Fix deflate decompression

[cipherbrain]

This solve issue 1462 by using 'deflate' algo instead of 'zlib' algo to decompress responses with the 'deflate' Content-Encoding.

The full series of tests (cargo test) still works after this.

My code (see isseu #1462) now works, where I was getting " deflate decompression error" before my patch.

[seanmonstar]

Makes sense to me, thanks!

[cipherbrain]

You're welcome.

But the CI doesn't seem to like the patch, and rejected it. I'm not sure if it's my patch, some unrelated error, or how I could reproduce the errors on my machine to have a chance to solve them.

Doing "cargo test --all-features" doesn't work even without my patch, and "cargo test" works with my patch.

If you have some idea on what I could do to have it pass the tests, I'm willing to try that.

Anyways, thanks for your work on this great library!

[seanmonstar]

Does running cargo test --feature deflate --test deflate work, with your patch? It looks like this is the failure:

deflate_single_byte_chunks' panicked at 'text: reqwest::Error { kind: Decode, source: Custom { kind: Other, error: DecompressError(General { msg: None }) } }', tests/deflate.rs:149:33

[T-Sujeeban]

The deflate test fails because it also uses Zlib instead of Deflate
The line 95 in tests/deflate.rs should be changed from
let mut encoder = libflate::zlib::Encoder::new(Vec::new()).unwrap();
to this
let mut encoder = libflate::deflate::Encoder::new(Vec::new());

[seanmonstar]

Thanks for finding that!

[jongiddy]

Is this correct behavior? RFC9110 says deflate coding is a deflate stream inside a zlib container, with a note that some non-conformant implementations use plain deflate.

[seanmonstar]

Ah, I hadn't checked that out, seems the proper fix is more involved. Would it be to use zlib, and if that fails then try deflate?

[inikulin]

Would it be to use zlib, and if that fails then try deflate?

Yes, this is usually how browsers handle it