chore: update flate2 version & use zlib-rs

[jorgehermo9]

As described in this blog post https://trifectatech.org/blog/zlib-rs-is-faster-than-c/ zlib-rs is a faster and pure Rust zlib alternative and those performance upgrades are included in flate2 1.1.

What are your thoughts on adopting it?

[jorgehermo9]

I think tests are broken due to what was commented in #569 or #572

[JosiahParry]

Would there be a possibility of creating a feature for using zlib-rs instead of libz-ng-sys as a backend? It would be great if the required system dependencies can be reduced.

[jorgehermo9]

Hi @drcaramelsyrup, Just noticed that I had this PR forgotten. Merged with main and now CI passes!

As @JosiahParry, Is there any need to let users configure their flate2 feature flag? or should we just use zlib-rs always?

[DarrenASmith]

Any chance this could be merged?
We are having problems with the dependency on "zlib-ng" which then tries to build with cmake and would like the option to be able to use "zlib-rs" instead

[JosiahParry]

sorry i never replied! 🙈 my specific use case is for a binary which doesn't need to expose feature flags to users

[jorgehermo9]

@JosiahParry

I'm not sure if I'm understanding you correctly

Do you mean to let the dependent packages of pingora-core to control that feature flag, through pingora-core's own features? So we let them control the flate2 features themselves... Or what do you mean?

[DarrenASmith]

@JosiahParry

I'm not sure if I'm understanding you correctly

Do you mean to let the dependent packages of pingora-core to control that feature flag, through pingora-core's own features? So we let them control the flate2 features themselves... Or what do you mean?

If that's possible, that would work for my use case.
ie. pingora with "zlib-rs" feature sets that feature on it's flate2 dependency.
(the PR merged as is with the dependency hard-coded to zlib-rs would also be fine for my use case)