You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Vendored crates should be checked for changesΒ #5805
I was building Firefox, and tried to do some changes to some rust crate. Doing so made the build fail with something like
error: the listed checksum of ... has changed:
expected: ...
actual: ...
directory sources are not intended to be edited, if modifications are required then it is recommended that [replace] is used with a forked copy of the source
That's an IMO annoying feature of cargo, but this is not what this issue is about. The problem is the following: if I actually build the crate first, and then make the changes, rebuilding just goes on acting as if I had made no change, saying Fresh $crate_name.
It seems to me cargo should fail with the exact same error as if I hadn't already built.
This is currently expected behavior in that Cargo treats vendored crates like it does crates.io, it doesn't check any mtimes as it assumes it's all read-only. It is a performance optimization to not even look at the filesystem for these crates on incremental builds.
The correct way to modify a vendored crate is to use [patch] or a path dependency, and Cargo will then correctly handle the crate on incremental rebuilds as it knows that it's not a readonly dependency.
The correct way to modify a vendored crate is to use [patch] or a path dependency, and Cargo will then correctly handle the crate on incremental rebuilds as it knows that it's not a readonly dependency.
Fortunately GNU Guix can bypass this pointless check, as I today discovered.
When I vendor code into my repository, it changes line endings (and nothing else) which triggers this check. Is there any way to disable or at least mitigate the error in this situation?
Instead of closing this now, I would like to see an improvement in the cargo-vendor doc. Could document that vendored sources are not intended to be modified, and [patch] is the correct way to do it.
Instead of closing this now, I would like to see an improvement in the cargo-vendor doc. Could document that vendored sources are not intended to be modified, and [patch] is the correct way to do it.
cargo-vendor should have an option to generate the necessary [patch] for the vendored crates.
Reacted by Yuriy Al. Shirokov, Hamilton Chapman and Danny Hammer
Activity
alexcrichton commentedon Jul 26, 2018
This is currently expected behavior in that Cargo treats vendored crates like it does crates.io, it doesn't check any mtimes as it assumes it's all read-only. It is a performance optimization to not even look at the filesystem for these crates on incremental builds.
The correct way to modify a vendored crate is to use
[patch]or apathdependency, and Cargo will then correctly handle the crate on incremental rebuilds as it knows that it's not a readonly dependency.ctrlcctrlv commentedon Oct 24, 2022
Fortunately GNU Guix can bypass this pointless check, as I today discovered.
https://fredrickbrennan.medium.com/how-to-bypass-rust-cargo-error-directory-sources-are-not-intended-to-be-edited-ab2f257b89da
Initial commit
k3d3 commentedon Jan 9, 2023
When I vendor code into my repository, it changes line endings (and nothing else) which triggers this check. Is there any way to disable or at least mitigate the error in this situation?
ctrlcctrlv commentedon Jan 10, 2023
Does revendor.scm not work for you?
epage commentedon Oct 23, 2023
As this is expected behavior and is important for performance reasons, I'm going to propose to the cargo team that we close this.
weihanglo commentedon Dec 22, 2023
Instead of closing this now, I would like to see an improvement in the cargo-vendor doc. Could document that vendored sources are not intended to be modified, and
[patch]is the correct way to do it.briansmith commentedon Dec 22, 2023
cargo-vendor should have an option to generate the necessary
[patch]for the vendored crates.LuuuXXX commentedon Feb 28, 2024
claim first, question it later~ @rustbot claim
Auto merge of #13512 - LuuuXXX:issue-5805, r=weihanglo
weihanglo commentedon Mar 1, 2024
#13512 has improved the doc a bit, so I am fine closing this.
@briansmith feel free to open an issue if you have an idea how it would look like. Thank you :)