From 0b6dac2de037bb6cb00ebd81849a7adf62c2251d Mon Sep 17 00:00:00 2001 From: Joey Hess Date: Wed, 25 Oct 2017 15:04:55 -0400 Subject: response --- ...ment_1_7bc271ddeafc92d36d2b90fcb1481891._comment | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) create mode 100644 doc/bugs/graft__47__graft_cleanup_commits_--_really_needed__63__/comment_1_7bc271ddeafc92d36d2b90fcb1481891._comment diff --git a/doc/bugs/graft__47__graft_cleanup_commits_--_really_needed__63__/comment_1_7bc271ddeafc92d36d2b90fcb1481891._comment b/doc/bugs/graft__47__graft_cleanup_commits_--_really_needed__63__/comment_1_7bc271ddeafc92d36d2b90fcb1481891._comment new file mode 100644 index 000000000..8de4c1915 --- /dev/null +++ b/doc/bugs/graft__47__graft_cleanup_commits_--_really_needed__63__/comment_1_7bc271ddeafc92d36d2b90fcb1481891._comment @@ -0,0 +1,21 @@ +[[!comment format=mdwn + username="joey" + subject="""comment 1""" + date="2017-10-25T18:56:49Z" + content=""" +They're really needed. Squashing them together would make the tree ref +that's being grafted in not be reachable, so it would be subject to GC, +which would lose data that the export feature needs to make sure is +retained for later. + +It would be possible to squash the graft cleanup with the export.log update +commit. However, I was unable to stage the graft +into the git-annex branch index file using git update-index, due to +possibly a bug in git, so I was not able to find a way to do that. +See [[!commit 5483ea90eca33f61c799fb6a3c2675657caa9c75]] + +The diff makes it look somehow big and expensive, but the actual overhead +on disk is two commit objects. The tree object being grafted in is the +exported tree, so retaining it does not really add any overhead. So +something in the area of 2 kb overhead, per export of a tree. +"""]] -- cgit v1.2.3