote a particular bit of code. Subversion makes that easy with the '''blame''' command:
 
<pre>
 
<pre>
 
$ svn blame file2
 
$ svn blame file2
Line 196: Line 312:
 
</pre>
 
</pre>
  
==blame==
+
You see the content of file2 and for each line the name of the author and the revision number. You could then fetch the log message for that particular revision to get more information.
 +
 
 +
<pre>
 +
$ svn log file2 --revision 3
 +
------------------------------------------------------------------------
 +
r3 | jprenaud | 2008-05-14 16:03:45 +0100 (Wed, 14 May 2008) | 1 line
 +
 
 +
More things.
 +
------------------------------------------------------------------------
 +
</pre>
 +
 
 +
However, this feature is not available for github hosted repositories:
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test/trunk$ svn blame README.md
 +
svn: Server does not support custom revprops via log
 +
</pre>
 +
 
 +
=Conflicts=
 +
 
 +
Sometimes, a commit or an update will fail because of conflicting changes. As a rule, you should always update before a commit so the example here will show a conflict created after an update.
 +
 
 +
==Creating the conflict==
 +
 
 +
As mentioned previously, conflicts arise when SVN cannot merge together changes to the same file--i.e. the changes are on the same line.
 +
 
 +
You can manufacture a conflict using two checkouts--let's call them A and B.  I could create two such working copies by typing the following:
 +
 
 +
<pre>
 +
svn co https://github.com/ggdagw/test ./testA
 +
svn co https://github.com/ggdagw/test ./testB
 +
</pre>
 +
 
 +
There is nothing to stop me having multiple checkouts on the same computer.  Now, that we have the raw materials:
 +
 
 +
# Ensure that both A and B are up-to-date.
 +
# edit line 1 of README.md in A and commit.
 +
# edit line 1 of README.md in B--do not commit.
 +
# Now attempt to update B.
 +
 
 +
Et voila, you will have a conflict.  Since SVN cannot resolve it, we must apply the old human grey matter to the task.  If you were working with a collaborator, this may well involve a phone or email conversation to decide on the best course of action.
 +
 
 +
The update does not immediately fail.  Rather, you are be presented with some options:
  
 
<pre>
 
<pre>
$ svn blame file2
+
Conflict discovered in 'README.md'.
     2  jprenaud Added some stuff
+
Select: (p) postpone, (df) diff-full, (e) edit,
 +
        (mc) mine-conflict, (tc) theirs-conflict,
 +
        (s) show all options: df
 +
</pre>
 +
 
 +
If you choose "df", then you will be presented with a summary of the 3-way difference: First how it was prior to your local change; second how it is in your working copy and lastly how it currently is in the repository.  You will also be presented with the list of options again.  If you choose mine-conflict, "mc", your local modifications will be preferred--at least in this working copy, since nothing has been committed back at this stage.  Theirs-conflict, "tc", will prefer the repository version.  If you elect to postpone, "p", then 'README.md' is flagged with the letter "C" indicating a conflict and you will notice new files in your working copy:
 +
<pre>
 +
$svn status
 +
?      README.md.r5
 +
?      README.md.r6
 +
?      README.md.mine
 +
C     README.md
 
</pre>
 
</pre>
  
=Resolving conflicts=
+
* README.md.r5 is README.md as at revision 5 (i.e. the one at your last update)
 +
* README.md.r6 is README.md at revision 6 (i.e. the one that is on the repository now)
 +
* README.md.mine is README.md as it was in working copy before the update
 +
* README.md contains an attempt at merging the changes (this will be similar to what you see with "df").
 +
 
 +
The last option, edit (e), will present to you the attempted merge in a text editor, for you to resolve as you see fit.  Note that if you type "svn status" after editing README.md in this situation, you will see that the file is still marked as conflicted and you will not be able to commit your changes until you have '''resolved''' the conflict by typing e.g.:
 +
 
