Repository navigation
Use published, versioned bootc internal crates - #403
Conversation
We don't need to track these by hash anymore, and more importantly the gitref dependencies are not grouped properly by renovate, so these crates keep getting bumped independently and they always conflict with each other. This also reverts bootc-dev#367 where I tried (and failed) to get them to group properly. Signed-off-by: John Eckersberg <dev@eckersberg.com>
| bootc-mount = { package = "bootc-internal-mount", git = "https://github.com/bootc-dev/bootc", rev = "83d15891428cda862c5c732397f5dd7c81dc4a23" } | ||
| bootc-utils = { package = "bootc-internal-utils", git = "https://github.com/bootc-dev/bootc", rev = "83d15891428cda862c5c732397f5dd7c81dc4a23" } | ||
| bootc-mount = { package = "bootc-internal-mount", version = "1.16.14" } | ||
| bootc-utils = { package = "bootc-internal-utils", version = "1.16.14" } |
There was a problem hiding this comment.
I think renovate may still try to bump these separately. The best fix is to re-export one of them from the other; most likely bootc-mount should re-export bootc_utils.
There was a problem hiding this comment.
I think they'll just all get bundled together into the "rust" one like #400
There was a problem hiding this comment.
Ah OK yeah probably true, but it's still a best practice to do the "re-export" thing so there's only one version. It's what we did with composefs-rs for bootc
There was a problem hiding this comment.
I don't think the re-export is necessary here because although mount uses utils (only for CommandRunExt) it doesn't leak any of the API from utils into its own API. The problem with bootc/composesfs-rs was that bootc exposed parts of the composefs-rs API in its own API, so you could get into version mismatches where the "inner" version of composefs-rs that bootc was built against was different from the external version used by the caller.
In the case here with bcvk it doesn't really matter too much except for a slightly larger binary size if the two diverge, but I think that's probably trivial in the grand scheme of things.
2 participants
We don't need to track these by hash anymore, and more importantly the gitref dependencies are not grouped properly by renovate, so these crates keep getting bumped independently and they always conflict with each other.
This also reverts #367 where I tried (and failed) to get them to group properly.