Attempt to revert 14026

[odersky]

This is an attempt to partially revert #14026. It has lots of failures. We need to find someone to complete the job or else do a total revert of #14026. #14026 is important to have, but I fear we cannot accept all the regressions it causes for code that should compile.

So I think it is better to back out now, and try to put it back in the 3.2 series.

Fixes #14494
Fixes #15178
Fixes #15184
Fixes #15216

[smarter]

A quick look at the test failures:

tests/neg/i12640.scala failed

Apparently that test is flaky in the number of errors it outputs, it was removed once before (#12720) but added back recently (#14873).

tests/neg/i8900.scala failed

That's expected since this is a soundness bug that #14026 fixed.

tests/run/i8861.scala failed

Also expected, that test was added in #14026.

tests/pos/i15174.scala failed

Pickling difference, so probably harmless? (edit: the difference is likely caused by a type avoidance bug, so it makes sense for it to be broken by this reversal)

tests/run-macros/i7519c failed

That one is weird:

-- Error: tests/run-macros/i7519c/Test_2.scala:3:17 --------------------------------------------------------------------
3 |    println(quote[Int])
  |            ^^^^^^^^^^
  |            Exception occurred while executing macro expansion.
  |            java.lang.ArrayIndexOutOfBoundsException: Index 1 out of bounds for length 1
  |            	at scala.collection.immutable.ArraySeq$ofRef.apply(ArraySeq.scala:331)
  |            	at dotty.tools.dotc.quoted.PickledQuotes$.$anonfun$1(PickledQuotes.scala:176)
  |            	at scala.collection.Iterator$$anon$9.next(Iterator.scala:577)
  |            	at scala.collection.mutable.Growable.addAll(Growable.scala:62)
  |            	at scala.collection.mutable.Growable.addAll$(Growable.scala:57)
  |            	at scala.collection.immutable.MapBuilderImpl.addAll(Map.scala:692)
  |            	at scala.collection.immutable.Map$.from(Map.scala:643)
  |            	at scala.collection.IterableOnceOps.toMap(IterableOnce.scala:1256)
  |            	at scala.collection.IterableOnceOps.toMap$(IterableOnce.scala:1255)
  |            	at scala.collection.AbstractIterator.toMap(Iterator.scala:1293)
  |            	at dotty.tools.dotc.quoted.PickledQuotes$.spliceTypes(PickledQuotes.scala:178)
  |            	at dotty.tools.dotc.quoted.PickledQuotes$.unpickleTerm(PickledQuotes.scala:86)
  |            	at scala.quoted.runtime.impl.QuotesImpl.unpickleExprV2(QuotesImpl.scala:3044)
  |            	at Macro_1$package$.quoteImpl(Macro_1.scala:12)
  |
  |---------------------------------------------------------------------------------------------------------------------
  |Inline stack trace
  |- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  |This location contains code that was inlined from Macro_1.scala:8
8 |inline def quote[T]: String = ${ quoteImpl[T] }
  |                              ^^^^^^^^^^^^^^^^^
   ---------------------------------------------------------------------------------------------------------------------

This test hasn't been touched in a long time, but this might be related to the quote refactoring /cc @nicolasstucki.

[odersky]

I also put the two reverted follow-on commits under flag Config.checkLevels and moved the failing tests to pending.

It would be good to clarify why i7519c fails without level checking. /cc @nicolasstucki

[odersky]

So, here's the reversal. It should be easy to re-enable this by

• setting Config.checkLevels = true
• re-instantiating the tests moved to pending in 44c2a3b
• Making sure the new tests still pass

[smarter]

LGTM except for the mysterious macro test error.

[odersky]

I discussed with @nicolasstucki. Nothing definite comes to mind, and it does not seem urgent to address i7519c. So I think we can merge and that way we have addressed all regressions so far.