 +
<pre>
 +
svn resolved README.md
 +
</pre>
  
 
=Other useful commands=
 
=Other useful commands=
==move==
+
 
Rename
+
==info==
 +
 
 +
From inside a working copy, e.g.:
 +
 
 +
<pre>
 +
svn info
 +
</pre>
 +
 
 +
will give:
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test/trunk$ svn info
 +
Path: .
 +
URL: https://github.com/ggdagw/test/trunk
 +
Repository Root: https://github.com/ggdagw/test
 +
Repository UUID: be566c2d-dc09-ebaf-f5e5-ce57b7db7bff
 +
Revision: 3
 +
Node Kind: directory
 +
Schedule: normal
 +
Last Changed Author: ggdagw
 +
Last Changed Rev: 3
 +
Last Changed Date: 2013-07-26 16:03:50 +0100 (Fri, 26 Jul 2013)
 +
</pre>
 +
 
 +
==list==
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test2/trunk$ svn list https://github.com/ggdagw/test
 +
</pre>
 +
 
 +
will give:
 +
 
 +
<pre>
 +
branches/
 +
trunk/
 +
</pre>
 +
 
 +
and
 +
 
 +
<pre>
 +
gethin@gethin-desktop:~/test2/trunk$ svn list https://github.com/ggdagw/test/trunk
 +
</pre>
 +
 
 +
will give:
 +
 
 
<pre>
 
<pre>
$ svn mv file2 new_file2
+
README.md
 +
foo.txt
 +
</pre>
 +
 
 +
==move & copy==
 +
If you rename a file or directory manually, you loose its history, this is because subversion needs to be notified that a tracked file or directory will have a new name. It is simpler to use the subversion '''move''' command. For instance, to rename "file2", do:
 +
<pre>
 +
$ svn move file2 new_file2
 
A        new_file2
 
A        new_file2
 
D        file2
 
D        file2
 
</pre>
 
</pre>
 +
 +
You notice that the new file is added and the old one deleted. You could have done this manually but the advantage of this is that the history of the new file before the new name is still available.
 +
 +
A close relation to '''move''' is '''copy'''.  This creates a new file, with a copy of the revision history of it's template:
 +
 +
<pre>
 +
$ svn copy havana havana2
 +
A        havana2
 +
</pre>
 +
 
==import==
 
==import==
==mkdir==
+
When you ask for a new repository, it is empty by default. To populate it, you can use the import command. (import is because the action is done from the server, it imports something). The syntax is:
 +
 
 +
<pre>
 +
$ svn import PATH URL/trunk --message "Log message."
 +
</pre>
 +
 
 +
* PATH is the path to the local folder (by default it ses "./", i.e. the current folder)
 +
* URL is the full URL of the repository. In the example, I also added "trunk/" at the end and the trunk would be created automatically.
 +
 
 +
==mkdir ==
 +
Often people ask how then can create the "branches/" directory in the repository to store some specific versions of their code. This can be done by invoking mkdir directly on the server. The syntax is:
 +
 
 +
<pre>
 +
$ svn mkdir URL/branches --message "Log message."
 +
</pre>
 +
 
 
==export==
 
==export==
==merge==
+
Sometimes, you want to get the files from the version control system but this will not be used as a working copy, for instance you are going to send the files to somebody who is not involved in the development. For instance, it is the command that was used for the [[Linux1]] and [[Linux2]] practicals.
 +
 
 +
You could do a checkout and remove all the hidden ".ssh" directories manually, but the easiest to to use the "export" command. It works exactly like a checkout except that you end up with a normal local folder, not a working copy. The syntax is:
 +
 
 +
<pre>
 +
$ svn export URL PATH
 +
</pre>
 +
 
 +
==branching and merge==
 +
 
 +
Subversion can support multiple development strands via the creation of branches.  The svn '''copy''' command is used to make a (space efficient) copy of an entire file tree from, say, the trunk to a subdir called "branches".  A popular reason for creating a branch is for a particular developer (or team) to work on something speculative or disruptive:
 +
 
