Oh that's odd. They weren't fatal when I ran them locally...I'll look into this.

sbt-spiewak makes them fatal on CI, but not locally. This is convenient when exploring and having unused imports, but it's frustrating when it comes PR time. I'm not sure I like the behavior, and we can turn it off, but that's why, and it's a recent change.

@rossabaker ah, that makes sense. I was grepping for a -Werror or -Xfatal-warnings I'd thought I'd missed.

I would be fine disabling that in CI. For just about all scopes other than docs I do want -Werror, but docs is the exception.

My attention has been split and there are a lot of exemptions floating around in a lot of plugins and I guess I need a mental reset.

So as of be0890c, we have several linking warnings. These are fatal in CI. Can we clean these up and leave the flags as they are? Sometimes I try to do too many things at once and rely on CI to alert me about broken docs, so I have a mild preference for leaving them on so they're seen as they happen. But they can be annoying -- last time I tried, we couldn't run mdoc cleanly on 2.13, which is why we just generate them from 2.12.

I'm pretty open to whatever solution you think is best here.

@rossabaker the remaining warnings are all related to overloaded symbols. I would love to clean them up, but for the life of me I can't seem to get the correct incantation. You are supposed to be able to provide part of the signature to distinguish between overloaded symbols, but nothing I tried seemed to work.

I'll attempt to another crack at it before we change the warning flags.

An unsatisfying solution, if all else fails, is also to just unlink them.

@rossabaker

An unsatisfying solution, if all else fails, is also to just unlink them.

I'd be hesitant to do that in this case. When there is an overloaded symbol in a Scaladoc link, Scaladoc will still generate a link to one of the members (I assume it just picks one at random), it just also emits a warning. So, from a user's perspective, they are still getting the nice linking. It's just a pain in the build.

Ah, that makes sense. Could we turn off fatal warnings for doc and unidoc, but leave them on for compiling and mdoc?

I'd like to land this one. It's also currently affected by typelevel/sbt-typelevel#226 I believe.