Check minimal version of Rust that compiles

[jongiddy]

Specify a minimum version of the Rust compiler that successfully builds the crate. This allows us to be aware if the minimal version increases and to avoid issues such as #370

The minimum version is a best-effort measured value and is different to the MSRV which is a support guarantee. The minimum version can be incremented by a PR in order to pass tests, as long as the MSRV holds. When the minimum version increases, the next release should be a minor version, to allow any affected users to pin to a previous minor version.

[jongiddy]

The minimum version for the updated miniz_oxide is higher than I expected. Since the rust-version in Cargo.toml can prevent Cargo from selecting the crate, it needs to be the minimum version supported by a common feature-set. We do not want to prevent someone using Rust 1.56.1 from downloading the crate if they are planning to use zlib-* feature.

My current thinking is to remove the rust-version directive from Cargo.toml but keep the test, and use it to document behavior. (e.g. if using Rust < 1.60 a different backend must be selected).

[oyvindln]

The current version of miniz_oxide works with 1.56.0 as of now if I test it on it's own locally though I don't want to give any guarantees that it will stay at that.

It looks like you may need to pin some other crates to older versions to make flate2 work with that version though judging by the CI build results.

[jongiddy]

Ah it's ppv-lite86 which is a transitive dependency in the dev-dependencies. v0.2.17 in my lock file has no dependencies, but the latest v0.2.20 used in GitHub Actions adds zerocopy which requires 1.60.

[dev-dependencies]
├── quickcheck v1.0.3
│   └── rand v0.8.5
│       ├── libc v0.2.153
│       ├── rand_chacha v0.3.1
│       │   ├── ppv-lite86 v0.2.17

[jongiddy]

Changed the CI workflow to only test that code builds, without running the tests.

[Byron]

Thanks a lot for making clearer the defacto way the MSRV was handled.

I'd also hope that adding the rust-version can't break anyone. In theory, those who use even more ancient compiler versions will benefit from those not respecting that field yet. So it should be safe especially with a value as low as that.

There is just one improvement that I think could be made to the MSRV paragraph, which would have helped me to understand it more quickly.

[Byron]

Suggested change

	The `Cargo.toml` file specifies a minimum version for which builds passed at some point.
	The `Cargo.toml` file specifies a minimum version for which builds passed at some point using the `rust-version` field.

[Byron]

Since 'minimum version' is really the rust-version field, I wonder if that name should rather be used to reduce the level of abstraction.

[Byron]

Thanks a lot!

[Byron]

CI now fails on main, and I have a feeling that this entry might be the cause of it?

It's surprising that it didn't fail CI before.

In any case, your help for a fix will be welcome.

[jongiddy]

Oh wow, CI should complain more noisily if that happens!

#422 should be the fix.