 +
<pre>
 +
svn copy https://svn.ggy.bris.ac.uk/subversion/ourproject/trunk \
 +
https://svn.ggy.bris.ac.uk/subversion/ourproject/branches/sally_dev \
 +
-m "The reason why I'm branching is..."
 +
</pre>
 +
 
 +
When you have different branches in your project, you might want to merge the changes from one branch to another. For instance somebody has fixed a bug on a branch that is still present in the trunk. You might want to apply the changes done on branch back onto the trunk. Subversion allows you to do this and it is called a merge operation. We will cover this topic very quickly here but you can refer to the [http://svnbook.red-bean.com Subversion Red Book] for more information about merging.
 +
 
 +
[[Image:Svn_merge_general.jpg|frame|centre|General situation for svn merge]]
 +
 
 +
For example, the following command will merge in a change set from revA to revB from the sally_dev branch of ourproject into your working copy:
 +
 
 +
<pre>
 +
svn merge -r revA:revB https://svn.ggy.bris.ac.uk/subversion/ourproject/branches/sally_dev
 +
</pre>
 +
 
 +
You can then commit these changes--should you so desire--which will end up in the development line of whatever you chose to checkout in order to obtain your working copy.
  
 
=To go further=
 
=To go further=
Subversion red Book
+
The [http://svnbook.red-bean.com/ Subversion Red Book] is the bible of subversion. Highly recommended.
=Pragmatic Programming=
+
 
That's all folks. We hope you have the right basis to be
+
In the book, you can see how to create your own repository, should you desire.  For example, some simple repository setup commands will provide you with a working facility via the filesystem (i.e. the repository is on the same computer that you typically work on), or SSH (i.e. you have SSH access to the machine that will host the respository).  See below.  '''However, note two things:'''
 +
* '''It is dangerous to create a repository on a files-ystem that is not backed-up.  You could lose all your work that way.'''
 +
* '''You can request a (backed-up) University of Bristol repository simply by contacting the service desk.'''
 +
 
 +
==Creating our own Repository==
 +
 
 +
<pre>
 +
svnadmin create $HOME/my_test_repo
 +
</pre>
 +
 
 +
Next you might import some files (that you have stored in a directory called 'projectX').  If you're working on the same filesystem you would use:
 +
 
 +
<pre>
 +
cd projectX
 +
svn import . file://$HOME/my_test_repo/trunk -m "my import message"
 +
svn list file://$HOME/my_test_repo/trunk
 +
...
 +
</pre>
 +
 
 +
Which--via SSH--would translate to:
 +
 
 +
<pre>
 +
cd projectX
 +
svn import . svn+ssh://user@host/absolute/path/to/my_test_repo/trunk -m "my import message"
 +
svn list svn+ssh://user@host/absolute/path/to/my_test_repo/trunk
 +
...
 +
</pre>
 +
 
 +
NB where the components of '''user@host/absolute/path/to''' are:
 +
* user: your username on the remote machine hosting your repository
 +
* host: the hostname of the remote machine
 +
* /absolute/path/to: you must specify the absolute path to your repository ($HOME will be evaluated on the machine to SSH'ing ''from''
 +
 
 +
And you're away!

Latest revision as of 11:24, 7 November 2014


Introduction

In this workshop, we'll look at using a particular Version Control System (VCS) called Subversion (often abbreviated to SVN). Before getting into the nitty-gritty of using SVN, we'll pause to consider the motivations for adopting version control and also the key concepts that are common to most available systems.

Why is Version Control useful?

