I'm confused, was there a platform-specific issue here, or wasn't there? if it was passing before on Linux, won't it start failing now? I guess let's see what the PR validator says. we can also do a https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-windows/ run before merging, for good measure

That would be nice. I believe the problem was (as Adriaan mentioned) that this went in before the compiler determinism PR, so platform issues could affect the names of synthetic methods that used a fresh name creator.

FWIW, the test case (before this change) was passing on my Mac when I made the earlier PR, and now fails, with the exact same system configuration.

(the same failure also turned up on another PR, so that removes any doubt this was a semantic merge conflict rather than a platform-specific issue)