On 8 February 2013 11:01, Lewis John Mcgibbney <[email protected]> wrote: > Hi All, > > I thought it best to bring this up once and for all. > > I use SVN, I make patches as follows > > svn diff > patch.patch > > which can be applied as follows > > patch -p0 -i patch.patch > > and can be reverted as follows > > patch -p0 -R < patch.patch > > Some resources for Git within the TheASF > > [0] http://wiki.apache.org/general/GitAtApache > [1] http://wiki.apache.org/hadoop/GitAndHadoop > > AFAIK, we are using SVN as the authoritative source code for Apache Any23 > (if there is appetite to change this then we can) but it is within our > interests to make using Git as easy as possible as well. We could however > do with some lightweight policy that specifies the standard method for > producing patches for the code.
I am all for using Git. The major benefit for me is simple and quick branching, which when you can do it easily, can be the standard practice for patching every issue. Most issues would have quick git branches based on the issue names, which can easily be reviewed and checked out, and updated if necessary, before merging to trunk/master. Ie, publishing a branch to git named ANY23-144 would be the standard way to provide a patch for that issue, or ANY23-144-followup if there are subsequent or alternative patches. If people want to attach patches to issues that is still available with Git, as the Hadoop guidelines show. However, pulling a branch is more convenient if we are using Git, and you can easily make fixes to a branch and push it back to a shared apache git repository if there are minor issues or if people are collaborating on a patch. > Any comments...? :0) Sorry that my system seems to be broken with regards to my failure to be able to run patch today. Will look into it. Cheers, Peter