OK, here's the sales pitch:

  • It removes confusion about versions. For example, you will no longer have to keep inventing names for different versions of essentially the same document e.g. blah.old, blah.sav, blah.older, blah.newest2 (look familiar?).
  • It makes collaborative working easier. Version control assists in coordination as it removes any confusion about versions, highlights conflicts, allows the use of independent working copies, records log messages and much more besides.
  • It makes distributing your code easier. A version control repository can be visible to the world (often as a URL). However, using some highly customisable access controls, you can arrange for some (perhaps anyone) to download your project while also specifying that only a select few may be trusted to upload.
  • It makes reproducing experiments easier. The ability to reproduce an experiment is a key characteristic of science. However, all too often, in the digital age, people are unable to run the same version of a model that they ran six months ago. With version control, you can always access any previous version of your model.
  • It aids disaster recovery. You computer is fried? No problem! Just checkout your code to another and you're working productively again in minutes.

Version Control Concepts

A picture can be worth a thousand words, so let's try illustrating some of the key version control concepts, before wading into acres of text:

Files stored in a repository; a checkout; modification and commit. All versions recorded.
A checkout adds a copy of the files held in the repository to your local computer.


Subversion is a centralised version control system. Centralised version control means that a copy of your project is held in a central location called the repository and the subversion server logs all operations happening on the repository: every time something is changed in the repository, the server logs the time and date, the changes, the author as well as a log message. The server can be configured to give privavcy; allowing some people actions which are disallowed for others. For instance, in the practical, the server allows anonymous read-only access but only a selected number of people can changes things.

All the operations described above (logging and authentication) happen on the server. However, the server is only accessible directly to system administrators. To interact with the server, a user makes use of a subversion client. Some of you might already know about some graphical subversion clients such as TortoiseSVN (see the screen grab below). This practical will show how the command line client can be used. The subversion client can be used to (1) ask information from and (2) send information to the server. The client can also be used to get information about your working copy which is the local copy of the project that resides on your filespace. You can use the client to ask questions such as:

  • which files have I modified since I last synchronised with the server?
  • when was that file last modified?
  • who, inadvertently, created a bug at line 18 in file foo.c?
  • what has changed in that file?

Acquiring a Repository

For the purposes of this practical, you can get yourself a repository from one of the hosting sites that can be found out in the cloud. Or you could use another repository that you have access to--perhaps hosted here in the University, some other portion of UK academia or elsewhere. I've prepared the examples using a repository accessed via the GitHub website (https://github.com). Note, however, that to obtain a free repository on GitHub, you must agree to it being readable by anyone.

NB With that in mind, you may want to have a think about the right home for any of your intellectual property.

OK, let's assume that you are happy to work with a GitHub hosted repository--at least for your initial steps learning about version control and subversion. (A natty feature of GitHub repositories is that they can be used with both Subversion and Git VCS.)

Registering and creating a repository is easy, just follow the instructions on the webpage:

The Github web interface

Be sure to check the box: Initialize this repository with a README

In addition to the command line client that we will describe in the following sections, you can manage your repository, view and even edit your files through the GitHub website:

A test repository

svn: The Subversion Command Line Client

The subversion command line client is called svn. To execute a subversion command, simply type:

$ svn command arguments

Some commands can also use options which are given with dashes:

svn command arguments --option optionvalue

Subversion provides extensive help about the commands to use. To get help for a particular subversion command, simply use:

$ svn help command

Checkout a Working Copy

Now that you have access to a repository, let's create a working copy of the files in the repository. To do this we use the svn checkout command, or svn co, for short:

gethin@gethin-desktop:~$ svn co https://github.com/ggdagw/test ./test
A    test/branches
A    test/trunk
A    test/trunk/README.md
Checked out revision 1.

This command gets a copy of the content at the URL and places it in a new directory called test. The letter "A" simply means that these files have been added to your working copy. You'll also notice two subdirectories called trunk and branches. This pattern follows an convention. Usually, subversion repositories are organised so that the main strand of development is in the trunk. Sometimes it is useful to store variants of the trunk version (more of that later) and the branches folder exists to accommodate those. (This is purely convention as far as subversion is concerned, however, and "trunk" and "branches" are merely two folders under the URL.)

