Recently I started investigating the two major Distributed Version Control Systems (DVCS) mainly due to the historical SVN deficiency in handling renames. You may say that you don’t need a DVCS for tracking renames … Yes, in fact I know… it was only an excuse to start learning a DVCS after all there are plenty differences between a regular VCS and a DVCS.
My first option
After analysing whether I should stick with Git or Hg I decided to go with Hg since I have a trauma of using native applications originally written for Linux on Windows. Not that I am a Windows only user, in fact for a long time I had been using Linux as a Desktop option instead of Windows but you can’t deny that there is still a huge crowd that won’t switch from Windows over anything. The problem with native Linux applications that highly depend on a collection of shell scripts and other Linux dependent solutions is that they usually have a suboptimal performance on Windows, either they miss some functionality or they depend on a myriad of rare libraries. Have said that, I went with Mercurial on my first attempt.
First attempt with Hg
I wasn’t really lucky on my first attempt to install Hg. My first mistake was to pay too much attention to python.org’s warn on main downloads page:
If you don’t know which version to use, start with Python 2.7;
This warning is probably updated after each stable version is released but if I had seen the other advice on releases page I’d have thought twice:
Consider your needs carefully before using a version other than the current production version.
I chose to download latest python and build Hg myself and obviously it prove to be not that smart as it was my first experience with Hg.
Comes Git
As I gave up on Hg I decided to give a try on Git. First thing was to download msysGIT and surprisingly enough (following this tutorial) it was rather easy to set it up but its drawbacks were related to its tooling. As soon as I setup Git and tried to clone a repository over HTTPS with authentication I realized that JGIT does not support authentication over HTTP and as it was what I planned (in fact SSH on Windows is not very advisable since I have never seen a good free port of a SSH Server for Windows).
I had to get back to Hg but I decided to check whether I was taking a complex approach since Git employs a similar approach and had been much easier, I used what I learned with the tutorial used for Git setup.
Second attempt on Hg
As already mentioned, I decided to do something similar to what I done on Git, so, I chose CGI. I’ll highlight the important points for the installation here:
- The file to be downloaded is now named hgweb.cgi and not hgwebdir.cgi
- Download python 2.5 as noted here
- Unzip library.zip as noted here and edit the sys.path.insert line and the first line (the one with the #! (sha-bang) ) to point to python executable
- Configure style and templates entries under [web] on hgweb.config
- Configure an entry under [paths] for each repository (eg.: repository = c:/users/hg/repository)
- Enable pushing for the configured repositories
- Configure authorization on Apache. Either using htpasswd or ldap, but authorization is really recommended.
- Configure SSL on Apache (there is a short explanation on how to do this in portugues over here, the only thing is that SSLPassPhraseDialog builtin is not supported on Windows, so instead, provide a .bat file with a simple @echo yourpassword and use exec instead of builtin (eg.: SSLPassPhraseDialog exec:C:/Progra~1/Apache~1/Apache2.2/bin/passphrase.bat
Perform an hg init for each configured repository, start Apache and try cloning the repository over HTTPS (remember to provide your credentials if you configured any authentication method).