Warning summary respects -nowarn

[som-snytt]

Previously, hasWarning meant a warning had been displayed
(as opposed to counted). Since last refactor, it means
counted, so warnings were summarized under -nowarn.
Until 2.13.1, there was no summary under -nowarn.

Let -nowarn mean -Xmaxwarns 0 (to mean show me no warnings)
and additionally don't summarize any warnings (including
erroring under -Werror).

Fixes scala/bug#11952

[lrytz]

Some tests failed because summaries no longer show in repl tests

[som-snytt]

Edit: fixed just the REPL output, which was due to withSuppressedWarnings also tamping maxwarns but not restoring it.

[som-snytt]

It still says 1 warning, which is better than 9.

[info] Non-compiled module 'compiler-bridge_2.13' for Scala 2.13.3-20200429-100920-dd427d7. Compiling...
1 warning

[som-snytt]

@lrytz That leaking warning is because reportSuspendedMessages (which sets suppressionsComplete) is at end of typer phase, so a warning from patmat (if not "suppressed") is issued immediately. Why does that happen after typer and not after the run, at summarize time?

The division of state and responsibility between PerRunReporting and the Reporter is confusing to me.

Probably just make ConsoleReporter not issue summary under -nowarn, then the warning count is irrelevant.

[som-snytt]

Also picks up a commit to check that compiler options to a partest DirectTest are kosher.

Noticed while linting run tests; maybe there is a neg test that intentionally tests bad args?

[lrytz]

That leaking warning is because reportSuspendedMessages (which sets suppressionsComplete) is at end of typer phase, so a warning from patmat (if not "suppressed") is issued immediately.

I don't really follow what happens here. suppressionsComplete is about @nowarn annotations within source code, which we only collect during typer. So to avoid displaying warnings that would be hidden by a @nowarn, all warnings are buffered until after typer.

All that should not have anything to do with settins.nowarn. Are you saying that some patmat warning is getting issued in 2.13.2 under -nowarn?

The division of state and responsibility between PerRunReporting and the Reporter is confusing to me.

I think the division exists because of being within or outside the global cake. But I see the point: it would probably be good to move everything to Reporter that doesn't need access to Symbol, and make RunReporting a few convenience methods.

[som-snytt]

Yes, patmat issues warning, debug stack trace:

        at scala.reflect.internal.Reporter.increment(Reporting.scala:121)
        at scala.reflect.internal.Reporter.filteredInfo(Reporting.scala:115)
        at scala.reflect.internal.Reporter.warning(Reporting.scala:110)
        at scala.tools.nsc.Reporting$PerRunReporting.issueWarning(Reporting.scala:116)
        at scala.tools.nsc.Reporting$PerRunReporting.checkSuppressedAndIssue(Reporting.scala:130)
        at scala.tools.nsc.Reporting$PerRunReporting.warning(Reporting.scala:251)
        at scala.tools.nsc.Reporting$PerRunReporting.warning(Reporting.scala:255)
        at scala.tools.nsc.typechecker.Contexts$ContextReporter.warning(Contexts.scala:1638)
        at scala.tools.nsc.typechecker.Contexts$Context.warning(Contexts.scala:813)
        at scala.tools.nsc.transform.patmat.Interface$MatchMonadInterface.reportMissingCases(PatternMatching.scala:196)

It is "counted", so ConsoleReporter on finish summarizes it.

This PR adds two checks for -nowarn at PerRunReporter.summarize and also ConsoleReporter. This doesn't seem terrible to me, as the semantics are simply, "Don't print warning-related messages, including summaries."

But if -nowarn sets -Xmaxwarns 0 why are there any warnings? Because maxwarns only tells reporters.Reporter to cap the warnings that are displayed; the count is "upstream" in internal.Reporter.

As a footnote, not only cake per se but the split due to reflect is ongoing confusion. By confusion, I mean I couldn't understand the whole picture as a casual contributor.

[som-snytt]

[error] Caused by: sbt.ForkMain$ForkError: java.io.IOException: No space left on device

[som-snytt]

/rebuild

[lrytz]

Thank you for giving so much of your patience and time budgeds!