Printing -v output to stderr makes it work well when any errors occur during compilation, or when using -x. Why is this a bug?

I see what you are saying. When compiling multiple packages, if a package has a compilation error you would like to know which package it is for. However, I believe this is already handled, because if you do go build and there is an error you will see the following output.

E:\gowork\src\github.com\joegrasse\passtest>go build
# github.com/joegrasse/passtest
.\passtest.go:26: syntax error: unexpected }

Notice the package name printed once vs the package name printed twice below with the -v flag.

E:\gowork\src\github.com\joegrasse\passtest>go build -v
github.com/joegrasse/passtest
# github.com/joegrasse/passtest
.\passtest.go:26: syntax error: unexpected }

My issue is when there are no "true" errors you still have output to stderr with the -v option. Which in turn leads to issue GoClipse/goclipse#119.

I agree that the error message tells you which package failed, but it's still convenient to not have strange interleaving in the output. Plus, -x.

I don't understand why goclipse is passing -v at all. This doesn't seem like a strong reason to change the way the go tool works.

I guess I would ask why -x is sent to stderr as well. Why not stdout for that, since it is information and not errors?

OK, forget -x. We can agree that -x and -v should use the same rule.

It's still convenient to not have strange interleaving in the output.

I don't see the fact that goclipse decides to pass -v as a compelling reason to change how the tool behaves, because I don't know a reason why goclipse should do that.

Working as intended. Diagnostic output goes to stderr; that's what stderr is for.