Check entries' paths rather than path_bytes

[smoelius]

This PR changes how db-dump.tar.gz entries are checked. Instead of using path_bytes, the new code uses path.

crates.io recently started publishing deleted crates in a file deleted_crates.csv: rust-lang/crates.io#12803

That file would match this line:

cargo-supply-chain/src/crates_cache.rs

Line 207 in 18db9ea

} else if entry.path_bytes().ends_with(b"crates.csv") {

Thus, if deleted_crates.csv appeared later than crates.csv in db-dump.tar.gz, deleted_crates.csv would be used to populate crates.json.

The net effect was the following. If a user did cargo supply-chain update --cache-max-age=0s, everything would appear to update fine. But if the user then did cargo supply-chain json, the wrong crates would be in crates.json and they would have to be downloaded (i.e., the cache would provide no benefit).

Nits are welcome.

---

Separate from the above, would you consider a PR to rename --cache-max-age to --max-age, but keep --cache-max-age as an alias?

[Shnatsel]

Separate from the above, would you consider a PR to rename --cache-max-age to --max-age, but keep --cache-max-age as an alias?

Not really. --max-age is an awkward middle ground that's neither descriptive nor quick to type. If you need something that can be quickly typed, we should add a single-letter alias.

[smoelius]

Thanks for your response, and thanks for the rapid merger of this PR! ⚡