The content that you saw through your browser is now in your own file space. You may also notice hidden directories called ".svn". It is very important that you do not touch these directories.

gethin@gethin-desktop:~$ cd test/
gethin@gethin-desktop:~/test$ ls -al
total 28
drwxr-xr-x   5 gethin gethin  4096 2013-07-26 12:29 .
drwxr-xr-x 117 gethin gethin 12288 2013-07-26 12:29 ..
drwxr-xr-x   3 gethin gethin  4096 2013-07-26 12:29 branches
drwxr-xr-x   6 gethin gethin  4096 2013-07-26 12:29 .svn
drwxr-xr-x   3 gethin gethin  4096 2013-07-26 12:29 trunk

Modifying your Working Copy

Right oh. The working copy is yours to work with so let's go ahead and modify the README.md file.

gethin@gethin-desktop:~/test$ cd trunk
gethin@gethin-desktop:~/test/trunk$ ls
README.md
gethin@gethin-desktop:~/test/trunk$ emacs -nw README.md 

(See, e.g. https://www.acrc.bris.ac.uk/acrc/pdf/emacs.pdf, if you'd like to use the emacs text editor, but are new to it.)

To see what files you have modified, you ask the client for the status of your working copy:

gethin@gethin-desktop:~/test/trunk$ svn status
M       README.md

The status shows the letter "M" for README.md, indicating it has been modified.

Note that this status only shows the things that have changed in your working copy. It does not show any changes made by others, either in the repository or in their own working copies.

You can also add a new file. Let's add a file called foo.txt:

gethin@gethin-desktop:~/test/trunk$ touch foo.txt
gethin@gethin-desktop:~/test/trunk$ svn status
?       foo.txt
M       README.md

The question mark shows that the subversion client knows nothing about the new file (i.e. it is not currently under the auspices to version control). By default, svn will ignore new files. To indicate that a new file should be versioned, use the add command:

gethin@gethin-desktop:~/test/trunk$ svn add foo.txt
A         foo.txt
gethin@gethin-desktop:~/test/trunk$ svn status
A       foo.txt
M       README.md

The letter "A" is used to indicate an addition.

Recording Changes in the Repository

Sending changes to the repository is called a commit. Here's the command I used to send our two local modifications:

gethin@gethin-desktop:~/test/trunk$ svn commit --message "Added text to README.md and added the empty file foo.txt"
Sending        trunk/README.md
Adding         trunk/foo.txt
Transmitting file data ..
Committed revision 2.

where the --message, or -m for short, allows us to write a log message inline.

Notice the revision number. These numbers encode the state of the whole repository at a given juncture and are the passport to retrieving earlier versions of your project. As you commit future changes to your repository, your revision numbers will steadily increase.

Sometimes, you want a long message to go with a commit. To do this, simply execute the commit without the --message option. A text editor will then pop-up to be used to write the message and by saving and exiting, the commit will be done. Note that svn uses the editor indicated by the EDITOR environment variable. The editor often defaults to vi if this variable is undefined. If you are an emacs fan, set the variable first:

$ export EDITOR=emacs

(Note that you can use :q! to get out of vi, if you started it by accident. You could also set EDITOR=nano or gedit etc. You can also use the SVN_EDITOR environment variable.)

Revert: Your "Get-Out-of-Jail Card"

Just as we can add files, we can delete, for example:

gethin@gethin-desktop:~/test/trunk$ svn delete foo.txt
D         foo.txt
gethin@gethin-desktop:~/test/trunk$ ls
README.md

The letter "D" indicates deletions and we see from typing 'ls' that the file has gone.

Subversion allows you to revert changes when you have made an error. Let's assume that 'foo.txt' was deleted in error. Fear not, you can get it back with:

gethin@gethin-desktop:~/test/trunk$ svn revert foo.txt
Reverted 'foo.txt'
gethin@gethin-desktop:~/test/trunk$ ls
foo.txt  README.md
gethin@gethin-desktop:~/test/trunk$ svn status
gethin@gethin-desktop:~/test2/trunk$

and foo.txt is back! A silent return from svn status, svn stat for short, indicates that there are no pending modifications in your working copy. Put another way, it exactly matches the repository version of the project when you made the checkout.

