We've been using git-lfs successfully for several years. However, we've periodically added to the file types that we're including in git-lfs. One of the more recent file types that we've moved over to git-lfs is all of our *.csv files.
The problem is that we've got several very large *.csv files in our history that didn't get caught in the initial migration, and I need to move those files to git-lfs as well. (We're moving from BitBucket to GitHub, and GitHub has stricter rules on the maximum size of files it will allow, so I need to make sure all the old large files are on git-lfs as well.)
The obvious way to do it is something like this:
git lfs migrate import --everything --include="*.csv"
And when I run that command, it appears to succeed. It rewrites history, and the output looks something like this:
PS E:\source\swyfft_web_ken> git lfs migrate import --everything --include="*.csv"
migrate: Sorting commits: ..., done.
migrate: Rewriting commits: 100% (131623/131623), done.
development 1b042eeebfb346af8f7e90d1f6c7b6abfc64d014 -> 1b042eeebfb346af8f7e90d1f6c7
b6abfc64d014
feature/ks/20200205_ImsWebServicesV2 e2b5b267d5215ef48d5ef4e3f0924613c6076cd7 -> e2b5b267d5215ef48d5ef4e3f092
4613c6076cd7
master 2029338210697c795a536faa9dd4c141c6edd6c2 -> 2029338210697c795a536faa9dd4
c141c6edd6c2
1.0 752143ee1b87020e9b285886b3986703d6c74a38 -> 752143ee1b87020e9b285886b398
6703d6c74a38
1.1 5fefda3de45651107218a953e789ea5121002756 -> 5fefda3de45651107218a953e789
ea5121002756
1.2 30f87f08e672d10de0029985ad5acc1bf5a82457 -> 30f87f08e672d10de0029985ad5a
cc1bf5a82457
1.2.1 a4206c0ffb957401012385306157b7e1d14332bb -> a4206c0ffb957401012385306157
b7e1d14332bb
1.3 dc5d6c88a9c64150a33c60963f9ba8b03b1399de -> dc5d6c88a9c64150a33c60963f9b
a8b03b1399de
1.3.1 e3306023d55c5195ed12936a0b7cafb752d7d1f8 -> e3306023d55c5195ed12936a0b7c
(lots more like that)
But then when I try to run git lfs migrate info --everything, it doesn't appear that those files have actually been touched.
PS E:\source\swyfft_web_ken> git lfs migrate info --everything
migrate: Sorting commits: ..., done.
migrate: Examining commits: 100% (131623/131623), done.
*.cs 3.1 GB 142030/142038 files(s) 100%
*.csv 1.3 GB 2102/2102 files(s) 100%
*.csproj 1.1 GB 15634/15634 files(s) 100%
*.exe 224 MB 43/43 files(s) 100%
*.ts 150 MB 6133/6134 files(s) 100%
And when I try to push the rewritten repository to GitHub, it still complains about there being *.csv files that are too large for GitHub to accept. (Can't get the exact error message, as the error takes about 8 hours to show up, and the last time I got it has scrolled off my screen.)
I've also tried pruning and GC'ing the repo afterwards, just in case that was interefering, but it doesn't seem to have made any difference.
PS E:\source\swyfft_web_ken> git reflog expire --expire-unreachable=now --all
PS E:\source\swyfft_web_ken> git gc --prune=now
Enumerating objects: 709351, done.
Counting objects: 100% (709351/709351), done.
Delta compression using up to 4 threads
Compressing objects: 100% (210504/210504), done.
Writing objects: 100% (709351/709351), done.
Total 709351 (delta 579094), reused 583494 (delta 492112)
Removing duplicate objects: 100% (256/256), done.
Checking connectivity: 709351, done.
Expanding reachable commits in commit graph: 131646, done.
PS E:\source\swyfft_web_ken> git lfs migrate info --everything --include="*.csv"
migrate: Sorting commits: ..., done.
migrate: Examining commits: 100% (131623/131623), done.
*.csv 1.3 GB 2102/2102 files(s) 100%
Am I misunderstanding something about how git lfs migrate is supposed to work? Or am I doing something wrong?