Dimitar Georgiev

Hello folks, and thanks for your awesome work! - Is the expected end-user solution for now to create manually an A-record pointing to `traefik`? - Do you have a rough...

Reproduced on proprietary codebase with a non-contrived example. Getting times in the same magnitude as cited on that repro readme.

Compiled with scalac profiling, you can obtain the results here https://github.com/dimitarg/skunk-test/blob/main/trace.trace (this file is openable e.g. via a chromium browser if you type `chrome://tracing/` in the address bar) If I...

P.S. my fork of the repro repo must be compiled with `jdk 8`. I had trouble using jdk 11 on this - the compiler with tracing turned on crashes due...

I added `-Ystatistics:patmat` to the repro compiler settings. Here are the results. ``` 2021.11.09 10:43:21 INFO *** Cumulative statistics at phase patmat 2021.11.09 10:43:21 INFO #class symbols : 6025 2021.11.09...

This diff exhibits the same pathology. So maybe at least we can rule out tilde `unapply` as the villain. https://github.com/dimitarg/skunk-test/commit/55e7d75a139c07e9d7deae81530c0e59c5470b94

This compiles instantaneously, proving that indeed the pattern match is where scalac spends its time. https://github.com/dimitarg/skunk-test/commit/acbc9c8ecc6336e25cc271d8220def75254f087c

This appears to be a viable workaround https://github.com/dimitarg/skunk-test/commit/06d813f12228a5bf7987db228b5545641020785b

This is a minimal reproduction of the issue without skunk or external dependencies, courtesy of @jdrphillips https://gist.github.com/jdrphillips/5dc1f533e8d44f9b34962c79798d8839

This is also tracked via https://github.com/scala/bug/issues/12489