Updating your Working Copy

You can update your working copy to synchronise it with the latest version (known as the HEAD) held in the repository. The general form of the update command is:

$ svn update
... <- list of files that have been added/modified
At revision X.

If I update my working copy now:

gethin@gethin-desktop:~/test/trunk$ svn update
At revision 2.

we see an empty list of files--i.e. there is nothing to update and my working copy perfectly matches the HEAD of the repository.

That needn't be the case, however. Let's imagine that you and a collaborator in Japan have access to the repository. You obviously work independently and, for good measure, in different time zones. Your collaborator may have committed some changes to the repository since you were last in front of a computer. That being the case, an update will bring all of her changes to your working copy.

A similar situation can arise if you are simultaneously operating two checkouts. Perhaps one at work and another on your home computer. If you had done some work at home yesterday evening and committed the fruits of your labours, and update will bring your work copy in line.

You can even update your working copy if you have some local modifications pending. In that situation, SVN will attempt to merge your changes with those from the southern hemisphere. If you both have edited the same line in a file, a conflict is flagged. More on that possibility later.

With all the foregoing in mind, status, commit and update will probably be your most widely used commands:

  1. update regularly to bring other people's work
  2. status to make sure all is well
  3. commit frequently so that you can always recover a version you care about

Investigating Changes

This section highlights some commands can be used to boost productivity.

log

To get a log of what happened in the repository, use the log command. To see the files that have been modified as well as the log messages, use the --verbose option:

gethin@gethin-desktop:~/test/trunk$ svn log --verbose
------------------------------------------------------------------------
r2 | ggdagw | 2013-07-26 14:58:08 +0100 (Fri, 26 Jul 2013) | 2 lines
Changed paths:
   M /trunk/README.md
   A /trunk/foo.txt

Added text to README.md and added the empty file foo.txt

------------------------------------------------------------------------
r1 | ggdagw | 2013-07-26 12:28:32 +0100 (Fri, 26 Jul 2013) | 2 lines
Changed paths:
   A /branches
   A /trunk
   A /trunk/README.md

Initial commit

------------------------------------------------------------------------

You can also invoke the log command on a particular file/path and provide a range of revisions. For instance to see which commits affected file1 between revisions 4 and 6, one could use:

$ svn log --verbose --revision 4:6 file1
... <- log output

diff

After you have modified something, it can be handy to highlight what you've done. You can do this using the diff command.

For instance add some text to 'README.md' and use diff to see what you have done.

gethin@gethin-desktop:~/test/trunk$ svn stat
M       README.md
gethin@gethin-desktop:~/test/trunk$ svn diff README.md 
Index: README.md
===================================================================
--- README.md	(revision 2)
+++ README.md	(working copy)
@@ -7,3 +7,16 @@
 ------------
 
 This file is formatted in [Markdown](https://github.com/adam-p/markdown-here/wiki/Markdown-Cheatsheet) and will be automatically rendered on your GitHub webpage.
+
+Here is an itemised list:
+* bread
+* butter
+* marmalade
+
+A Table:
+
+| Name    | Colour        | Price         |
+| ------- |:-------------:|--------------:|
+| Thomas  | centered      | right-aligned |
+| Gordon  | blue          |         £3.56 |
+| Henry   |         green | £2.81         |

You can also use diff to highlight differences between two versions of some file, as stored in the repository:

$ svn diff -r73:74 foo.txt
...

blame (praise)

Sometimes, you want to know who wrote a particular bit of code. Subversion makes that easy with the blame command:

$ svn blame file2
     2   jprenaud Added some stuff
     3   jprenaud Another line
     4   jprenaud A third line.

