| 7 Aug 2025 |
 emily | I don't see why | 15:56:52 |
| emily | the files are tagged appropriately at the relevant point in the history | 15:56:56 |
 K900 | But also like that history is objectively garbage | 15:57:32 |
| K900 | Because none of that code matters | 15:57:35 |
| K900 | And it's only there because vlinkz had no idea how to calamares | 15:57:44 |
| emily | I will be really unhappy if you break my git blame and git bisect. | 15:58:03 |
| K900 | I wonder if git blame even works | 15:58:27 |
| K900 | Without the history rewrite | 15:58:33 |
| K900 | Also I can 100% promise you that history won't be bisectable anyway | 15:58:48 |
| K900 | Because the specific changes on the installer will be specific to nixpkgs revisions | 15:59:05 |
| K900 | And none of that will linearize properly | 15:59:12 |
| emily | 
Download image.png | 16:01:38 |
| emily | lol okay | 16:01:41 |
| emily | I guess I'm fine with rewriting then. though it'd be nice to add the package prefix if we're going to be doing that | 16:02:58 |
| K900 | OK sorry I have a real bad migraine but I can cook up some filter-branch fuckery later | 16:04:45 |
| K900 | Or feel free to force push if you want | 16:04:48 |
| emily | it's alright, no rush | 16:05:14 |
| emily | I'm okay with the current state if it's what's required to get Git to behave | 16:05:23 |
| emily | as annoying as it is | 16:05:31 |
| K900 | Git is just terrible | 16:06:20 |
| emily | I wouldn't know anything about that 😈 | 16:15:06 |
| emily | I would make a Jujutsu brag but Jujutsu currently lacks the bloom filter stuff Git uses to make git log on individual paths fast so I actually just use git log for that still. | 16:15:29 |
| K900 | jj can't really fix the underlying data model | 16:16:37 |
 Alyssa Ross | In reply to @k900:0upti.me
I wonder if git blame even works It should | 16:20:11 |
| Alyssa Ross | I blame into subtrees relatively often | 16:20:26 |
| emily | fwiw I tried (with a non-rewriting non-squashing subtree merge) and it didn't work | 16:21:15 |
| emily | there is no underlying data model problem with following the log here | 16:21:32 |
| emily | --follow is just bad | 16:21:34 |
| emily | but also there is no "data model" for copies/moves given Git is snapshot-based anyway | 16:21:48 |
| K900 | You can't have a subtree merge that both updates paths in the history and preserves original commit hashes | 16:22:25 |
