Various minor fixes and improvements

[Osse]

• Add top-level Cargo.toml defining workspace

• Run cargo fmt

• lager: Fix tests

  Test tests wouldn't compile because of wrong array size.

• halide-cache: Add crude system test

  This started out as a shell script, but add a very simple Rust wrapper
  in order to be able to call it as cargo test. Rewrite later.

• lager: Make additional error variant and simplify

  With the extra InvalidAddress variant the Runtime variant does not need
  an optional source error inside it. This makes lager::Error interact
  more nicely with anyhow.

  Leverage the #[from] attribute of thiserror to simplify errors from
  dependent libraries.

• halide-cache: Use anyhow for top-level errors

  Error propagation because easier and main() easier to read.

• halide-cache: Be stricter about cache errors

  Comparted to the previous implementation of cache_hit() this one
  attempts to retrieve both the header and object before doing any error
  handling. In the happy path we end up calling both anyway so that's not
  an issue.

  It might not be a good ideea in the future to let cache failures be
  fatal like this, but it helps with development.

• halide-cache: Collect dependencies in a struct

  Instead of converting things to vector of strings and so on, make a
  simple struct to store the dependency info. Add make_address() to that
  struct to generate the required address directly. We now hash the data
  directly instead of doing hashes of hashes.

  Additionally, generate the 64 bytes directly.

[RMPR]

Looks very nice! If you rebase you should be able to build on Windows.

I am not a big fan of the last commit as it most likely breaks the build on Windows, but we will see what the CI says.

[Osse]

Looks very nice! If you rebase you should be able to build on Windows.

I am not a big fan of the last commit as it most likely breaks the build on Windows, but we will see what the CI says.

The last, really? Do you actually mean the system test commit? Or perhaps the workspace one? I have no issue removing either one.

[RMPR]

The last, really?

Yup, the one that introduces the unix specific extensions

error[E0433]: failed to resolve: could not find `unix` in `os`
 --> halide-cache\src\main.rs:4:14
  |
4 | use std::os::unix::ffi::OsStrExt;
  |              ^^^^ could not find `unix` in `os`
  |
note: found an item that was configured out
 --> /rustc/01f6ddf7588f42ae2d7eb0a2f21d44e8e96674cf\library\std\src\os\mod.rs:29:5
  |
  = note: the item is gated here
note: found an item that was configured out
 --> /rustc/01f6ddf7588f42ae2d7eb0a2f21d44e8e96674cf\library\std\src\os\mod.rs:84:41
  |
  = note: the item is gated here
error[E0599]: no method named `as_bytes` found for reference `&std::ffi::OsStr` in the current scope
  --> halide-cache\src\main.rs:33:45
   |
33 |         hasher.update(self.path.as_os_str().as_bytes());
   |                                             ^^^^^^^^
   |
help: there is a method `as_encoded_bytes` with a similar name
   |
33 |         hasher.update(self.path.as_os_str().as_encoded_bytes());
   |                                                ++++++++
Some errors have detailed explanations: E0433, E0599.
For more information about an error, try `rustc --explain E0433`.
error: could not compile `halide-cache` (bin "halide-cache") due to 2 previous errors
Caused by:

https://github.com/zivid/halide-cache/actions/runs/23163614443/job/67370342216?pr=1

Now that we have CI that actually works.

[RMPR]

This doesn't build on Windows.

[Osse]

I didn't notice it. Using as_encoded_bytes() is good enough for this usecase. No OS-specific traits needed.

[RMPR]

LGTM