MSysGit:MSysGitHerald1

OBSOLETE CONTENT

This wiki has been archived and the content is no longer updated. Please visit git-scm.com/doc for up-to-date documentation.

Good morning git land!

This lovely morning in York is as good an occasion as any to try a new format for news on msysGit, the effort to bring one of the most powerful Source Code Management systems to the poor souls stuck with Windows.

This is the 1st issue of the Herald (let's see if there are more ;-):

• WinGit 0.1

• The war between cmd and bash

• Missing make.exe, problems to clone

• msysGit breaks through 1000 downloads

• No concrete mingw/git.git merge plans yet

WinGit 0.1

Hot from the press: a first installer for native git on Windows has been released, marked as an alpha version. It is named "WinGit" -- the similarity to "wing it" is purely intentional.

This program installs git and some necessary dependencies from MSys into a directory of your choice.

Then it installs shortcuts on the Desktop, a QuickLaunch and a StartMenu item for a Git Shell.

WinGit is sibling of msysGit (a full blown installer containing the complete development environment needed to hack on Git, and a checkout of the git source code) and GitMe (a network installer version of msysGit, which clones the repository from repo.or.cz).

The war between cmd and bash

Windows does not come with a proper shell, as many of you know, but with a limited command line, called "command" in DOS based Windowses, and "cmd" in NT based ones. There are a few deviations between the two, and worse, there are deviations between cmds of different Windows versions.

Nevertheless, regular Windows users are used to that program, which is a bit more than just a shell: it is also the window in which that shell runs. This is in the good tradition of Windows not really separating different tasks cleanly, (at least sometimes) for the benefit of a better user experience.

In the early days of the MinGW port there was no question that git was run in Rxvt, the terminal emulator that comes with MSys, and from bash, which also comes from MSys.

Then, one day came along Johannes Sixt, who was not only interested in getting git to run in that POSIX hostile environment, but also to use it. Which is the perfect pole position to make something work smoothly, as opposed to just make it compile.

Since code talks, those who write the patches get to decide how the program runs, and therefore the MinGW port was soon made to run from within cmd.exe rather than bash.exe. All the goodies of bash notwithstanding, like tab completion, consistent and portable command set, cmd is the preferred environment of many Windows users.

A big discussion ensued, and is indeed still waging, if cmd or bash should be the default starting point for WinGit. Probably the best argument for cmd is that Windows people tend to know that, and would have to learn bash. Probably the best argument for bash is that it is sort of the lingua franca to get help with git.

Missing make.exe, problems to clone

While many people reported success with one or more of the 3 installers, there have been reports that make.exe is not installed, and therefore git cannot be compiled, or that a clone fails (even if the initial cloning of the msysgit repository worked fine).

These issues are not resolved, but are investigated.

msysGit breaks through 1000 downloads

Version 0.6 of msysGit, the first version of msysGit in a "long" time, has been downloaded more than 1000 times. We're getting popular.

No concrete mingw/git.git merge plans yet

msysgit uses a modified git source code base, maintained in the 4msysgit repository on http://repo.or.cz/w/git/mingw/4msysgit.git/, which is based on the MinGW port on http://repo.or.cz/w/git/mingw.git/, which itself is a fork of git.git.

There are still deviations in 4msysgit.git from mingw.git, since the setups are not quite identical (see cmd/bash war). So in msysgit, we can rely on a setup which is more or less controlled by us. We know the precise version of bash, perl or Tcl/Tk, for example. And we can just add programs as we need them.

There has been a lengthy post to the git mailing list describing what would be needed to merge the MinGW efforts back into git.git, which would be nice in the 1.5.3.* timeframe.

That post contains a long list of pretty well contained tasks. However, as of time of writing, no volunteer (except for the assiduous-as-ever Johannes Sixt) stepped up to tackle any of the tasks.