This series introduces an --object-id option to git merge-file such that, instead of reading and writing from files on the system, it reads from and writes to the object store using blobs. This is in use at GitHub to produce conflict diffs when a merge fails, and it seems generally useful, so I'm sending it here. The only tricky piece is the fact that we have to special-case the empty blob since otherwise it isn't handled correctly. brian m. carlson (1): merge-file: add an option to process object IDs Documentation/git-merge-file.txt | 20 +++++++++++ builtin/merge-file.c | 58 +++++++++++++++++++++++--------- t/t6403-merge-file.sh | 58 ++++++++++++++++++++++++++++++++ 3 files changed, 120 insertions(+), 16 deletions(-)
