Unnecessary symbol occurrence for new on compound types

[tanishiking]

Looks like semanticdb randomly(?) extract unnecessary symbol occurrence of new literal for compound types.

scalameta/tests/jvm/src/test/resources/example/Types.scala

Lines 58 to 60 in 83a261a

	val compoundType4/*<=types.Test.C#compoundType4.*/ = new /*<=local2*/{ def k/*<=local3*/: Int/*=>scala.Int#*/ = ???/*=>scala.Predef.`???`().*/ }
	val compoundType5/*<=types.Test.C#compoundType5.*/ = new /*<=local4*/M/*=>types.Test.M#*/ with N/*=>types.Test.N#*/
	val compoundType6/*<=types.Test.C#compoundType6.*/ = new /*<=local5*/M/*=>types.Test.M#*/ with N/*=>types.Test.N#*/ { def k/*<=local6*/: Int/*=>scala.Int#*/ = ???/*=>scala.Predef.`???`().*/ }

see: #2457 (comment)

[kitbellew]

@tanishiking @tgodzik what's expected here? the /*<=localX*/ notation not added to new Somethingvsnew { .. }, or new Somethingvsnew Something { ... }`?

[tgodzik]

We should not het any of those local symbols local2, local4 and local5 should not be added as occurrences

[kitbellew]

We should not het any of those local symbols local2, local4 and local5 should not be added as occurrences

does that mean you propose to undo #1902 completely? or partially (if so, when)?

[tgodzik]

Och wow, I didn't realize it was me who changed that 😅 I need to figure out why we needed those ocurrences.

[tgodzik]

Och ok, I think this was used to have a position to show the symbol for the implementation of an abstract class. Looks like this should be in the synthetics instead and that would serve us well.