You see the content of file2 and for each line the name of the author and the revision number. You could then fetch the log message for that particular revision to get more information.

$ svn log file2 --revision 3
------------------------------------------------------------------------
r3 | jprenaud | 2008-05-14 16:03:45 +0100 (Wed, 14 May 2008) | 1 line

More things.
------------------------------------------------------------------------

However, this feature is not available for github hosted repositories:

gethin@gethin-desktop:~/test/trunk$ svn blame README.md 
svn: Server does not support custom revprops via log

Conflicts

Sometimes, a commit or an update will fail because of conflicting changes. As a rule, you should always update before a commit so the example here will show a conflict created after an update.

Creating the conflict

As mentioned previously, conflicts arise when SVN cannot merge together changes to the same file--i.e. the changes are on the same line.

You can manufacture a conflict using two checkouts--let's call them A and B. I could create two such working copies by typing the following:

svn co https://github.com/ggdagw/test ./testA
svn co https://github.com/ggdagw/test ./testB

There is nothing to stop me having multiple checkouts on the same computer. Now, that we have the raw materials:

  1. Ensure that both A and B are up-to-date.
  2. edit line 1 of README.md in A and commit.
  3. edit line 1 of README.md in B--do not commit.
  4. Now attempt to update B.

Et voila, you will have a conflict. Since SVN cannot resolve it, we must apply the old human grey matter to the task. If you were working with a collaborator, this may well involve a phone or email conversation to decide on the best course of action.

The update does not immediately fail. Rather, you are be presented with some options:

Conflict discovered in 'README.md'.
Select: (p) postpone, (df) diff-full, (e) edit,
        (mc) mine-conflict, (tc) theirs-conflict,
        (s) show all options: df

If you choose "df", then you will be presented with a summary of the 3-way difference: First how it was prior to your local change; second how it is in your working copy and lastly how it currently is in the repository. You will also be presented with the list of options again. If you choose mine-conflict, "mc", your local modifications will be preferred--at least in this working copy, since nothing has been committed back at this stage. Theirs-conflict, "tc", will prefer the repository version. If you elect to postpone, "p", then 'README.md' is flagged with the letter "C" indicating a conflict and you will notice new files in your working copy:

$svn status
?      README.md.r5
?      README.md.r6
?      README.md.mine
C      README.md
  • README.md.r5 is README.md as at revision 5 (i.e. the one at your last update)
  • README.md.r6 is README.md at revision 6 (i.e. the one that is on the repository now)
  • README.md.mine is README.md as it was in working copy before the update
  • README.md contains an attempt at merging the changes (this will be similar to what you see with "df").

The last option, edit (e), will present to you the attempted merge in a text editor, for you to resolve as you see fit. Note that if you type "svn status" after editing README.md in this situation, you will see that the file is still marked as conflicted and you will not be able to commit your changes until you have resolved the conflict by typing e.g.:

svn resolved README.md

Other useful commands

info

From inside a working copy, e.g.:

svn info

will give:

gethin@gethin-desktop:~/test/trunk$ svn info
Path: .
URL: https://github.com/ggdagw/test/trunk
Repository Root: https://github.com/ggdagw/test
Repository UUID: be566c2d-dc09-ebaf-f5e5-ce57b7db7bff
Revision: 3
Node Kind: directory
Schedule: normal
Last Changed Author: ggdagw
Last Changed Rev: 3
Last Changed Date: 2013-07-26 16:03:50 +0100 (Fri, 26 Jul 2013)

list

gethin@gethin-desktop:~/test2/trunk$ svn list https://github.com/ggdagw/test

will give:

branches/
trunk/

and

gethin@gethin-desktop:~/test2/trunk$ svn list https://github.com/ggdagw/test/trunk

will give:

README.md
foo.txt

move & copy

If you rename a file or directory manually, you loose its history, this is because subversion needs to be notified that a tracked file or directory will have a new name. It is simpler to use the subversion move command. For instance, to rename "file2", do:

$ svn move file2 new_file2
A         new_file2
D         file2

You notice that the new file is added and the old one deleted. You could have done this manually but the advantage of this is that the history of the new file before the new name is still available.

