investigate 2.13 community build breakage for 2.13.0-RC1

[SethTisue]

current nightly: 2.13.0-pre-59975bb 2.13.0-pre-af24410 2.13.0-pre-b4926ef

I did a sweep through the latest run log at https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/ , findings as follows (numbers are "number of downstream projects blocked")

advice for those who want to help is now up at https://github.com/scala/community-builds/wiki/Troubleshooting-a-failure . I can expand and update it in response to questions. you can also help simply by eyeballing a failure and investigating it by reading code, and posting your findings here

might possibly have hit as-yet-unknown 2.13 regressions:

• 3 "http4s-parboiled2", // diverging implicit expansion -- could be a Scala regression, looks tricky to troubleshoot
• 1 "circe-jackson", // appears to have run afoul of recent (Feb? Mar?) changes to overloading resolution? could possibly be a Scala regression
• 1 "parboiled2", // same as http4s-parboiled2

I also found some projects that are only failing for trivial-looking reasons. we could submit upstream fixes, or just put fixes or workarounds in our forks, and see if that uncovers any real issues (or, in some cases, unblocks downstream stuff):

• 10 "monix" // compile errors because of new PartialFunction overload, should be easy workaround
• 3 "silencer", // test failure, maybe just a changed error message wording?
• 3 "scala-logging", // runs afoul of new rules about varargs overloading
• 2 "grizzled", // M5 compat is in place, but there are collections-related compile errors, maybe just needs post-M5 tweaks?
• 0 "scalajson", // looks like it might be hitting Som's change where import shadows local identifiers?
• 0 "scallop", // trivial post-M5 remove vs removed fix needed

and then this is a dilemma:

• 13 "scalapb", // dilemma: on master they have M5 support, but they also require Fastparse 2. what to do?

fixed

• 9 "scopt", // no arguments allowed for nullary method thank you @xuwei-k
• 2 "specs2-more", // overly sensitive test thank you @etorreborre
• 1 "base64", // needs workaround for Can't call scala.StringBuilder#length_= in 2.13.0-M5 #11158 which won't be fixed in RC1 thank you @xuwei-k and @marklister
• 1 "scalariform", // looks like a trivial compile error perhaps related to removal of foo bar () syntax thank you @Philippus and @godenji
• 0 scala-parallel-collections thank you @eed3si9n
• 0 scala-collection-laws scala/community-build@53ff214
• 0 "twotails", // "no arguments allowed for nullary method" thank you @Philippus
• 0 "scodec", // we just needed to use the branch where 2.13 support is present
• 0 "airframe", // serialization related test failures, thank you @xerial

not green, but investigated, and no unreported Scala regression seems present

• 0 scala-js scala/community-build@e5c65e2 thank you @adriaanm
• 0 "kittens", // hashing-related test failure thank you @joroKr21
• 0 "scala-swing", // looks like it needs tweaks for post-M5 changes thank you @Sciss

[SethTisue]

we're at 94 green projects

removed the "blocker" label, as I believe that except for scala-java8-compat, we're in good enough shape here for RC1, though work here will continue up through the 2.13.0 release and beyond

tracking scala-java8-compat-in-the-community-build separately at #11452

[Philippus]

@SethTisue I was looking at "scala-logging", I think the specs could be fixed with some clever annotations or asInstanceOf-stuff.
But I think it requires a nice proper fix, as it will probably hit real end-users of "scala-logging" eventually.
Maybe open an issue on the project?

[SethTisue]

@Philippus yes please :-)

(I would eventually get to it myself, but not this week)

[SethTisue]

3 "http4s-parboiled2", // diverging implicit expansion -- could be a Scala regression, looks tricky to troubleshoot
1 "parboiled2", // same as http4s-parboiled2

this is not eyeball-able, too hairy to troubleshoot ourselves, let's not hold the release up over it

1 "circe-jackson", // appears to have run afoul of recent (Feb? Mar?) changes to overloading resolution? could possibly be a Scala regression

also hairy, very likely fixable in a minor version even if turns out to be Scala's fault, ditto and not holding things up

[SethTisue]

there have been 4 regressions:

• unexpected FAILED: scala-xml — fixed by @adriaanm (https://github.com/scalacommunitybuild/scala-xml/tree/community-build-2.13)
• unexpected FAILED: scala-parallel-collections — fixed by @adriaanm (scala/scala-parallel-collections@3647df7)
• unexpected FAILED: jackson-module-scala (https://github.com/scala/scala-dev/issues/623) — investigated by @adriaanm, not a release blocker
• unexpected FAILED: spire — "fixed" by @adriaanm (scalacommunitybuild/spire@366f435)

Adriaan and team are looking into the remaining two

[sjrd]

scala-js should be back on track, now. Let me know if something breaks again.

[SethTisue]

https://scala-ci.typesafe.com/job/scala-2.13.x-integrate-community-build/1923/ running on 2.13.0-pre-63769af is SUCCESS 94 UNEXPECTED 0, so we're good to go again for RC1, knock on wood

@sjrd will do (after I unfreeze all the project SHAs after RC1 is out)

[SethTisue]

Adriaan suggested testing scala-js now rather than waiting, so I pushed scala/community-build@3306bd8 and to get fast feedback did a local, scala-js-only run. there were 2 test failures: https://gist.github.com/SethTisue/d83b051d0be8890346ad8612122aa245

the tests are testing error messages, and it looks plausible to me that the additional errors the compiler are now generating are probably reasonable errors coming from @adriaanm's recent rechecks changes; or if they do indicate that refchecks is now a bit overzealous in a few cases, it's probably not a release blocker? /cc @sjrd

[sjrd]

Oh I had forgotten that the community build performed the compiler/test. Yeah, those were broken ~2 weeks ago, and I haven't had time to investigate them. I don't consider them a blocker, but I can have a look today.

[adriaanm]

Yeah, the error messages (I assume) are related to refchecks, which now checks all pairs in the override chain.

[sjrd]

OK, I can now confirm that the new error messages are progressions.

[SethTisue]

• ticket on circe-jackson failure: compilation failure in Scala 2.13 community build circe/circe-jackson#54
• ticket on http4s failure: 2.13.0-RC1: new diverging expansion in DSLTest sirthias/parboiled2#200

[SethTisue]

I updated https://github.com/scala/community-builds/blob/2.13.x/report/Report.scala a day or two ago to include all the failure reasons (to the extent they're known)