Hi, Bit of weirdness I ran into. I have a SVN tree, with a branch that gets some backported fixe (they usually apply cleanly). I commit the fix to the trunk + git-svn dcommit, and then switch to the other branch, and git-cherry-pick the fix. For git-svn commits, the log has: On the trunk side, there is a two blank line before git-svn-id. On the branch side, there are three blank lines before git-svn-id. If I cherry-pick from that branch to a third branch, I get 4 blank lines. For regular SVN commits, the log has: Only a single blank line before git-svn-id. I think the source of the problem is that git-svn-id is being removed when cherry-picking, but NOT the leading blank line. This all leads to the hashes of the commits diverging badly, and then git-cherry gets very confused when asked to compare the trunk vs. the branches. How do we solve it? This is only on a conceptual level here, looking at the present state of git-svn, I'm not sure how best to solve it. git-svn fetch: 1. Get commit message from SVN. 2. Trim ALL trailing blank lines (and existing git-svn-id lines). 3. If the last line was header-style (Signed-off-by, CC, etc) 3.1. if it was, do not add a blank line. 3.2. If not, add a single blank line. 4. Insert git-svn-id header. git-svn dcommit: 1. Get commit message (git cat-file commit ....) 2. Remove the git-svn-id line. 3. Remove all trailing blank lines. 4. Commit. -- Robin Hugh Johnson Gentoo Linux Developer & Council Member E-Mail : robbat2@xxxxxxxxxx GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85
Attachment:
pgpxQ7mGONcpQ.pgp
Description: PGP signature
