I think #1866 should be treated as a breaking change.

I think #1866 should be treated as a breaking change.

What code do you think it breaks? It seems to strictly make more code work, in particular code that trivially worked pre-#1731 and for which there is no obvious/simple alternative. Some of my downstream code turns out to be affected, too, in that if call SomeServiceClient::new() and start calling methods on it, that just fails if I give it a https URL. How should I rewrite that code, in your opinion?

Makes sense. I understand that it can be treated as not breaking change.

I've revised the changelog and added more things to it. I think this is ready. @LucioFranco can you get it released?

It does not matter if it is a side effect, it is whether it is intended or not.

So when you wrote #1731, did you consider (and intend) that SomeServiceClient::new() would no longer be able to handle https endpoints? I did not consider that while reviewing, and would have objected if I did. From that perspective, it is not entirely intentional.

From your comments it seems that you are picking up pieces of information and judging whether something is a bug or not based on your own preferences.

The 0.12.0 release is a major release that assumes the existence of breaking changes, so it is legitimate for it to include any breaking changes. Whether or not you like a change has no bearing upon whether or not it is a bug. It is a technical problem.

IMO it is reasonable to consider it a bug that SomeServiceClient::new() still compiled in 0.12.0 but failed to work with https endpoint URLs. While it may be technically allowed by semver, I think the effect of the changes in #1731 on that scenario are problematic for this reason. In general, if one (or two) maintainers think something is reasonable but most of the downstream users disagree, I'm inclined to think the users are right by default.

So when you wrote #1731, did you consider (and intend) that SomeServiceClient::new() would no longer be able to handle https endpoints?

It works with adding appropriate TLS configuration after that change. It is intended and the one of the main purposes of the changes.

This was published: https://crates.io/crates/tonic/0.12.2.