starseeker (Jul 17 2018 at 01:44): starseeker (Jul 17 2018 at 01:44):@Sean When I started taking apart the cvs-git repo, I noticed a rather odd behavior of the conversion in that things were committed to non-master git branches for most of the repo history. Very odd, and it complicates the analysis somewhat.
Sean (Jul 17 2018 at 01:46):huh anything in the tool docs or history that would indicate why that might happen?
starseeker (Jul 17 2018 at 02:17):Not that I spotted right offhand
Sean (Jul 17 2018 at 02:18):well if they're not going to master, what were they going to?
starseeker (Jul 17 2018 at 02:19):cvs branches, looks like (rt, rt-2, etc.)
starseeker (Jul 17 2018 at 02:21):I'm eventually planning on a variation of your original proposal - compare all the diffs (sha1 signatures actually) and try to correct the svn dates, then convert off of that history
starseeker (Jul 17 2018 at 02:24):The other alternative is to take the git fast export stream from the cvs-git conversion, post-process it to do the commits on master and turn what appear to be full git branches into tags (which is how they look in gitk, to a casual inspection) and then proceed to splice in the svn conversion. That's my "preferred" approach, but I need to get my steam back up to finish the git side of the parsing logic
Sean (Jul 17 2018 at 02:35):I don't know the downsides or upsides, so whatever you think is best. I wouldn't be too concerned whether they are branches or tags -- heck keep branches and tag the ends. or leave them as branches and we can tag/restructure later. no info is lost.
Last updated: Oct 09 2024 at 00:44 UTC