Best practices for Subversion and Visual Studio projects

Solution 1:

According to MSDN:

You can add the following files to Visual Studio source control:

  • Solution files (*.sln).
  • Project files, for example, *.csproj, *.vbproj files.
  • Application configuration files, based on XML, used to control run-time behavior of a Visual Studio project.

Files that you cannot add to source control include the following:

  • Solution user option files (*.suo).
  • Project user option files, for example, *.csproj.user, *.vbproj.user files.
  • Web information files, for example, *.csproj.webinfo, *.vbproj.webinfo, that control the virtual root location of a Web project.
  • Build output files, for example, *.dll and *.exe files.

Solution 2:

I would suggest using AnkhSVN - a Subversion source control plugin for Visual Studio 2008/2010.

You can use it to perform your initial add and commit of the solution, projects and sources to the repository and it won't add any of the build artefacts. It won't add anything that is generated by your build, only files that are referenced by your solution. If there are any other bits and pieces you need that aren't in your solution, you can add them afterwards.

Solution 3:

Put the following files in version control:

  • .dsw (VS6 workspace)
  • .dsp (VS6 project)
  • .sln (VS Solution)
  • .*proj (VS Project files of various types)
  • of course your source files and other artifacts you create

Do not put the following files into version control:

  • .ncb (something to do with browsing or intellsense)
  • .suo (user workspace settings like window placement, etc - I think)
  • .user (user project settings like breakpoints, etc - I think)

Also, don't put in any object files, executables, auto-generated files (like headers that might be generated).

As for executables and other generated files - there might be an exception if you want to be able to archive releases. That might be a good idea, but you'll probably want to manage that a little differently and possibly in a different place than your source code. If you do this, also archive your .pdb files so you can debug the stuff later. you might want to use a Symbol Server to store you archived symbols (see Debugging Tools for Windows for the symbol server and its documentation).

Here's my list of VS-specific files that I exclude from SVN:

Ankh.Load
*.projdata
*.pdb
*.positions
*proj.user
*proj.*.user
*.ncb
*.suo
*.plg
*.opt
*.ilk
*.pch
*.idb
*.clw
*.aps

Solution 4:

Solution level:

  • add the .sln solution file
  • ignore the .suo solution user options file

Project level:

  • add the .csproj, .vbproj (and c++ proj?) files
  • ignore the .csproj.user, .vbproj.user files
  • ignore the bin directory
  • ignore the obj directory
  • ignore any files/directories that get generated during runtime (ie. logs)

If you use and VS addins, they may generate files that also need ignoring (ie. ReSharper generates .resharper and .resharper.user files).

The ignore items can either be ignored explicitly by filename (ie. MyProject.csproj), or by a wildcard pattern (ie. *.csproj.user).


Once you have your ignores set up, checking out a clean copy of your source then building should then show no modifications (ie. no new unversioned files).