Move deny warnings to CI

[markmandel]

What is nice about this is that we deny warnings for the stable version of Rust that we support, and we ignore warnings for the nightly version, which has more changes more often.

Closes #189

[iffyio]

looking at #204 and the ci failure and I think I'm wondering why the warning triggered now 😮 if we're using the same nightly binary we should have the same warning rules right?
thinking we should keep warnings on for nightly as well? Like in this case the warning is actually useful since its something we'll need to fix eventually anyways, we might as well do it as soon as its introduced?

 error: use of deprecated associated function `std::sync::atomic::AtomicUsize::compare_and_swap`: Use `compare_exchange` or `compare_exchange_weak` instead
Step #4 - "external-doc-test":    --> src/extensions/filters/local_rate_limit/mod.rs:138:57
Step #4 - "external-doc-test":     |
Step #4 - "external-doc-test": 138 | ...                   refilled = available_tokens.compare_and_swap(
Step #4 - "external-doc-test":     |                                                   ^^^^^^^^^^^^^^^^
Step #4 - "external-doc-test":     |
Step #4 - "external-doc-test": note: the lint level is defined here
Step #4 - "external-doc-test":    --> src/lib.rs:18:9
Step #4 - "external-doc-test":     |
Step #4 - "external-doc-test": 18  | #![deny(warnings)]
Step #4 - "external-doc-test":     |         ^^^^^^^^
Step #4 - "external-doc-test":     = note: `#[deny(deprecated)]` implied by `#[deny(warnings)]`
Step #4 - "external-doc-test": 
Step #4 - "external-doc-test": error: use of deprecated associated function `std::sync::atomic::AtomicUsize::compare_and_swap`: Use `compare_exchange` or `compare_exchange_weak` instead
Step #4 - "external-doc-test":    --> src/extensions/filters/local_rate_limit/mod.rs:169:38
Step #4 - "external-doc-test":     |
Step #4 - "external-doc-test": 169 |             if self.available_tokens.compare_and_swap(

[markmandel]

looking at #204 and the ci failure and I think I'm wondering why the warning triggered now

Because the CI image got rebuilt (I didn't do a pull/push, I rebuilt the build script to point to the new location), we got a new nightly build -- which is more up to date.

thinking we should keep warnings on for nightly as well

I have no problem with warnings being visible 👍 (in fact I tested this when I updated nightly locally, and the warnings still show up when running the --doc tests), but didn't want them to actually fail the test. So I think this works well, in that we can see the warnings coming in the nightly tests, but they aren't going to block us from finishing work, when the nightly system is constantly in flux (especially on local machines of external contributors).

I'm assuming that having warnings be visible, but not cause an abort of the process is what you mean by "on" ?

[iffyio]

Because the CI image got rebuilt (I didn't do a pull/push, I rebuilt the build script to point to the new location), we got a new nightly build -- which is more up to date.

Ah that explains why it warns now make sense!

I'm assuming that having warnings be visible, but not cause an abort of the process is what you mean by "on" ?

I was referring to failing the build as usual if nightly produces warnings as well but yeah I think as you mentioned it makes sense to not do so since people might have different nightlys and we don't lose much otherwise. the current fix still LGTM!