Profile cover photo
Profile photo
Johan Herland
103 followers -
Software engineer at Cisco. Open source enthusiast. Jazz piano junkie.
Software engineer at Cisco. Open source enthusiast. Jazz piano junkie.

103 followers
About
Posts

Post has attachment
Johan Herland commented on a post on Blogger.
I agree that the current solution is not good for git-svn users. Ideally (say, if git-svn was introduced today), git-svn should make use of the remote-helpers infrastructure in git (see: https://www.kernel.org/pub/software/scm/git/docs/git-remote-helpers.html ). This would make use of git-svn much more seamless, as instead of using separate "git svn" commands, you could now clone/fetch from a Subversion server, and push back to it with regular git commands, and the git-svn code would be a more-or-less invisible middleware that made it all Just Work(tm).

Furthermore, with remote-helpers, the git-svn remote-tracking branches would actually be valid remote-tracking branches (and not just masquerading as such without a proper git "remote" backing them up), and the tracking behavior you describe above would also Just Work(tm).

I believe there have been some attempts at rewriting git-svn as a remote-helper, but I haven't heard about it in a long while, so I don't know the current status.

Meanwhile, if the change in tracking behavior seriously hampers the usability of git-svn, I guess you could lobby for reintroducing the prefixcmp(real_ref, "refs/remotes/") line in branch.c:create_branch(), although I would consider that a regression (albeit small) for non-git-svn users.

Post has attachment
I love Peff's walk-through of how he investigated and fixed a pack corruption in a Git repo. Lots to learn about low-level Git commands, and how a pack file is structured. Also a good example of how Git's hashing and checksumming notices corruptions that other systems might not: http://article.gmane.org/gmane.comp.version-control.git/236238

Post has attachment
My #Git presentation from #ACCU2013  is finally online!

Post has attachment
Controlling my mouse pointer with 4 different devices (trackpoint, trackpad, external mouse and #LeapMotion). Experimental code: https://github.com/jherland/LeaPointer
Photo

Today I learned not to put "~/foo"-style paths in my $PATH. Although the ~/ is expanded properly by bash (and probably most other shells) it isn't expanded when calling e.g. subprocess.call() from Python. Explains why a simple command that worked on the command line refused to work from within my Python script. Using $HOME instead of ~ does wonders...

Post has attachment

Post has attachment

Post has attachment
Raspberry Pi x 3 :)
Photo

Post has attachment

Let's see how Google+ compares to Twitter and Facebook...
Wait while more posts are being loaded