ltar cache failures in CI

Stream: general

[Eric Wieser]

CI seems to be failing with

thread '<unnamed>' panicked at 'called `Result::unwrap()` on an `Err` value: Error { kind: UnexpectedEof, message: "failed to fill whole buffer" }', /project/src/ltar.rs:57:34
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
thread '<unnamed>' panicked at 'called `Result::unwrap()` on an `Err` value: Error { kind: UnexpectedEof, message: "failed to fill whole buffer" }', /project/src/ltar.rs:57:34
thread '<unnamed>' panicked at 'called `Result::unwrap()` on an `Err` value: Error { kind: UnexpectedEof, message: "failed to fill whole buffer" }', /project/src/ltar.rs:57:34
thread '<unnamed>' panicked at 'called `Result::unwrap()` on an `Err` value: Error { kind: UnexpectedEof, message: "failed to fill whole buffer" }', /project/src/ltar.rs:57:34

[Oliver Nash]

This just happened in two bors runs (admittedly one was a rerun of the same PR).

[Anatole Dedecker]

I noticed that just rerunning the failed task doesn’t work, you have to redo everything for it to fix itself

[Oliver Nash]

Do we understand what is going on here? It's totally opaque to me.

[Mario Carneiro]

That error message means that the .ltar file was truncated in the middle

[Mario Carneiro]

usually this happens when the download fails or is cancelled and the cleanup doesn't work for some reason

[Mario Carneiro]

although it could also be that the cache file uploaded to the server was truncated

[Alex J. Best]

Can there be some sort of checksum at the start of ltar files or something to check that the file isn't corrupted / truncated and emit a good error message / redownload in that case?