A close relation to move is copy. This creates a new file, with a copy of the revision history of it's template:

$ svn copy havana havana2
A         havana2

import

When you ask for a new repository, it is empty by default. To populate it, you can use the import command. (import is because the action is done from the server, it imports something). The syntax is:

$ svn import PATH URL/trunk --message "Log message."
  • PATH is the path to the local folder (by default it ses "./", i.e. the current folder)
  • URL is the full URL of the repository. In the example, I also added "trunk/" at the end and the trunk would be created automatically.

mkdir

Often people ask how then can create the "branches/" directory in the repository to store some specific versions of their code. This can be done by invoking mkdir directly on the server. The syntax is:

$ svn mkdir URL/branches --message "Log message."

export

Sometimes, you want to get the files from the version control system but this will not be used as a working copy, for instance you are going to send the files to somebody who is not involved in the development. For instance, it is the command that was used for the Linux1 and Linux2 practicals.

You could do a checkout and remove all the hidden ".ssh" directories manually, but the easiest to to use the "export" command. It works exactly like a checkout except that you end up with a normal local folder, not a working copy. The syntax is:

$ svn export URL PATH

branching and merge

Subversion can support multiple development strands via the creation of branches. The svn copy command is used to make a (space efficient) copy of an entire file tree from, say, the trunk to a subdir called "branches". A popular reason for creating a branch is for a particular developer (or team) to work on something speculative or disruptive:

svn copy https://svn.ggy.bris.ac.uk/subversion/ourproject/trunk \
https://svn.ggy.bris.ac.uk/subversion/ourproject/branches/sally_dev \
-m "The reason why I'm branching is..."

When you have different branches in your project, you might want to merge the changes from one branch to another. For instance somebody has fixed a bug on a branch that is still present in the trunk. You might want to apply the changes done on branch back onto the trunk. Subversion allows you to do this and it is called a merge operation. We will cover this topic very quickly here but you can refer to the Subversion Red Book for more information about merging.

General situation for svn merge

For example, the following command will merge in a change set from revA to revB from the sally_dev branch of ourproject into your working copy:

svn merge -r revA:revB https://svn.ggy.bris.ac.uk/subversion/ourproject/branches/sally_dev

You can then commit these changes--should you so desire--which will end up in the development line of whatever you chose to checkout in order to obtain your working copy.

To go further

The Subversion Red Book is the bible of subversion. Highly recommended.

In the book, you can see how to create your own repository, should you desire. For example, some simple repository setup commands will provide you with a working facility via the filesystem (i.e. the repository is on the same computer that you typically work on), or SSH (i.e. you have SSH access to the machine that will host the respository). See below. However, note two things:

  • It is dangerous to create a repository on a files-ystem that is not backed-up. You could lose all your work that way.
  • You can request a (backed-up) University of Bristol repository simply by contacting the service desk.

Creating our own Repository

svnadmin create $HOME/my_test_repo

Next you might import some files (that you have stored in a directory called 'projectX'). If you're working on the same filesystem you would use:

cd projectX
svn import . file://$HOME/my_test_repo/trunk -m "my import message"
svn list file://$HOME/my_test_repo/trunk
...

Which--via SSH--would translate to:

cd projectX
svn import . svn+ssh://user@host/absolute/path/to/my_test_repo/trunk -m "my import message"
svn list svn+ssh://user@host/absolute/path/to/my_test_repo/trunk
...

NB where the components of user@host/absolute/path/to are:

  • user: your username on the remote machine hosting your repository
  • host: the hostname of the remote machine
  • /absolute/path/to: you must specify the absolute path to your repository ($HOME will be evaluated on the machine to SSH'ing from

And you're away!