[Nut-upsdev] Repo conversion progress report
Charles Lepple
clepple at gmail.com
Fri Dec 30 04:02:15 UTC 2011
What I might need to do is move this to a box with a wider screen, and compare the raw git-svn results side-by-side with the reposurgeon output. I think I see a number of dangling branches that may have had extra commits trimmed (commits that would have anchored them to a branch).
I also saw a few commits that reference a commit with the new ID format, but the corresponding target commit doesn't seem to match. For example, "First shot of limiting the impact of 2009-10-21T23:52:59Z!clepple+nut at gmail.com to the mge-hid subdriver." The corresponding Git date, "Wed Oct 21 23:52:59 2009 +0000", does not seem to be related. I might try my hand at dumping a table of commit IDs in reposurgeon and Git format. (Or is this related to the "fossil reference map"?)
Another corner case is when there are two old-style references on the same line (usually seen in a merge commit where "from" and "to" revisions are listed). There are a number of occurrences of "[[SVN:...]]" in the git log output after the timestamp-basd commit IDs. I think this is being substituted near line 5777 of reposurgeon, but shouldn't the "while" loop get all instances in each event? I'm a little suspicious of the in-place transformation of the comments, but in theory the new commit IDs shouldn't collide with the old ones.
--
Charles Lepple
clepple at gmail
More information about the Nut-upsdev
mailing list