| 9 Nov 2024 |
 aidalgol | Semi-related to this: I found that you can nix flake update input-a from repo B when input A points to a local checkout with uncommitted changes, and repo B will now be pointing to the uncommitted changes in your local repo A. | 23:19:58 |
 KFears (annoying) | In reply to @benjamin:computer.surgery
I cannot imagine a situation where the current behavior would be desirable Me neither, but with flakes, better safe than sorry :P
Still, it's like a CLI issue and I would expect people to run into actually insane breakages due to old transitive inputs, it just doesn't help anyone, and it's thankfully isolated to CLI I think | 23:20:05 |
| aidalgol | It was somewhat useful, but it's also a bit of a footgun. | 23:20:21 |
| KFears (annoying) | In reply to @aidalgol:matrix.org
Semi-related to this: I found that you can nix flake update input-a from repo B when input A points to a local checkout with uncommitted changes, and repo B will now be pointing to the uncommitted changes in your local repo A. Wait wut, can you give an inputs example for it? Does it work only for path, or for git+file too? | 23:22:06 |
| aidalgol | IIRC, I was using git+file. | 23:22:29 |
 olivia | cursed | 23:22:38 |
| KFears (annoying) | No way, that sounds broken | 23:22:41 |
| KFears (annoying) | Eww | 23:22:48 |
| olivia | I would expect that to work with path: but not git+file | 23:22:52 |
| KFears (annoying) | Wait, so what does ref and rev give you in the lockfile in that case? | 23:23:18 |
| aidalgol | I ran into it while working on https://github.com/aidalgol/valheim-server-flake and https://github.com/nix-community/steam-fetcher, when I needed to change something in steam-fetcher but test it via valheim-server-flake. | 23:23:49 |
| aidalgol | This was before I switched to Lix, though. Let me see if I can make it happen again. | 23:24:02 |
| KFears (annoying) | Oh, reading 2.18 manual, there's this:
When git+file is used without specifying ref or rev, files are fetched directly from the local path as long as they have been added to the Git repository. If there are uncommitted changes, the reference is treated as dirty and a warning is printed.
| 23:24:21 |
| aidalgol | What's the URL for whatever you're reading? | 23:24:57 |
| KFears (annoying) | Gahh, I hate it. I also disable warn-dirty because it's so ridiculously noisy for normal workflows... | 23:26:33 |
| aidalgol | OK, here we go. I'm going to abbreviate the repos to vsf and sf. What I did was.
- Made a code change in my local checkout of
sf (did not stage or commit anything)
- In my local
vsf checkout, I changed the input for sf to git+file:///path/to/sf
- Ran
nix flake update sf
This is the git diff of flake.lock in my local vsf checkout.:
diff --git a/flake.lock b/flake.lock
index dae015e..638ec92 100644
--- a/flake.lock
+++ b/flake.lock
@@ -29,17 +29,16 @@
]
},
"locked": {
+ "dirtyRev": "12f66eafb7862d91b3e30c14035f96a21941bd9c-dirty",
+ "dirtyShortRev": "12f66ea-dirty",
"lastModified": 1714795926,
- "narHash": "sha256-PkgC9jqoN6cJ8XYzTA2PlrWs7aPJkM3BGiTxNqax0cA=",
- "owner": "nix-community",
- "repo": "steam-fetcher",
- "rev": "12f66eafb7862d91b3e30c14035f96a21941bd9c",
- "type": "github"
+ "narHash": "sha256-nyLyYPhzZ76Gltyeyl3/b4sLPc+W7L7m2BTTT3icyOw=",
+ "type": "git",
+ "url": "file:///home/aidan/src/nix-steam-fetcher"
},
"original": {
- "owner": "nix-community",
- "repo": "steam-fetcher",
- "type": "github"
+ "type": "git",
+ "url": "file:///home/aidan/src/nix-steam-fetcher"
}
}
},
| 23:30:53 |
| KFears (annoying) | It will probably be unfixable too, until we can fork off flakes :/
Given that dirty warnings pop up even when the only local flake you have is the one you're directly working with, it seems like self is treated as git+file, and removing the dirty behavior will break iterative improvements on self extremely hard, and replacing git+file with path will probably break too much stuff people are used to | 23:32:07 |
| aidalgol | Er, correction. I did not actually do step 3. What causes the flake.lock to be updated is direnv. | 23:32:13 |
| aidalgol | I have use flake in .envrc in this repo. | 23:32:24 |
| aidalgol | Is direnv supposed to mess with the lock file? | 23:32:41 |
| KFears (annoying) | In reply to @aidalgol:matrix.org
OK, here we go. I'm going to abbreviate the repos to vsf and sf. What I did was.
- Made a code change in my local checkout of
sf (did not stage or commit anything)
- In my local
vsf checkout, I changed the input for sf to git+file:///path/to/sf
- Ran
nix flake update sf
This is the git diff of flake.lock in my local vsf checkout.:
diff --git a/flake.lock b/flake.lock
index dae015e..638ec92 100644
--- a/flake.lock
+++ b/flake.lock
@@ -29,17 +29,16 @@
]
},
"locked": {
+ "dirtyRev": "12f66eafb7862d91b3e30c14035f96a21941bd9c-dirty",
+ "dirtyShortRev": "12f66ea-dirty",
"lastModified": 1714795926,
- "narHash": "sha256-PkgC9jqoN6cJ8XYzTA2PlrWs7aPJkM3BGiTxNqax0cA=",
- "owner": "nix-community",
- "repo": "steam-fetcher",
- "rev": "12f66eafb7862d91b3e30c14035f96a21941bd9c",
- "type": "github"
+ "narHash": "sha256-nyLyYPhzZ76Gltyeyl3/b4sLPc+W7L7m2BTTT3icyOw=",
+ "type": "git",
+ "url": "file:///home/aidan/src/nix-steam-fetcher"
},
"original": {
- "owner": "nix-community",
- "repo": "steam-fetcher",
- "type": "github"
+ "type": "git",
+ "url": "file:///home/aidan/src/nix-steam-fetcher"
}
}
},
Oh. So it gets worse | 23:32:48 |
| KFears (annoying) | 💀💀💀💀💀💀💀 | 23:32:55 |
| * aidalgol dies | 23:33:03 |
| KFears (annoying) | Thanks for the report, I'll try to document it somewhere | 23:33:30 |
| aidalgol | Is this worth raising as a bug on the Lix issue tracker just so it's on record? | 23:33:34 |
| aidalgol | Or I'll leave it to you. :P | 23:33:38 |
| KFears (annoying) | I'm not sure that it even makes sense to open a bug D:
Seems like a design issue in flakes that we'll have to live with, because actually fixing it will break the universe | 23:34:18 |
| KFears (annoying) | In reply to @aidalgol:matrix.org
Is direnv supposed to mess with the lock file? I think it is, at least it has been updating my lock files for me for years | 23:34:53 |
| aidalgol | Yeah, best not to do that until such time as it is worth it for Lix to diverge from CppNix. | 23:35:16 |
| 10 Nov 2024 |
|  kurotoshiro joined the room. | 13:00:38 |
