Mingun

Not right now. I think, we should not introduce new dependences just because we can. As I already mentioned, if this PR will introduce some useful feature that we doesn't...

The only required dependency is `memchr` and our MSRV restricted only by it. Actually, we use some features (I don't mean cargo features here) that requires 1.52. But enabling some...

I started to write some docstring examples (well, almost couple of years ago), but realized, that we should fix our constructors. Currently we trust to users that the strings passed...

Could you post all code that reproduces the problem (ideally as Rust testcase)? Because we have exactly this testcase here: https://github.com/tafia/quick-xml/blob/64c42495c0596933436d2887f5eb658c21083ae6/tests/serde-de-seq.rs#L3326-L3339

The problem is in namespace prefix. Namespaces are not properly supported in serde yet. If you remove `ns2:` from `cancelReason` here, the test is passed: ```rust #[test] fn issue683() {...

Because your XML use namespaces, try to look at [xmlserde](https://lib.rs/crates/xmlserde) which should support namespaces better. It is based on quick-xml, so should be as fast as quick-xml's deserializer.

Thanks for PR! It is useful. I (or someone else who is interested) use it when I'll deal with the problem, but for now let it hang open.

Probably this bug is OS-dependent. I tried to increase for loops to `0..3000` and didn't get errors on Windows 11

In case if GitHub delete logs -- error is ``` failures: ---- test_cancel_future stdout ---- thread 'test_cancel_future' panicked at tests/async-tokio.rs:79:22: called `Result::unwrap()` on an `Err` value: IllFormed(MismatchedEnd { expected: "

Well, then I can say, that you probably should try to debug and fix it yourself. I'll plan to make a PR with completely new parser soon, which maybe will...